Modal Assurance Criterion - Álvaro Rivero

DINÁMICA
Y
FATIGA
“The Modal Assurance Criterion (MAC)”
Profesor: R. Avilés
Asignatura: Dinámica y Fatiga
Alumnos: Rivero González, Álvaro
18 de Diciembre de 2009
Modal Assurance Criterion (MAC) 2
Index
Abstract 4
1. State of the Art. Antecedents. review of existing methods 5
1.1. Historical Development of MAC 5
1.1.1. Modal Vector Orthogonality 6
1.1.2. Modal Vector Consistency 8
1.1.3. Modal Assurance Criterion (MAC) Zero 10
1.1.4. Modal Assurance Criterion (MAC) Unity 11
1.1.5. MAC Presentation Formats 12
1.2. Other Similar Assurance Criteria 14
1.3. Uses of the Modal Assurance Criterion 17
1.4. Abuses of the Modal Assurance Criterion 18
1.5. Current Developments 20
2. Software 21
2.1. LMS Virtual.Lab Overview 21
2.2. LMS Virtual.Lab Correlation 25
2.2.1. Systematic Validation from the bottom up 26
2.2.2. LMS Virtual.Lab Correlation. Features and Benefits 27
2.2.3. LMS Virtual.Lab Model Updating. Features and Benefits 29

3. Uses for Aircraft Design and Testing/Certificating Companies 31
3.1. Using MSC/NASTRAN and LMS/PRETEST to find an

optimal sensor placement for modal identification and
correlation of aerospace structures 31
3.1.1. MSC/NASTRAN and LMS/PRETEST 33
3.1.2. Target Mode Selection 33
3.1.3. Sensor Placement 37
3.1.4. Shaker Positioning 41
3.2. Modal Test of L-610G Aeroplane 43
Conclusions 47
Bibliography 48
Abstract
This report firstly reviews the development of the original modal assurance criterion
(MAC) together with other related assurance criteria that have been proposed over the last twenty
years. Some of the other assurance criteria that will be discussed include the coordinate modal
assurance criterion (COMAC), the frequency response assurance criterion (FMAC), partial mode
assurance criterion (PMAC) and modal assurance criterion using reciprocal modal vectors (MACRV).
Several uses of MAC that may not be obvious to the casual observer will be identified; the common
problems with the implementation and use of modal assurance criterion computations will also be
identified. Afterward, the LMS Virtual.Lab is briefly presented, stressing the module related to the
modal correlation, the LMS Virtual.Lab Correlation. Finally, this dossier introduces some particular
uses of the model assurance criterion that could be interesting for aircraft design and
testing/certificating companies.
Este informe revisa en primer lugar el desarrollo del criterio de confianza modal (MAC)
junto con otros criterios de confianza o correlación relacionados que han sido propuestos a lo largo de
los últimos veinte años. Otros criterios de correlación que serán tratados son el criterio de correlación
modal por coordenadas (COMAC), el criterio de correlación de respuesta en frecuencia (FMAC), el
criterio de correlación de modo parcial (PMAC) y el criterio de correlación modal utilizando vectores
modales recíprocos (MACRV). Varios usos del MAC que pueden no ser obvios para el observador
casual serán identificados; los problemas habituales con la implementación y uso de computaciones
del criterio de confianza modal serán también identificados. Seguidamente, el LMS Virtual.Lab es
brevemente presentado, haciendo hincapié en el módulo relacionado con la correlación modal, el LMS
Virtual.Lab Correlation. Finalmente, este informe introduce algunos usos particulares del criterio de
correlación modal que podrían resultar interesantes para empresas de diseño y de prueba/certificación
de aeronaves.
1. State of the art. Antecedents. Review of existing methods
The development of the modal assurance criterion over twenty years ago has led to a
number of similar assurance criteria used in the area of experimental and analytical structural
dynamics. It is important to recognize the mathematical similarity of these varied criteria in order to be
certain that conclusions be correctly drawn from what is essentially a squared, linear regression
correlation coefficient. The modal assurance criterion is a statistical indicator, just like ordinary
coherence, which can be very powerful when used correctly but very misleading when used
incorrectly. This first section will first review the historical development of the modal assurance
criterion. Other similar assurance criteria will then be identified although the list is not intended to be
comprehensive. Typical uses of the modal assurance criterion will be discussed and finally, typical
abuses will be identified.
Before starting with the review of the methods, it is necessary to present the nomenclature
that will be used in the following paragraphs:
L = Number of matching pairs of modal vectors.

A* = Complex conjugate of A.
Ni = Number of inputs.
No = Number of outputs (assumed to be larger than Ni).
Ne = Number of experimental modal vectors.
Na = Number of analytical modal vectors.
𝐻𝑝𝑞 (𝑤) = Measured frequency response function.
𝐻𝑝𝑞 (𝑤) = Synthesized frequency response function.
𝜑𝑞𝑟 = Modal coefficient for degree-of-freedom q, mode r.
𝜑𝑝𝑞𝑟 = Modal coefficient for reference p, degree-of-freedom q, mode r.
𝜑 𝑇 = Transpose of {φ}.
𝜑 𝐻 = Complex conjugate transpose (Hermitian) of {φ}.
𝜑𝑟 = Modal vector for mode r.
𝜑𝑝𝑟 = Modal vector for reference p, mode r.
1.1. Historical Development of MAC
The historical development of the modal assurance criteria originated from the need for a
quality assurance indicator for experimental modal vectors that are estimated from measured
frequency response functions. The standard of the late 1970s, when the modal assurance criterion was
developed, was the orthogonality check. The orthogonality check, however, coupled errors in the
analytical model development, the reduction of the analytical model and the estimated modal vectors
into a single indicator and was, therefore, not always the best approach. Many times, an analytical
model was not available which renders the orthogonality check impractical.
The original development of the modal assurance criterion was modeled after the
development of the ordinary coherence calculation associated with computation of the frequency
response function. It is important to recognize that this least squares based form of linear regression
analysis yields an indicator that is most sensitive to the largest difference between comparative values
(minimizing the squared error) and results in a modal assurance criterion that is insensitive to small
changes and/or small magnitudes. In the original thought process, this was considered an advantage
since small modal coefficient values are often seriously biased by frequency response function (FRF)
measurements or modal parameter estimation errors.
In the internal development of the modal assurance criterion at the University of

Cincinnati, Structural Dynamics Research Lab (UCSDRL), a little modal assurance criterion (Little
MAC), a big modal assurance criterion (Big MAC) and a multiple modal assurance criterion (Multi-
MAC) were formulated as part of the original development. Little MAC and Multi-MAC were
primarily testing methods and are not discussed further here. The modal assurance criterion that
survives today is what was originally identified as Big MAC. Since the “Big Mac” acronym was
already in use at that time, MAC is the designation that has persisted.
1.1.1. Modal Vector Orthogonality
The primary method that has historically been used to validate an experimental modal
model is the weighted orthogonality check comparing measured modal vectors and an appropriately
sized (the size of the square weighting matrix must match the length and spatial dimension of the
modal vector) analytical mass or stiffness matrix (weighting matrix). Variations of this process include
using analytical modal vectors together with experimental modal vectors and the appropriately sized
mass or stiffness matrix. This latter comparison is normally referred to as a pseudo-orthogonality
check (POC).
In the traditional orthogonality check, the experimental modal vectors are used together
with a mass matrix, normally derived from a finite element model, to evaluate orthogonality of the
experimental modal vectors. In the pseudo-orthogonality check, the experimental modal vectors are
used together with a mass matrix, normally derived from a finite element model, and the analytical
modal vectors, normally derived from the same finite element model, to evaluate orthogonality
between the experimental and analytical modal vectors. The experimental and analytical modal vectors
are scaled so that the diagonal terms of the modal mass matrix are unity. With this form of scaling, the
off-diagonal values in the modal mass matrix are expected to be less than 0.1 (10 percent of the
diagonal terms).
Theoretically, for the case of proportional damping, each modal vector of a system will be
orthogonal to all other modal vectors of that system when weighted by the mass, stiffness or damping
matrix. In practice, these matrices are made available by way of a finite element analysis and normally
the mass matrix is considered to be the most accurate. For this reason, any further discussion of
orthogonality will be made with respect to mass matrix weighting. As a result, the orthogonality
relations can be stated as follows:
𝑇
For r ≠ s: 𝜑𝑟 𝑀 𝜑𝑠 = 𝑂 (1)
𝑇
For r = s: 𝜑𝑟 𝑀 𝜑𝑠 = 𝑀𝑟 (2)
Experimentally, the result of zero for the cross orthogonality calculations (r ≠ s, Eq. 1) can
rarely be achieved but values up to one tenth of the magnitude of the generalized mass of each mode
are considered to be acceptable. It is a common procedure to form the modal vectors into a normalized
set of mode shape vectors with respect to the mass matrix weighting. The accepted criterion in the
aerospace industry, where this confidence check is made most often, is for all of the generalized mass
terms to be unity and all cross orthogonality terms to be less than 0.1. Often, even under this criterion,
an attempt is made to adjust the modal vectors so that the cross orthogonality conditions are satisfied.
Note that, in general, experimental modal vectors are not always real-valued and Eqs. 1 and 2 are
developed based upon normal or real-valued modal vectors. This complication has to be resolved by a
process of real normalization of the measured modal vectors prior to utilizing Eqs. 1 and 2 or by
applying an equivalent procedure involving the state-space form of the weighted orthogonality
relationship.
In Eqs. 1 and 2, the mass matrix must be an No×No matrix corresponding to the
measurement locations on the structure. This means that the finite element mass matrix must be
modified from whatever size and distribution of grid locations required in the finite element analysis to
the No×No square matrix corresponding to the measurement locations. This normally involves some
sort of reduction algorithm as well as interpolation of grid locations to match the measurement
situation.
When Eq. 1 is not sufficiently satisfied, one (or more) of three situations may exist. First,
the modal vectors can be invalid. This can be due to measurement errors or problems with the modal
parameter estimation algorithms. This is a very common assumption and many times contributes to the
problem. Second, the mass matrix can be invalid. Since the mass matrix does not always represent the
actual physical properties of the system when it is built or assembled, this probably contributes
significantly to the problem. Third, the reduction of the mass matrix can be invalid. This can
certainly be a realistic problem and cause severe errors. The most obvious example of this situation
would be when a relatively large amount of mass is reduced to a measurement location that is highly
flexible, such as the center of an unsupported panel. In such a situation, the measurement location is
weighted very heavily in the orthogonality calculation of Eq. 2, but may represent only incidental
motion of the overall modal vector.
In all probability, all three situations contribute to the failure of orthogonality or pseudo-
orthogonality criteria on occasion. When the orthogonality conditions are not satisfied, this result does
not indicate where the problem originates. From an experimental point of view, it is important to try to
develop methods that indicate confidence that the modal vector is, or is not, part of the problem.
1.1.2. Modal Vector Consistency
Since the frequency response function matrix contains redundant information with respect
to a modal vector, the consistency of the estimate of the modal vector under varying conditions such as
excitation locations (references) or modal parameter estimation algorithms can be a valuable
confidence factor to be utilized in the process of evaluation of experimental modal vectors. The
common approach to estimation of modal vectors from frequency response functions is to measure
several complete rows or columns of the frequency response function matrix. The estimation of modal
vectors from this frequency response function matrix will be a function of the data used in the modal
parameter estimation algorithms and the specific modal parameter estimations algorithms used. If the
modal vectors are not well represented in the frequency response function matrix, the estimation of the
modal vector will contain potential bias and variance errors. In any case, the modal vectors will
contain potential variance errors.
Frequently, different subsets of the frequency response function matrix and/or different
modal parameter estimation algorithms are utilized to estimate separate, redundant modal vectors for
comparison purposes. In these cases, if different estimates of the same modal vectors are generated,
the modal vectors can be compared and contrasted through an evaluation that consists of the
calculation of a complex modal scale factor (relating two modal vectors) and a scalar modal assurance
criterion (measuring the consistency or linearity between two modal vectors).
The function of the modal scale factor (MSF) is to provide a means of normalizing all
estimates of the same modal vector, taking into account magnitude and phase differences. Once two
different modal vector estimates are scaled similarly, elements of each vector can be averaged (with or
without weighting), differenced or sorted to provide a best estimate of the modal vector or to provide
an indication of the type of error vector superimposed on the modal vector. In terms of modern,
multiple reference modal parameter estimation algorithms, the modal scale factor is a normalized
estimate of the modal participation factor between two references for a specific mode of vibration.
The function of the modal assurance criterion (MAC) is to provide a measure of

consistency (degree of linearity) between estimates of a modal vector. This provides an additional
confidence factor in the evaluation of a modal vector from different excitation (reference) locations or
different modal parameter estimation algorithms.
The modal scale factor and the modal assurance criterion also provide a method of easily
comparing estimates of modal vectors originating from different sources. The modal vectors from a
finite element analysis can be compared and contrasted with those determined experimentally as well
as modal vectors determined by way of different experimental or modal parameter estimation
methods. In this approach, methods can be compared and contrasted in order to evaluate the mutual
consistency of different procedures rather than estimating the modal vectors specifically. If an
analytical and an experimental vector are deemed consistent or similar, the analytical modal vector,
together with the modal scale factor, can be used to complete the experimental modal vector if some
degrees of freedom could not be measured.
The modal scale factor is defined, according to this approach, as follows:
𝑁0 ∗
𝑞 =1 𝜑 𝑐𝑞𝑟 𝜑 𝑑𝑞𝑟
𝑀𝑆𝐹𝑐𝑑𝑟 = 𝑁0 ∗
(3a)
𝑞=1 𝜑 𝑑𝑞𝑟 𝜑 𝑑𝑞𝑟
or:
{𝜑 }𝑇 {𝜑 ∗ }
𝑀𝑆𝐹𝑐𝑑𝑟 = {𝜑 𝑐𝑟 }𝑇 {𝜑𝑑𝑟
∗ } (3b)
𝑑𝑟 𝑑𝑟
Since the modal scale factor is a complex-valued scalar, this is also equivalent to:
{𝜑 }𝐻 {𝜑 ∗ }
𝑀𝑆𝐹𝑐𝑑𝑟 = {𝜑 𝑑𝑟 }𝐻 {𝜑 ∗𝑐𝑟 } (3c)
𝑑𝑟 𝑑𝑟
Eq. 3 implies that the modal vector d is the reference to which the modal vector c is
compared. In the general case, modal vector c can be considered to be made of two parts. The first part
is the part correlated with modal vector d. The second part is the part that is not correlated with modal
vector d and is made up of contamination from other modal vectors and any random contribution. This
error vector is considered to be noise. The modal assurance criterion is defined as a scalar constant
relating the degree of consistency (linearity) between one modal and another reference modal vector as
follows:
𝑁0 2
∗
𝑞 =1 𝜑 𝑐𝑞𝑟 𝜑 𝑑𝑞𝑟
𝑀𝐴𝐶𝑐𝑑𝑟 = 𝑁0 ∗ 𝑁0 ∗
(4a)
𝑞=1 𝜑 𝑐𝑞𝑟 𝜑 𝑐𝑞𝑟 𝑞 =1 𝜑 𝑑𝑞𝑟 𝜑 𝑑𝑞𝑟
or:
∗ 2
𝜑 𝑐𝑟 𝑇 𝜑 𝑑𝑟
𝑀𝐴𝐶𝑐𝑑𝑟 = ∗ ∗ (4b)
𝜑 𝑐𝑟 𝑇 𝜑 𝑐𝑟 𝜑 𝑑𝑟 𝑇 𝜑 𝑑𝑟
Since the modal assurance criterion is a real-valued scalar, this is also equivalent to:
2
𝜑 𝑑𝑟 𝐻 𝜑 𝑐𝑟
∗
𝑀𝐴𝐶𝑐𝑑𝑟 = (4c)
𝜑 𝑑𝑟 𝐻 𝜑 𝑑𝑟 𝜑 𝑐𝑟 𝐻 𝜑 𝑐𝑟
or:
𝜑 𝑑𝑟 𝐻 𝜑 𝑐𝑟 𝜑 𝑐𝑟 𝐻 𝜑 𝑑𝑟
𝑀𝐴𝐶𝑐𝑑𝑟 = (4d)
𝜑 𝑑𝑟 𝐻 𝜑 𝑑𝑟 𝜑 𝑐𝑟 𝐻 𝜑 𝑐𝑟
or:
𝑀𝐴𝐶𝑐𝑑𝑟 = 𝑀𝑆𝐹𝑑𝑟 𝑀𝑆𝐹𝑑𝑟𝑐 (4e)
The modal assurance criterion takes on values from zero – representing no consistent
correspondence, to one – representing a consistent correspondence. In this manner, if the modal
vectors under consideration truly exhibit a consistent, linear relationship, the modal assurance criterion
should approach unity and the value of the modal scale factor can be considered reasonable. Note that,
unlike the orthogonality calculations, the modal assurance criterion is normalized by the magnitude of
the vectors and, thus, is bounded between zero and one.
The modal assurance criterion can only indicate consistency, not validity or orthogonality.
If the same errors, random or bias, exist in all modal vector estimates, this is not delineated by the
modal assurance criterion. Invalid assumptions are normally the cause of this sort of potential error.
Even though the modal assurance criterion is unity, the assumptions involving the system or the modal
parameter estimation techniques are not necessarily correct. The assumptions may cause consistent
errors in all modal vectors under all test conditions verified by the modal assurance criterion.
1.1.3. Modal Assurance Criterion (MAC) Zero
If the modal assurance criterion has a value near zero, this is an indication that the modal
vectors are not consistent. This can be due to any of the following reasons:
 The system is nonstationary. This can occur if the system is nonlinear and two data sets have
been acquired at different times or excitation levels. System nonlinearities will appear
differently in frequency response functions generated from different exciter positions or
excitation signals. The modal parameter estimation algorithms will also not handle the different
nonlinear characteristics in a consistent manner.
 There is noise on the reference modal vector. This case is the same as noise on the input of a
frequency response function measurement. No amount of signal processing can remove this
type of error.
 The modal parameter estimation is invalid. The frequency response function measurements
may contain no errors but the modal parameter estimation may not be consistent with the data.
 The modal vectors are from linearly unrelated mode shape vectors. Hopefully, since the
different modal vector estimates are from different excitation positions, this measure of
inconsistency will imply that the modal vectors are orthogonal.
If the first four reasons can be eliminated, the modal assurance criterion can be interpreted
in a similar way as an orthogonality calculation.
1.1.4. Modal Assurance Criterion (MAC) Unity
If the modal assurance criterion has a value near unity, this is an indication that the modal
vectors are consistent. This does not necessarily mean that they are correct. The modal vectors can be
consistent for any of the following reasons:
 The modal vectors have been incompletely measured. This situation can occur whenever too
few response stations have been included in the experimental determination of the modal
vector.
 The modal vectors are the result of a forced excitation other than the desired input. This would
be the situation if, during the measurement of the frequency response function, a rotating piece
of equipment with an unbalance is present in the system being tested.
 The modal vectors are primarily coherent noise. Since the reference modal vector may be
arbitrarily chosen, this modal vector may not be one of the true modal vectors of the system. It
could simply be a random noise vector or a vector reflecting the bias in the modal parameter
estimation algorithm. In any case, the modal assurance criterion will only reflect a consistent
(linear) relationship to the reference modal vector.
 The modal vectors represent the same modal vector with different arbitrary scaling. If the two
modal vectors being compared have the same expected value when normalized, the two modal
vectors should differ only by the complex valued scale factor, which is a function of the
common modal coefficients between the rows or columns.
Therefore, if the first three reasons can be eliminated, the modal assurance criterion
indicates that the modal scale factor is the complex constant relating the modal vectors and that the
modal scale factor can be used to average, difference or sort the modal vectors.
Under the constraints mentioned previously, the modal assurance criterion can be applied
in many different ways. The modal assurance criterion can be used to verify or correlate an
experimental modal vector with respect to a theoretical modal vector (eigenvector). This can be done
by computing the modal assurance criterion between Ne modal vectors estimated from experimental
data and Na modal vectors estimated from a finite element analysis evaluated at common stations.
This process results in a Ne×Na rectangular modal assurance criterion matrix with values that
approach unity whenever an experimental modal vector and an analytical modal vector are
consistently related.
Once the modal assurance criterion establishes that two vectors represent the same
information, the vectors can be averaged, differenced or sorted to determine the best single estimate or
the potential source of contamination using the modal scale factor. Since the modal scale factor is a
complex scalar that allows two vectors to be phased the same and to the same mean value, these
vectors can be subtracted to evaluate whether the error is random or biased. If the error appears to be
random and the modal assurance criterion is high, the modal vectors can be averaged (using the modal
scale factor) to improve the estimate of a modal vector. If the error appears to be biased or skewed, the
error pattern often gives an indication that the error originates due to the location of the excitation or
due to an inadequate modal parameter estimation process. Based upon partial but overlapping
measurement of two columns of the frequency response function matrix, modal vectors can be sorted,
assuming the modal assurance function indicates consistency, into a complete estimate of each modal
vector at all measurement stations.
The modal assurance criterion can be used to evaluate modal parameter estimation
methods if a set of analytical frequency response functions with realistic levels of random and bias
errors is generated and used in common with a variety of modal parameter estimation methods. In this
way, agreement between existing methods can be established and new modal parameter estimation
methods can be checked for characteristics that are consistent with accepted procedures. Additionally,
this approach can be used to evaluate the characteristics of each modal parameter estimation method in
the presence of varying levels of random and bias error.
The concept of consistency in the estimate of modal vectors from separate testing
constraints is important considering the potential of multiple estimates of the same modal vector from
numerous input configurations and modal parameter estimation algorithms. The computation of modal
scale factor and modal assurance criterion results in a complex scalar and a correlation coefficient that
does not depend on weighting information outside the testing environment. Since the modal scale
factor and modal assurance criterion are computed analogous to the frequency response function and
coherence function, both the advantages and limitations of the computation procedure are well
understood. These characteristics, as well as others, provide a useful tool in the processing of
experimental modal vectors.
1.1.5. MAC Presentation Formats
One of the big changes in the application of the Modal Assurance Criteria over the last
twenty years is in the way the information is presented. Historically, a table of numbers was usually
presented as shown in Table 1.
Table 1.- Numerical presentation of MAC values.
Today, most computer systems routinely utilize color to present magnitude data like MAC
using a 2D or 3D plot as shown in Figures 1 and 2. It is important to remember, however, that MAC is
a discrete calculation and what appears as a color contour plot really only represents the discrete mode
to mode comparison. Nevertheless, a color plot does allow for more data to be presented in an
understandable form in a minimum space.
Figure 1.- 2-D and 3-D presentation of MAC values.

1.2. Other Similar Assurance Criteria
The following brief discussion highlights assurance criteria that utilize the same linear,
least squares computation approach to the analysis (projection) of two vector spaces as the modal
assurance criterion. The equations for each assurance criterion are not repeated unless there is a
significant computational difference that needs to be clarified or highlighted. This list is by no means
comprehensive nor is it in any particular order of importance but includes most of the frequently cited
assurance criterion found in the literature.
Weighted Modal Analysis Criterion (WMAC): A number of authors have utilized a weighted
modal assurance criterion (WMAC) without developing a special designation for this case. WMAC is
proposed for these cases. The purpose of the weighting matrix is to recognize that MAC is not
sensitive to mass or stiffness distribution, just sensor distribution, and to adjust the modal assurance
criterion to weight the degrees-of-freedom in the modal vectors accordingly. In this case, the WMAC
becomes a unity normalized orthogonality – or pseudo-orthogonality – check where the desirable
result for a set of modal vectors would be ones along the diagonal (same modal vectors) and zeros off-
diagonal (different modal vectors) regardless of the scaling of the individual modal vectors. Note that
the weighting matrix is applied as an inner matrix product for the single numerator vector product and
both vector products in the denominator.
Partial Modal analysis Criterion (PMAC): The partial modal assurance criterion (PMAC) was
developed as a spatially limited version of the modal assurance criterion where a subset of the
complete modal vector is used in the calculation. The subset is chosen based upon the user‟s interest
and may reflect only a certain dominant sensor direction (X, Y and/or Z) or only the degrees-of-
freedom from a component of the complete modal vector.
Modal Assurance Criterion Square Root (MACSR): The square root of the modal assurance
criterion (MACSR) is developed to be more consistent with the orthogonality and pseudo-
orthogonality calculations using an identity weighting matrix. Essentially this approach utilizes the
square root of the MAC calculation, which tends to highlight the cross terms (off diagonal) that are
generally very small MAC values.
Scaled Modal Assurance Criterion (SMAC): The scaled modal assurance criterion (SMAC) is
essentially a weighted modal assurance criterion (WMAC) where the weighting matrix is chosen to
balance the scaling of translational and rotational degrees-of-freedom included in the modal vectors.
This development is needed whenever different data types (with different engineering units) are
included in the same modal vector to normalize the magnitude differences in the vectors. This is
required since the modal assurance criterion minimizes the squared error and is dominated by the
larger values.
Modal Assurance Criterion Using Reciprocal Vectors (MACRV): A reciprocal modal vector is
defined as the mathematical vector that, when transposed and premultiplied by a specific modal
vector, yields unity. When the same computation is performed with this reciprocal modal vector and
any other modal vector or any other reciprocal modal vector, the result is zero. The reciprocal modal
vector can be thought of as a product of the modal vector and the unknown weighting matrix that will
produce a perfect orthogonality result. Reciprocal modal vectors are computed directly from measured
frequency response functions and the experimental modal vectors and are, therefore, experimentally
based.
The modal assurance criterion using reciprocal modal vectors (MACRV) is the comparison of
reciprocal modal vectors with analytical modal vectors in what is very similar to a pseudo-
orthogonality check (POC). The reciprocal modal vectors are utilized in controls applications as modal
filters and the MACRV serves as a check of the mode isolation provided by each reciprocal modal
vector compared to analytical modes expected.
Modal Assurance Criterion with Frequency Scales (FMAC): Another extension of the modal
assurance criterion is the addition of frequency scaling to the modal assurance criterion. This extension
of MAC “offers a means of displaying simultaneously the mode shape correlation, the degree of
spatial aliasing and the frequency comparison in a single plot.” This development is particularly useful
in model correlation applications (model updating, assessment of parameter variation, etc.).
Coordinate Modal Assurance Criterion (COMAC): An extension of the modal assurance

criterion is the coordinate modal assurance criterion (COMAC). The COMAC attempts to identify
which measurement degrees-of-freedom contribute negatively to a low value of MAC. The COMAC
is calculated over a set of mode pairs, analytical versus analytical, experimental versus experimental or
experimental versus analytical. The two modal vectors in each mode pair represent the same modal
vector, but the set of mode pairs represents all modes of interest in a given frequency range. For two
sets of modes that are to be compared, there will be a value of COMAC computed for each
(measurement) degree-of-freedom.
The coordinate modal assurance criterion (COMAC) is calculated using the following approach,
once the mode pairs have been identified with MAC or some other approach:
𝐿 2
𝑟=1 𝜑 𝑞𝑟 𝜙 𝑞𝑟
𝐶𝑂𝑀𝐴𝐶𝑞 = 𝐿 ∗ 𝐿 ∗ (5)
𝑟=1 𝜑 𝑞𝑟 𝜑 𝑞𝑟 𝑟=1 𝜑 𝑞𝑟 𝜑 𝑞𝑟
Note that the above formulation assumes that there is a match for every modal vector in the two
sets and the modal vectors are renumbered accordingly so that the matching modal vectors have the
same subscript. Only those modes that match between the two sets are included in the computation.
The Enhanced Coordinate Modal Assurance Criterion (ECOMAC): One common problem
with experimental modal vectors is the potential problem of calibration scaling errors and/or sensor
orientation mistakes. The enhanced coordinate modal assurance criterion (ECOMAC) was developed
to extend the COMAC computation to be more aware of typical experimental errors that occur in
defining modal vectors such as sensor scaling mistakes and sensor orientation (plus or minus sign)
errors.
Mutual Correspondence Criterion (MCC): The mutual correspondence criterion (MCC) is the
modal assurance criterion applied to vectors that do not originate as modal vectors but as vector
measures of acoustic information (velocity, pressure, intensity, etc.). The equation in this formulation
utilizes a transpose and will only correctly apply to real valued vectors.
Modal Correlation Coefficient (MCC): One of the natural limitations of a least squares based
correlation coefficient like the modal assurance criterion is that it is relatively insensitive to small
changes in magnitude, position by position, in the vector comparisons. The modal correlation
coefficient (MCC) is a modification of MAC that attempts to provide a more sensitive indicator. This
approach is particularly important when using modal vectors in damage detection situations where the
magnitude changes of the modal vectors being measured are minimal.
Inverse Modal Assurance Criterion (IMAC): An alternative approach to increasing the

sensitivity of the modal assurance criterion to small mode shape changes is the inverse modal
assurance criterion (IMAC). This approach uses essentially the same computational scheme as MAC
but utilizes the inverse of the modal coefficients. Therefore, small modal coefficients become
significant in the least squares based correlation coefficient computation. Naturally, this computation
suffers from the possibility that a modal coefficient could be numerically zero.
Frequency Response Assurance Criterion (FRAC): Any two frequency response functions
representing the same input-output relationship can be compared using a technique known as the
frequency response assurance criterion (FRAC). The simplest example is a validation procedure that
compares the FRF data synthesized from the modal model with the measured FRF data. The basic
assumption is that the measured frequency response function and the synthesized frequency response
function should be linearly related (unity scaling coefficient) at all frequencies. Naturally, the FRFs
can be compared over the full or partial frequency range of the FRFs as long as the same discrete
frequencies are used in the comparison. This approach has been utilized in the modal parameter
estimation process for a number of years under various designations (parameter estimation correlation
coefficient, synthesis correlation coefficient and response vector assurance criterion (RVAC)). This
procedure is particularly effective as a modal parameter estimation validation procedure if the
measured data were not part of the data used to estimate the modal parameters. This serves as an
independent check of the modal parameter estimation process.
𝑤2 2
∗ (𝑤)
𝑤 =𝑤 1 𝐻𝑝𝑞 𝑤 𝐻𝑝𝑞
𝐹𝑅𝐴𝐶𝑝𝑞 = 𝑤2 ∗ 𝑤2 ∗ (𝑤 ) (6)
𝑤 =𝑤 1 𝐻𝑝𝑞 𝑤 𝐻𝑝𝑞 (𝑤 ) 𝑤 =𝑤 1 𝐻𝑝𝑞 𝑤 𝐻𝑝𝑞
Complex Correlation Coefficient (CCF): A significant variation in the frequency response

assurance criterion is the complex correlation coefficient (CCF), which is computed without squaring
the numerator term, thus yielding a complex valued coefficient. The magnitude of the coefficient is the
same as the FRAC computation but the phase describes any systematic phase lag or lead that is present
between the two FRFs. In situations where analytical and experimental FRFs are compared, the CCF
will detect the common problem of a constant phase shift that might be due to experimental signal
conditioning problems, etc.
Frequency Domain Assurance Criterion (FDAC): A similar variation in the frequency

response assurance criterion is the frequency domain assurance criterion (FDAC), which is a FRAC-
type of calculation evaluated with different frequency shifts. Since the difference in impedance (FRF)
model updating is often an FRF that is in question due to frequencies of resonances or anti-resonances,
the FDAC is formulated to identify this problem. A related criterion, the modal FRF assurance
criterion (MFAC), combines analytical modal vectors with measured frequency response functions
(FRFs) in an extension of FRAC and FDAC that weights or filters the FRF data based upon the
expected, analytical modal vectors.
Coordinate Orthogonality Check (CORTHOG): The coordinate orthogonality check

(CORTHOG) is a normalized error measure between the pseudo-orthogonality calculation, comparing
measured to analytical modal vectors, and the analytical orthogonality calculation, comparing
analytical to analytical modal vectors. Several different normalizing or scaling methods are used with
this calculation.
1.3. Uses of the Modal Assurance Criterion
Most of the potential uses of the modal assurance criterion are well known but a few may
be more subtle. A partial list of the most typical uses that have been reported in the literature are as
follows:
 Validation of experimental modal models.
 Correlation with analytical modal models (mode pairing).
 Correlation with operating response vectors.
 Mapping matrix between analytical and experimental modal models.
 Modal vector error analysis.

 Modal vector averaging.
 Experimental modal vector completion and/or expansion.
 Weighting for model updating algorithms.
 Modal vector consistency/stability in modal parameter estimation algorithms.
 Repeated and pseudo-repeated root detection.
 Structural fault/damage detection.
 Quality control evaluations.
 Optimal sensor placement.
1.4. Abuses of the Modal Assurance Criterion
Many of the alternate formulations of the modal assurance criterion were developed to
address some of the shortcomings of the original modal assurance criterion formulation. When users
utilize the original modal assurance criterion in these situations, a poor result will often follow. For the
purposes of this discussion, this is referred to as misuse or abuse. The misuse or abuse of the modal
assurance criterion generally results due to one of five issues. These issues can be summarized as:
 The modal analysis criterion is not an orthogonality check.
 The wrong mathematical formulation for the modal assurance criterion is used.
 The modal assurance criterion is sensitive to large values (wild points?) and insensitive to
small values.
 The number of elements in the modal vectors (space) is small.
 The modal vectors have been zero padded.
These issues can be further explained in the following paragraphs.

The modal analysis criterion is not an orthogonality check: It is important to recognize that the
modal assurance criterion effectively weights the computation based upon the spatial distribution of
the degrees-of-freedom included in the modal vectors. The modal assurance criterion does not weight
the modal vectors with a mass or stiffness matrix and, therefore, cannot compensate for situations
where a very limited number of degrees-of-freedom (sensors) have been placed on a massive sub-
structure of a mechanical system. The typical example involves the engine of an automobile. If few or
no sensors are placed on the engine and a large number are placed on the surface of the automobile
body, several modal vectors at different modal frequencies will have very high MAC numbers
indicating that the modal vectors are the same. This example indicates to the user that an incomplete
modal vector was measured and the user has violated one of the primary assumptions of experimental
modal analysis (observability).
The wrong mathematical formulation for the modal assurance criterion is used: Frequently,
users implement the modal assurance criterion, or a related similar computation, using a vector
transpose in the numerator and denominator calculations rather than an Hermitian (conjugate
transpose). This error causes no problem as long as analytical vectors or real-valued experimental
vectors are involved in the calculation. However, in the general case, where some of the vectors are
complex-valued, this does not give the correct result. The original mathematical formulation assumes
the general case but has been reported incorrectly in some literature. This innocent error often occurs
when the author is utilizing real-valued vectors and notices no problem. However, users who do not
recognize this issue are often led astray in subsequent applications involving complex-valued vectors.
The modal assurance criterion is sensitive to large values (wild points?) and insensitive to
small values: The modal assurance criterion is based upon the minimization of the squared error
between two vector spaces. This means that the degrees-of-freedom involving the largest magnitude
differences between the two modal vectors will dominate the computation while small differences will
have almost no effect. Therefore, nodal information (small modal coefficients) will generally not have
much effect on the MAC calculation and large modal coefficients will potentially have the greatest
effect. This also means that, if there have been erroneous data included in the modal vectors due to
calibration errors, modal parameter estimation mistakes, etc., these wild points may dominate the
MAC calculation.
The number of elements in the modal vectors (space) is small: Since the modal assurance
criterion is essentially a statistical computation where the number of averages comes from the number
of elements in the modal vectors, if the modal vectors have only a limited number of degrees-of-
freedom, this will skew the meaning of the numerical MAC value. This frequently happens when high
order, multiple reference modal parameter estimation algorithms estimate the stability or consistency
diagram. Modal vector stability or consistency is identified using a MAC computation where the
vectors include only the degrees-of-freedom at the reference locations, typically two to five. In these
situations, there may be great variability in the MAC computation, particularly if the modal vector is
not well excited from one or more of the reference locations. Vectors with many elements reduce the
sensitivity of MAC to this problem.
The modal vectors have been zero padded. Frequently, when modal vectors are exported from
one computational environment to another, the modal vectors include zero values when no value was
ever measured, or computed, for that degree-of-freedom. For example, in an experimental situation,
one (X) or two dimensions (X,Y) of translational response may be measured at some degrees-of-
freedom rather than three dimensions (X,Y,Z). In the commonly used Universal File Format for modal
vectors (File Format 55), this is the case since there is no designation for not measuring the
information. When the modal assurance criterion is calculated for this case, there will be a problem if
some other vectors, with nonzero information at these degrees-of-freedom, are included in the
computation. This can be avoided if information is dropped from the computation when either vector
includes a perfect zero (within computational precision) at a degree-of-freedom, but is rarely done.
1.5. Current Developments
Currently, many users are utilizing more statistical approaches to understand the meaning
and bounds of experimental modal parameters. This approach extends to the modal assurance criterion
as well. Examples are the bootstrap and jackknife approaches to the evaluation of the mean and
standard deviation of discrete sets of experimental data. These approaches remove and/or replace
portions of the computation (bootstrap uses replicative resampling, jackknife uses sequential
elimination) to evaluate the bounds or limits on the MAC values. In this way, the sensitivity of the
MAC computation can be more effectively evaluated than with the current single number indicating
the degree of linearity between two modal vectors that are being compared.
2. Software:
One of the most used computer program, and recommended by the Head of Dynamics
Programs of the CTA, is the LMS Virtual.Lab. LMS Virtual.Lab offers an integrated software suite
to simulate and optimize the performance of mechanical systems for structural integrity, noise and
vibration, system dynamics and durability. LMS Virtual.Lab covers all the process steps and required
technologies to perform an end-to-end design assessment in each key discipline. Using LMS
Virtual.Lab, engineering teams can build accurate simulation models, simulate their real-life
performance, quickly assess multiple design alternatives and optimize designs before prototype
construction.
2.1. LMS Virtual.Lab Overview
LMS Virtual.Lab Desktop: LMS Virtual.Lab Desktop provides a common environment for
multiple functional performance applications. With LMS Virtual.Lab Desktop, users have seamless
access to models and load data, geometry and simulation models from industry-standard CAD and
CAE tools as well as test data. LMS Virtual.Lab Desktop also offers a complete visualization
environment for part and assembly models, functional performance engineering data, time and
frequency functions and much more.
Figure 2.- LMS Virtual.Lab Desktop.

LMS Virtual.Lab Structures: LMS Virtual.Lab Structures offers a scalable solution for
structural modeling and analysis, integrating advanced model creation and manipulation tools to
efficiently generate component, subsystem and full-system models. LMS Virtual.Lab Structures offers
full meshing capabilities and captures the complete modeling and analysis process from CAD drawing
to multiattribute simulation results. It offers multi-solver support for Abaqus, Ansys, CATIA CAE and
Nastran (MD, MSC, NX, NEi).
Figure 3.- LMS Virtual.Lab Structures.
LMS Virtual.Lab Motion: LMS Virtual.Lab Motion offers a highly efficient, completely
integrated solution to build multibody models that simulate the full-motion behavior of complex
mechanical system designs. Users can easily create a complete and accurate system model from
scratch or import geometry models from any industry-standard CAD system. LMS Virtual.Lab Motion
applies forces and motion to simulate the actual operational behavior of the new design. The resulting
simulation is excellent input to optimize the design‟s dynamic performance. The resulting loads can
also be used for structural analysis, durability, and noise and vibration studies.
Figure 4.- LMS Virtual.Lab Motion.

LMS Virtual.Lab Acoustics: LMS Virtual.Lab Acoustics offers an integrated solution to

minimize radiated noise or optimize the sound quality in new designs. Convenient modeling
capabilities combined with efficient solvers and easy-to-interpret visualization tools enable users to
quickly gain insight to the acoustic performance of their product. LMS Virtual.Lab Acoustics
simulates both internal and external acoustic radiation and offers dedicated applications for structural
noise radiation, engine acoustics, transmission loss through panels, aero-acoustic phenomena and
much more.
Figure 5.- LMS Virtual.Lab Acoustics.
LMS Virtual.Lab Noise and Vibrations: LMS Virtual.Lab Noise and Vibration is developed to
efficiently analyze, refine and optimize the vibro-acoustic behavior of a design. It offers all the
required tools to create system-level models, build realistic load cases and simulate noise and vibration
responses. It includes a wide range of visualization and analysis tools to analyze noise and vibration
performance and accurately pinpoint the most critical contributors to noise and vibration issues.
Convenient tools enable engineers to quickly perform design modifications and assess the noise and
vibration performance of a design variant in minutes.
Figure 6.- LMS Virtual.Lab Noise and Vibrations.

LMS Virtual.Lab Correlation: LMS Virtual.Lab Correlation allows users to combine test-based
and virtual component models into system-level models for more productive simulation. It offers
direct access to standard FE and test data formats and a unique export to LMS Test.Lab. LMS
Virtual.Lab quickly compares and validates FE models to test data and identifies possible modeling
errors to systematically improve existing simulation models.
Figure 7.- LMS Virtual.Lab Correlation.
LMS Virtual.Lab Durability: LMS Virtual.Lab Durability allows engineers to design reliable
products right from the start. It predicts fatigue hotspots and system-level fatigue life by combining
dynamic component loads with stress results and fatigue material parameters. LMS Virtual.Lab
Durability provides direct feedback regarding critical fatigue areas and the root cause of fatigue
problems. This immediate insight enables engineering teams to validate more design variants for
fatigue life within ever-shorter development cycles.
Figure 8.- LMS Virtual.Lab Durability.

LMS Virtual.Lab Optimization: LMS Virtual.Lab Optimization lets design and engineering
teams automatically select the optimal design while accounting for multiple performance targets.
Users can easily identify the key variables that have the most influence on the functional performance
of a mechanical system. LMS Virtual.Lab Optimization automatically explores a multitude of design
alternatives using design of experiment and response surface modeling techniques. It also analyzes
design robustness and reliability according to Design for Six Sigma criteria.
Figure 9.- LMS Virtual.Lab Optimization.
2.2. LMS Virtual.Lab Correlation
To guarantee realistic high fidelity simulations, it is essential that simulation models meet
stringent accuracy standards. Ensuring reliable simulation results requires component, subsystem and
full-system models to be compared with experimental data, or alternatively validated models of similar
structures. Building and validating system models from the bottom up is the only way to prevent
accumulating inaccuracies. Besides more reliable what-if analyses, validated models provide a better
understanding of assumptions made regarding material properties, connections, joints and boundary
conditions.
Correlating structural characteristics: Although static physical tests serve many design
purposes, models used for vibro-acoustic simulations usually require systematic test-based validation
of dynamic properties. LMS Virtual.Lab Correlation helps correlate physical test results and prepare
structural tests. A comprehensive tool set significantly facilitates simulated and measured mode shape
comparison and operational deflection shapes and response functions. Using the original FE model as
a basis to provide optimal comparison positions, and the required number of excitation and response
points helps avoid testing errors and redundancy.
Validation-driven model updating: Deducing model improvements using validation output is

not always obvious. To facilitate model updating driven by validations, LMS Virtual.Lab Correlation
offers specialized features to identify specific locations that need improvement. For example, it runs
sensitivity analyses that efficiently retain the most influential specified parameters. Users can also
automatically update models using internal and external algorithms, such as Nastran Solution 200,
which focuses on tuning modal frequencies and response functions.
2.2.1. Systematic validation from the bottom up
Figure 10.- Systematic Validation from the bottom up.
Pre-Test: When preparing measurements for physical structures, one can use modal information
of preliminary Finite Element models to define the optimal measurement set-up. For a modal test set-
up, this means defining a set of measuring points and excitation points. LMS Virtual.Lab Correlation
provides tools to quickly carry out this pre-test analysis in a user-friendly way. The objective is to
obtain a measurement set-up that guarantees high quality measurement data.
Correlation: Once good test data for the physical model is available, LMS Virtual.Lab
Correlation allows its users to quantify the geometrical and dynamic (FRF and Modal) resemblance
between the test model and its FE equivalent model. Several correlation metrics, like MAC and
FRAC, are available to study (mode) shape or frequency response function correlation interactively.
Specialized algorithms and post-processing tools allow to localize the problem locations of bad shape
correlation and give insight in stiffness differences between the two models.
Sensitivity and updating: After the dynamic correlation between two models has been
quantified, LMS Virtual.Lab allows users to easily setup and drive Nastran Sol200 to obtain the
sensitivity of FE dynamic properties towards a set of design parameters to decide which parameters to
change to obtain better correlation results. Using Sol200 sensitivities, MAC and frequency difference
sensitivities are derived to use for modal updating. A broad range of sensitivities can also be computed
using LMS Virtual.Lab Optimization: the user can define a variety of dynamic properties to optimize
for a very broad range of design parameters. Once the set-up of design parameters (inputs) and
correlation metrics (outputs) is in place, LMS Virtual.Lab offers the possibilities to carry out Design of
Experiments, Response Surface Modeling and Updating with several local and global optimization
algorithms.
2.2.2. LMS Virtual.Lab Correlation. Features and Benefits
LMS Virtual.Lab Correlation offers tools to ensure that high-quality FE models are used in
a CAE environment and that correct sensor and excitation locations are employed in a dynamic
physical structure test environment.
For pre-test analysis, users can create an optimal test geometry from an existing FE model.
LMS Virtual.Lab Correlation interactively creates a test wireframe on top of the FE mesh and directly
quantifies its quality according to relevant mode capturing and modal excitation. In case of poor
sensor location set quality, LMS Virtual.Lab Correlation provides an easy way to analyze why the
model was off-target. Users can easily change the test geometry and directly assess new quality levels
using the MAC (Modal Assurance Criterion). The DPR (Driving Point Residue) criterion is used for
the excitation point set.
LMS Virtual.Lab Correlation also lets users easily and quickly compare the dynamic
behavior of two models and deal with incompatible meshes (test and/or FE). It helps users to
quantitatively articulate the degree of shape correlation using a MAC matrix. If the MAC values are
too low to subjectively correlate the modes, the MAC Contribution (MACCo) criterion points out the
differences to be examined. In this way, users can verify different modeling assumptions by
comparing reference or measurement data. This improves model and simulation reliability. An
orthogonality check between two models adds a degree of correlation accuracy by using the mass
matrix to compare system dynamics. For this, LMS Virtual.Lab Correlation sets up the Nastran DMIG
Solution to obtain reduced system mass matrices required for orthogonality checks between test and
FE modes. The FRAC (Frequency Response Assurance Criterion) compares transfer functions
between two models and provides information about global stiffness and mass modeling errors.
Features:
 Universal access to test and FE data for models, modes and frequency spectra.
 Modal Assurance Criterion (MAC) and MAC Contribution (MACCo) support error
localization.
 Visual Shape correlation for side-by-side model animation (FE or Test).

 Frequency Response Assurance Criterion (FRAC).
 Orthogonality check for better dynamic correlation.
 Driving Point Residue (DPR) for shaker location identification.
 Export data to LMS Test.Lab or a universal file format.
Benefits:
 Maximum test information with minimized excitation and measurement locations.
 Increase measurement productivity with direct LMS Test.Lab integration.
 Confirm FE simulation model validity using measurements.
 Identify modeling errors or evaluate modeling strategies.
 Improve simulation model reliability.
Figure 11.- LMS Virtual.Lab Correlation uses.

2.2.3. LMS Virtual.Lab Model Updating. Features and Benefits
LMS Virtual.Lab Model Updating is a model correlation and updating tool that improves
simulation model quality based on reference data. With LMS Virtual.Lab Model Updating, analysts
can make models that match reality more closely. FE models are first correlated with reference
models, which are typically test models, but can also be FE models. The next step is to compute
dynamic property sensitivity with respect to design parameter uncertainties. This can be done by
inserting a Nastran Sol200 case from LMS Virtual.Lab Desktop. In this way, users can easily define
element group properties for sensitivity analysis, including material and property data. Dynamic
targets can be the total system mass, a specific eigenfrequency that is poorly correlated, vibration
levels for unit load conditions or mode shapes. Sensitivity information is then used to update or
optimize the Nastran model to match real-life condition better. For non-Nastran users, the FE model
can still be updated or optimized, using LMS Virtual.Lab Optimization.
LMS Virtual.Lab Model Updating easily handles incompatible geometries that typically
occur when comparing test and FE models. Models can be correlated geometrically through alignment,
sizing and mapping procedures. LMS Virtual.Lab Model Updating provides numerical tools, such as
MAC (Modal Assurance Criterion), FRAC (Frequency Response Assurance Criterion) as well as tools
to check orthogonality between two models, directly driving the Nastran Guyan reduction.
The MAC combined with Nastran Sol200 sensitivities for mode shapes and
eigenfrequencies helps users to compute and study MAC and frequency difference sensitivity for
mode pair sets. These sensitivities help obtain the best dynamic match between two models. LMS
Virtual.Lab Model Updating can deal adequately with mode switching during the updating process.
This ensures that the correct FE shape is used in correlation with the reference model during the
automated updating process.
Features:
 Input design parameters for material and element properties.
 Targets for mass, modal frequencies and vibration levels.
 Modal Assurance Criterion (MAC), Mode Pair Table and MAC Contribution (MACCo).
 Frequency Response Assurance Criterion (FRAC).
 Frequency difference sensitivity and MAC sensitivity.
 DOE, Response Surface Modeling and several optimization algorithms.

Benefits:
 Confirm FE simulation model validity using measurements.
 Identify modeling errors or evaluate modeling strategies.
 Improve Nastran model reliability with integrated optimization capabilities.
Figure 12.- LMS Virtual.Lab Model Updating Applications.

3. Uses for aircraft design and testing/certificating companies
In the following pages two specific examples, where the modal assurance criterion is used,
will be presented: “Using MSC/NASTRAN and LMS/PRETEST to find an optimal sensor placement
for modal identification and correlation of aerospace structures” and “Modal Test of L-610G
Aeroplane”. In both cases, the use of the MAC is essential to validate and/or check the tests performed
and the predicted FEA results.
3.1. Using MSC/NASTRAN and LMS/PRETEST to find an

optimal sensor placement for modal identification and correlation of
aerospace structures
The objective of an effective integration of finite element analysis with structural testing is
to combine the advantages of both approaches in a more valuable synergistic approach. The analytical
approach is predictive and can be used for predicting the flight loads and assessing the structural
integrity prior to the prototype production. The experimental approach, based on modal surveys on the
prototype, observes the actual behavior of the structure under controlled laboratory (ground vibration
test) or real operating conditions (in-flight testing). The benefits of such a combined approach are that:
 Testing provides reliable information to cross-check predicted FEA results (Correlation

Analysis): Testing can provide reliable estimates for system damping and resonance frequencies.
Furthermore, analyzing the experimentally obtained mode shapes, and comparing them with the
results from FEA, is critical in assessing the value of the analytical model and its interpretation. After
the difficult geometry mapping (geometrical correlation) that aligns both topologies, several modal
based assessment criteria are used to validate the analytical model. In aerospace, commonly used tools
therefore are the Modal Assurance Criteria (MAC) and the Cross-Orthogonality Criteria.
The modal assurance criterion (MAC) is used to evaluate the correlation between two modes
ignoring the effects of the system mass. It is an easy criterion and has been used primarily to check the
independence of two modes.
The cross-orthogonality is used to identify the corresponding test mode that associates with an
analytical mode, including the effects of system mass. A generally accepted requirement for the cross-
orthogonality is to have all diagonal terms larger than 0.9 and all the off-diagonal terms less than 0.1.
Since the outcome of the cross-orthogonality calculation is also dependent on the quality of the
measured test modes, the orthogonality matrix of the test modes with respect to the analytical reduced
mass matrix is used to assess the quality of thereof. The test data is acceptable if the off-diagonal terms
of this orthogonality matrix are less than 0.1 when the diagonal terms are normalized to 1.0.
The requirement for modal frequencies of corresponding experimental and analytical modes is to
have a discrepancy within 5%. If both criteria, cross-orthogonality and frequency discrepancy, are met,
the analytical model is said to be test-verified.
This in-depth correlation analysis will provide understanding of the discrepancies between the
analytical results and the test results, and will teach the designer how to improve his design.
 Testing results can be used to enhance the Analytical Model (FEA model Updating): The
outcome of the correlation analysis will decide if it is necessary to modify the analytical model so that
it better describes the results observed from testing. An improved analytical model is obtained by
changing analytical model parameters such that the discrepancy between test and FE resonance
frequencies is minimized. Such a structural optimization (updating) can be performed using the
MSC/NASTRAN Sol200 capability and thus are the changeable parameters shell thickness, beam
cross-sections, spring stiffnesses and such.
 The FEA results can be used to better design the Test (Pretest Analysis): FEA information can
complement the Test Engineer‟s expertise in selecting optimal ways of stimulating and measuring the
dynamic behavior of the test structure. Moreover, it will make the geometry mapping of both
topologies trivial since the experimental geometry was originally created from the FEA model.
This synergistic approach consists thus of the following steps, see Figure 13:
1. FEA Modeling and Analysis, using MSC/PATRAN and MSC/NASTRAN
2. Pretest Analysis using MSC/NASTRAN and LMS/PRETEST
3. Modal Testing & Analysis using LMS CADA-X Modal
4. Correlation Analysis using LMS/Correlation and MSC/NastranForLink
5. FEM Model Updating using LMS/Updating and MSC/NASTRAN Sol200
Figure 13.- Linking test and FE.

3.1.1. MSC/NASTRAN and LMS/PRETEST
A typical pre-test analysis will usually consist of different steps, see Figure 14. Starting
from a CAD model, an analytical model is created and the dynamic behavior is calculated in terms of
resonance frequencies, mode shapes and system‟s mass and stiffness matrices. Out of all these modes,
a limited set of target modes has to be selected and sensors and shakers have to be placed such that
they efficiently capture and excite all of these target modes. Lots of techniques and methodologies
have been developed already and are still being developed and most of them are implemented by
means of user programming (DMAP) in MSC/NASTRAN.
Using both MSC/NASTRAN and LMS/PRETEST in combination offers the structural

dynamicist an additional surplus because the outcome of most of his MSC/NASTRAN dynamic
calculations becomes available for interpretation in nice displays at the same time. In addition to that,
LMS/Pretest offers some additional tools.
Figure 14.- the different steps in a pre-test analysis.
3.1.2. Target Mode Selection
A first, very important step in the pre-test analysis is the selection of the target modes,
especially since the modal density of launch vehicles and other aerospace structures within the
frequency range of interest is usually very high. It is however not necessary to „capture‟ all these
closely spaced modes during a modal survey test, because only some of them will contribute
significantly to the critical component responses. These critical responses are usually located in the
areas of hardware concern.
These important structural modes are called „target‟ modes, and their selection is critical
for the generation of a validated analytical model. A test-verified model will, by definition, have a
good correlation between the test target modes and the analytical target modes, but not necessarily for
the non-target modes. It follows that a poor selection of the target modes could result in an analytical
model, which would not accurately predict the structural responses and member loads.
Before any criteria are used to determine the target modes, all modes coming from an
analytical modal analysis should be described in detail by a „simple‟ visual inspection. This inspection
gives the fundamental insight in the modal behavior of the structure and will also serve to interpret all
used target mode selection criteria.
Generally, there are four methods or combinations thereof widely used in the aerospace
industry. These are the rigid body modal effective mass, the constraint modal effective mass, the
modal kinetic energy fraction and the modal strain energy fraction. Another method, which uses a
somewhat different approach, is the use of modal participation factors. Besides these, also other
techniques are reported already.
Rigid Body Modal and Constraint Modal Effective Mass: The rigid body effective mass
associated with each deformation mode represents the amount of system mass participating in that
mode. Therefore, a mode with a large effective mass is usually a significant contributor to the system‟s
response. These criteria are in other words used to find the important system modes. A typical
requirement for the selection of target modes is that modes with a translational effective mass equal to
or greater than 2 percent of the total mass are target modes. If the modes are calculated using mass
normalization, the formula becomes:
𝑇 2
𝑀𝑒𝑓𝑓 = 𝜙𝑑 𝑀𝑠 𝜙𝑟𝑏
Note that this is the same as the root of the mass orthogonality between the deformation modes and the
rigid body modes. The constraint modal effective mass is similar to the rigid body modal effective
mass, but the constraint modes are used instead of the rigid body modes. This formula makes more
sense if the component (e.g. payload) is over constrained.
Kinetic/Strain Energy and Kinetic/Strain Energy Fraction: Since the modal effective mass
criteria look at the structure‟s dynamic behavior on a global basis, they are usually able to identify the
important system modes but they are less useful for the determination of important local modes. To
include the significant local modes of a subsystem for improving the response prediction, the kinetic
and/or strain energy fraction of that subsystem is calculated. The kinetic energy fraction is defined as
the amount of kinetic energy in that subsystem relative to that of the whole system. The selection
criterion to consider a component mode as target mode is an energy content of 50% of the total system
energy. These target modes will be added to the target mode set if not yet been selected by the
previous criteria. If the modes are again mass normalized the formula for the kinetic energy fraction
becomes:
𝑇
𝐷𝑖𝑎𝑔 𝜙𝑐 𝑀𝑐 𝜙𝑐 𝑇
𝐾𝐸𝐹 = 𝑇
= 𝐷𝑖𝑎𝑔 𝜙𝑐 𝑀𝑐 𝜙𝑐
𝐷𝑖𝑎𝑔 𝜙𝑠 𝑀𝑠 𝜙𝑠
The kinetic energy fraction for the first deformation mode of a scale model of a Boeing 747 is shows
in Figure 15 on top of the geometry. The fuselage, both wings including engines and the tail wings are
clearly visible as being the different components. Ultimately, it is possible to visualize the kinetic
energy of each element in model separately, see Figure 16.
Figure 15.- The modal kinetic energy for the first deformation
mode for several parts of Boeing 747. The lower plane is the undeformed
mode shape and each color represents a different group.
Figure 16.- The modal kinetic energy for the first deformation
mode for each element separately.
Mode Participation Factors: Although the previous methods may identify the most of the target
modes, some relevant modes critical to the payload or component responses may not be selected
because none of them takes the excitation into account. The structural integrity depends not only on
the structure‟s resonance frequencies, the mode shapes and the damping, but also on the frequency
characteristics of the excitations. Therefore, a tool that includes the excitation characteristics in the
target mode selection process will ensure the completeness of the target mode set. Useful in this
context are the mode participation factors, which are calculated during the dynamic solution
sequences, defined as (if mass normalization is used):
𝜙 𝑇 𝐹𝑖
𝑃𝐹𝑖 =
𝜆2𝑠 − 𝑤 2
The output is related to these participation factors by:

𝑁
𝑉 = 𝑃𝐹𝑖 𝜙𝑖
𝑖=1
Important is that these participation factors are independent of the output. The participation
factor is frequency dependant and its amplitude is determined by the structure‟s resonance behavior
(for w close to ls) and by the excitation spectra (for w far from ls), as can be seen in Figure 17, for the
PF of the first five modes. Plotting the participation factors of all modes for a certain frequency band
of interest results in the colormap diagram of Figure 18. It is now easy to investigate if some modes
are still being missed in the target mode set.
Figure 17.- The first five PF. On the x-axis the frequency
bandwidth, on the y-axis the amplitude of the PF.
Figure 18.- The PF for each mode (each vertical line

is a PF) in a colormap display.
3.1.3. Sensor Placement
Once the set of target modes has been defined, the measurement locations and their
corresponding degrees of freedom have to be chosen such that all target modes can be observed by the
modal survey test. This can be extremely though for large space structures where the target modes can
be closely spaced. Since it is not practical to instrument the test article in all degrees of freedom
corresponding to those of the analytical model, the challenge is to use a minimal number of sensors,
especially for in-flight testing, in order to sufficiently define the spatial resolution of all the target
modes. An erroneous or too limited subset of sensor locations will lead to an incomplete geometric
definition of the mode shapes, a phenomenon that is called „Spatial Aliasing‟.
To asses the correlation of the mathematical model predictions which in general do not
have dynamic degrees of freedom uniquely one to one with the modal test measurements, a reduction
(usually Guyan) to the test-analysis model (TAM) is required. Since this dynamic reduction is done
towards the measured degrees of freedom, the choice of the sensor set is also extremely important for
the outcome and the interpretation of the dynamic correlation tools.
Since the analytical model sizes of complete assemblies are way too big for a manual
selection of the sensor locations, a systematic approach in which the test engineer‟s experience is
central, see Figure 19, must thus be used. First a sensor set is searched to meet the observability
criterion. This set is then eventually modified to obtain a qualitative TAM model.
Figure 19.- Systematic approach to find the optimal sensor (and shaker) set.
Sensor set definition to meet the observability criterion: The methodology used to evaluate the
quality of a possible subset of the available analytical nodes and their corresponding degrees of
freedom, usually all three translational degrees of freedom, is a Modal Assurance Criterion (MAC)
calculation. If the off-diagonal terms of this MAC matrix are smaller than 0.1 or 0.2, the cross-
correlation between the target modes is sufficiently low and the chosen set of measurement point will
be able to observe all target modes.
If the initial group of points is not able to discriminate all target modes, a maximum
offdiagonality MAC (MODMAC) can be launched. This algorithm aims at the completion of the
initial subset with extra points/degrees of freedom that are chosen out of an additional subset such that
a resulting group of points/degrees of freedom is kept that, given a set of target modes, shows
offdiagonal MAC values below a given threshold. The initial group, the additional group and the target
modes are the only input to this algorithm.
Although MAC and MODMAC calculations are straightforward and powerful, the results and
especially the final amount of sensors still depend on the quality of the selection of initial and
additional set of possible measurement points. Different tools may assist the test engineer‟s experience
in the selection of those groups.
 Master DOF Selection – Geometrical Spread: This tool constructs a group with a user-
specified number of nodes that are maximally spread out over the structure. The
spreading can be performed either on all nodes of the structure, as is illustrated in Figure
20 where 50 points are spread out over the outer shell of the X-33 Advanced Technology
Demonstrator. To avoid the risk of clustering, it is possible to ask for a minimal distance
between the chosen locations. If a lot of component target modes were selected, it is also
necessary to have a sensor distribution on these components. Figure 21 shows a spread of
nodes on the internal LO2 tanks of the X-33.
Figure 20.- 50 nodes (triax) spread of the whole

structure of the X-33: Reusable Launch Vehicle.
Figure 21.- 50 Points spread of an internal tank of the X-33. A

character line wireframe gives the position of the tank in the whole model.
 Mode Shape Summation: This tools calculates the sum of a set or subset of (target)
modes and for this set of modes and within the selected nodes (assembly or component),
a user-specified number of nodes with the highest (summed) deformation will be
grouped. An example for a part of a satellite is shown in Figure 22. The summed mode
shape is shown together with the two most moving points.
Figure 22.- The summed mode shape (in color and deformation)
together with the undeformed mesh. Two points were asked as output.
 Group definition tools: Of course also different tools are available such as manual
group creation and editing, creation of groups by clicking points in the geometry,
grouping all nodes that correspond with a certain element type…
A possible strategy to find an optimal set of measurement points can be starting with a
relative small number of a-priori know response locations and launching a MODMAC
with as additional group a spread of points over the structure. If the target threshold
cannot be reached, for instance because there are local component modes amongst the
target modes, a second MODMAC can be launched with an additional group that
contains a spread of points only of that component…
Sensor set definition to meet also the cross-orthogonality criterion: Once a set of possible
measurement locations is found that meets the observability criterion, one still has to check if this set
of points can be used to obtain a high quality TAM model, by performing the actual reduction in
MSC/NASTRAN. If we suppose Guyan reduction, one can check if the mass distribution by the
calculation of the orthogonality between the spatially reduced modes and the Guyan reduced mass
matrix.
𝑇
𝑋𝑂𝑅 = 𝜙 𝑀𝑇𝐴𝑀 𝜙
An example is given in Figure 23. The target put forward for this orthogonality matrix is that the
diagonal terms are larger than 0.9, and the off-diagonal terms are smaller than 0.1 Therefore, the TAM
model produced by the chosen set of measurement points in the example, is only valid for the first 13
modes (including the 6 rigid body modes).
Figure 23.- Cross-Orthogonality using original

modes and Guyan reduced mass matrix.
It is also necessary to check if the modes from the TAM model are similar to the original target
modes. The correlation between the original target modes and the reduced TAM modes and the
resonance frequency discrepancy can be investigated using a MAC calculation. Typical for Guyan
reduction is that the reduction deteriorates for higher frequencies, as can be seen also in the example of
Figure 24. It is obvious that the current set of measurement points is only valid for the first 7 target
modes, although it is possible that more target modes can be observed using this set.
Figure 24.- Correlation between the original target modes and

the reduced TAM modes and the resonance frequency discrepancy.
If the cross-orthogonality and modal assurance criteria are not met, it is possible to add some
extra measurement points to the A-set. This process is more or less trial and error. If using Guyan
reduction, it is however possible to use the Master DOF selection – Ratio M/K tool.
3.1.4. Shaker Positioning
The third stage in the pre-test analysis is the selection of the exciter locations out of the
resulting group of measurement points in order to optimally stimulate all the modes of interest. If the
structure were to be excited close to a node of a particular mode, the corresponding resonance would
be difficult to observe in the measurement data, and the experimental modal model would be hard to
identify.
The tool that is used in LMS/Pretest for the selection of excitation locations is the
calculation of the „driving point residues‟ (DPR‟s). DPR‟s are stated to be equivalent to modal
participation factors, and are a measure of how much each mode is excited, or participated in the
overall response, at the driving point. As such also the modal participation factors in all possible
measurement points can be used. The definition of the driving point residue, for mode k and node i, is:
2
𝜙𝑖𝑘
𝐷𝑃𝑅𝑘 𝑖 =
2𝑚𝑘 𝑤𝑘
The degrees of freedom with maximum average DPR over all mode shapes are considered to be the
best excitation dofs for the specific set of target modes. An example is given for a tail boom problem
of a helicopter. The averaged DPR for all target modes is given, and the amplitude and the direction of
the red arrow show the best position and direction to place the shaker.
Figure 25.- Averaged DPR for all possible excitation points.
Figure 26.- A typical tail boom mode together

with the undeformed mesh.
3.2. Modal Test of L-610G Aeroplane
The L-610G is a high-wing monoplane with a T-shaped tail and a pressurised cabin
powered by two General Electric turboprop engines. Maximum take-off mass with forty passengers is
15 000 kg. The test was intended to determine the effect of significant structural changes on
aeroplane's modal characteristics and to obtain data to tune up an analytical finite-element model and
also to carry out detailed investigation into flight control circuits in all failure-present conditions
tolerated by the rules.
During the test the aeroplane was standing on under-inflated tyres of the main landing
gear, the nose fuselage was elastically suspended. Illustration of L-610G model test arrangement is
shown in Figure on envelope. A total of 19 exciters were used for excitation, the response was
measured in 280 points. The flap plays, which cause disturbing shocks and deteriorate measurement
results, were eliminated by additional masses of 20 kg suspended on the trailing edges by soft rubber
bundles. The additional mass natural frequency was less than 1 Hz and so it did not have to be
considered in the aeroplane total weight. The rubber bundle stiffness was taken into account in making
calculation corrections for additional masses and stiffness of exciters and transducers and for
aeroplane suspension.
The initial part of the test consisted in identification of the aeroplane natural frequencies.
Some important parts of the test are FRFs from all 280 transducers measured as recorded in different
configurations of swept sinusoidal excitation.
The aeroplane modal parameters were investigated by the method of sinusoidal excitation
of isolated normal modes (method of appropriated forces). Relative damping and generalised mass
were measured by two techniques, the complex power method and the method of forces in quadrature.
One problem of every modal test is evaluating the quality of received natural frequencies,
generalised masses, damping and modal vectors. Factors influencing modal test results fall into three
categories:
 Properties of the structure tested - modal density; linearity; damping intensity and
distribution; access to vibrating structural parts.
 Effect of the experiment - suspension of the structure tested; number, position and
magnitude of exciting forces; number and position of transducers; influence of moving
parts of exciters, transducers and suspension; method used; time for experiment
available; experimenter's skill.
 Technical level of experimental facilities - calibration; proper use of particular circuits.
The disturbing effect of the factors mentioned above must be minimalised. After
measuring each mode the result must be evaluated immediately so as to clear up uncertainties, if any,
or repeat the measurement. At the L-610G test several criteria were used for verification of linearity,
effects of moving parts of the test equipment and quality of natural modes isolation. Mode shapes were
checked by their graphic representation and also by verifying their orthogonality and by Modal
Assurance Criteria (MAC).
Table 2 shows the matrix of generalised masses verifying orthogonality of the L-610G
symmetric modes. Table 3 contains Auto-MAC values calculated for the same modes for all 280
measured points. In the two matrices there are following modes:
Mode f[Hz] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
SWB1 4,17 1 100
SEMVB 6,69 2 4,7 100
FVB1 8,41 3 0,1 -4,4 100
SWHB1 9,34 4 6,4 0,8 2,5 100
SWB2 10,83 5 3,1 -2,6 6,7 2,8 100
SEY 11,57 6 4,5 -2,4 -4,2 -3,2 -2,9 100
SEP 14,05 7 -2,4 3,4 -4,5 -4,7 -3,3 0,8 100
SEMHB 15,13 8 5,9 -7,0 -1,4 12,9 1,5 12,8 -5,3 100
SHTB1 15,88 9 -0,9 0,9 0,8 2,2 0,0 -0,4 -0,7 9,1 100
SWHB2 21,47 10 2,5 -0,9 -1,3 3,7 7,3 5,5 4,9 25,5 1,5 100
SWB3 22,8 11 -0,3 -3,2 -7,8 -4,2 -1,6 -0,7 7,4 0,2 1,1 23,0 100
SWT1 27,81 12 6,2 4,0 0,1 9,9 3,8 14,0 -4,9 5,2 -0,4 5,5 13,6 100
SWB4 36,78 13 -7,0 -2,9 4,1 15,7 7,5 -4,1 -7,6 3,5 1,4 -0,3 -6,3 13,4 100
SWHB3 46,1 14 -2,2 1,5 4,4 0,9 3,7 -1,6 -13,3 3,4 -0,6 -8,9 3,6 1,8 18,9 100
SWT2 46,6 15 4,8 -2,2 4,4 -8,1 -1,2 6,5 -1,5 3,3 1,3 10,6 12,1 1,0 -2,0 25,4 100
SHTHB1 54,0 16 -4,2 5,3 21,0 -1,8 -5,9 -9,5 -1,0 -5,9 -3,9 19,9 4,7 -2,0 -4,0 -4,3 -3,2 100
SHTT1 56,46 17 0,7 -0,1 7,6 0,3 -7,0 0,8 -3,8 0,7 10,2 -2,3 -2,8 0,1 0,5 0,8 0,0 25,1 100
SHTB2 60,1 18 0,6 -1,0 -4,5 -1,2 -5,5 -1,9 2,5 -0,7 -4,5 4,5 3,1 -1,0 0,8 -1,5 -1,0 -6,4 12,2 100
Table 2.- Orthogonality of L-610G aircraft symmetric mode shapes.

Mode f[Hz] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
SWB1 4,17 1 100
SEMVB 6,69 2 24,3 100
FVB1 8,41 3 0,3 0,4 100
SWHB1 9,34 4 13,9 2,1 16,8 100
SWB2 10,83 5 23,3 2,4 6,3 11,3 100
SEY 11,57 6 8,2 1,2 10,3 5,0 15,8 100
SEP 14,05 7 19,2 14,8 15,0 9,1 4,1 2,4 100
SEMHB 15,13 8 1,8 1,5 12,2 5,1 0,5 11,0 15,0 100
SHTB1 15,88 9 0,6 0,1 16,4 6,9 11,5 22,2 5,5 18,7 100
SWHB2 21,47 10 0,1 0,7 0,2 3,6 2,4 4,2 2,8 7,2 0,1 100
SWB3 22,8 11 6,7 4,2 0,1 3,7 20,4 9,5 34,8 3,1 0,3 3,2 100
SWT1 27,81 12 22,0 11,5 0,5 0,6 0,6 4,1 11,2 1,6 0,0 0,0 1,1 100
SWB4 36,78 13 3,8 2,1 0,1 0,2 0,2 1,4 4,9 0,7 0,0 0,1 0,5 16,5 100
SWHB3 46,1 14 0,4 2,0 0,0 0,3 0,3 1,4 2,4 2,9 0,3 1,1 1,5 2,1 2,8 100
SWT2 46,6 15 0,0 0,0 0,0 1,0 1,2 0,3 1,4 0,6 0,0 2,1 0,1 5,1 13,8 19,0 100
SHTHB1 54,0 16 0,0 0,1 4,3 0,2 0,9 0,6 0,4 0,0 1,9 0,5 0,3 0,0 0,4 0,7 0,0 100
SHTT1 56,46 17 0,0 0,0 0,2 0,1 0,4 0,0 0,0 0,1 0,6 0,0 0,0 0,0 0,0 0,1 0,0 30,3 100
SHTB2 60,1 18 0,0 0,0 0,6 0,2 0,1 0,0 0,0 0,4 4,1 0,0 0,0 0,0 0,0 0,1 0,0 19,5 1,2 100
Table 3.- Auto MAC of L-610G aircraft symmetric mode shapes.
SWB1 symmetric first wing bending SWHB2 symmetric second wing horizontal bending
SEMVB symmetric engine mounting vertical bending SWB3 symmetric third wing bending
FVB1 first fuselage vertical bending SWT1 symmetric first wing torsion
SWHB1 symmetric first wing horizontal bending SWB4 symmetric forth wing bending
SWB2 symmetric second wing bending SWHB3 symmetric third wing horizontal bending
SEY symmetric engines yaw SWT2 symmetric second wing torsion
SHTHB1 symmetric first horizontal tailplane horizontal
SEP symmetric engines pitch
bending
SEMHB symmetric engine mounting horizontal bending SHTT1 symmetric first horizontal tailplane torsion
SHTB1 symmetric first horizontal tailplane bending SHTB2 symmetric second horizontal tailplane bending
In the two matrices the values are given in per cent. Most non-diagonal elements are very
small under 10 per cent. Only a few values are slightly greater. Most of them are "horizontal" modes
with movement in longitudinal direction. In this direction there was limited number of measured
points on fuselage and engines. It is assumed that this is the reason why these values are rather higher.
With the Auto-MAC values the situation is similar, the elements over 10 per cent have again
"horizontal" mode shapes and in addition, mode shapes of engines. But it may be seen that higher
values may also be found with some shapes featuring good orthogonality. MAC depends to some
extent on choice of the points (and directions) on the structure whose measured displacements were
incorporated into the calculation. In analysing orthogonality and particularly the Auto-MAC values, it
was useful to assess the effect on the resulting value of motion components in the co-ordinate axes of
particular aeroplane parts.
Figure 27.- Test of L-610G Aeroplane.

Conclusions
Over the last twenty years, the modal assurance criterion has demonstrated how a simple
statistical concept can become an extremely useful tool in the field of experimental modal analysis and
structural dynamics. The use of the modal assurance criterion and the development and use of a
significant number of related criteria, has been remarkable and is most likely due to the overall
simplicity of the concept. New uses of the modal assurance criterion and new criteria will be
developed over the next years as users more fully understand the limitations of the current criteria.
Certainly in the next few years, the increased use of other statistical methods as well as further
development of singular value/vector methods are related areas that will generate useful tools in this
area.
Even so, it will always be important to recognize the origins and limitations of tools like
the modal assurance criterion to avoid misuse of the methodology. Simplistic tools like the modal
assurance criterion are limited in their meaningful application. The development of related assurance
criteria has been initiated by shortcomings, real or perceived, of the original modal assurance criterion.
Dissatisfaction often has resulted from the misuse of these tools by users, removed from the actual
development or unaware of application limitations in subsequent implementations. It is clear that users
will continue to need more feedback concerning quality assurance information relative to experimental
modal parameters and that new techniques, particularly statistical methods that utilize the redundant
information present in the measured data, will continue to be developed with strengths and
weaknesses, just like the modal assurance criterion.
In regarding to the uses for aircraft design and testing/certificating companies, this report
has tried to explain why a carefully performed pre-test analysis really is necessary to end-up with
meaningful modal survey test results. Good test results are really a sine qua non for the interpretation
of all dynamic correlation tools that are used in the analytical model verification and validation.
An overview is given of the commonly used techniques to address the target mode
selection, the sensor location placement and the positioning of the exciters. It may be clear that the use
of the available tools and the user programming capabilities of MSC/NASTRAN form a crucial aspect
for the calculation of all described tools. The synergy of MSC/NASTRAN and LMS/Pretest gives the
engineer the additional benefit that, a unique environment becomes available that guides the engineer
through the complete process, from pre-test analysis over correlation to end up eventually at the model
updating step, that the interpretation of the calculations can be visualized and that both program
communicate directly with each other.
Bibliography
[1] Allemang, R. J.: The Modal Assurance Criterion – Twenty Years of Use and Abuse. Sound and
Vibration, pp. 14-21, August 2003.
[2] Van Langenhove and T., Brughmans, M.: Using MSC/NASTRAN and LMS/PRETEST to find
an optimal sensor placement for modal identification and correlation of aerospace structures.
LMS International, Belgium, 1999.
[3] Cerný, O.: Primera Modal Analysis Systems and Software/Aircraft Modal Testing. Report
VZLÚ 3/98.
[4] LMS: LMS Virtual.Lab Introduction. LMS International (Belgium).
[5] LMS: LMS Virtual.Lab Correlation. LMS International.
[6] Braun, R., Madsen, N. and Meruane, V.: Análisis Modal Experimental de una Estructura
Aeronáutica. Universidad de Chile, Facultad de Ciencias Físicas y Matemáticas, Dpto. de
Ingeniería Mecánica, Chile, Julio de 2003.
[7] López-Díez, J., Marco-Gómez, V. and Cuerno-Rejado, C.: Modal Test Correlation and Error
Localization for Finite Elements Models of Spacecraft structures. C.A.S.A. División Espacio,
Spain, 1999.
[8] Garbayo, M. E. and Pintor, J. M.: Ajuste y validación de modelos teóricos mediante ensayos de
vibración sobre el componente. Universidad Pública de Navarra, Dpto. Ingeniería Mecánica
Energética y Materiales, Spain, 2006.

Modal Assurance Criterion - Álvaro Rivero

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modal Assurance Criterion - Álvaro Rivero

Uploaded by

Copyright:

Available Formats

DINÁMICA

“The Modal Assurance Criterion (MAC)”

Asignatura: Dinámica y Fatiga

Alumnos: Rivero González, Álvaro

1. State of the Art. Antecedents. review of existing methods 5

1.1. Historical Development of MAC 5

1.1.1. Modal Vector Orthogonality 6

1.1.2. Modal Vector Consistency 8

1.1.3. Modal Assurance Criterion (MAC) Zero 10

1.1.4. Modal Assurance Criterion (MAC) Unity 11

1.1.5. MAC Presentation Formats 12

1.2. Other Similar Assurance Criteria 14

1.3. Uses of the Modal Assurance Criterion 17

1.4. Abuses of the Modal Assurance Criterion 18

1.5. Current Developments 20

2.1. LMS Virtual.Lab Overview 21

2.2. LMS Virtual.Lab Correlation 25

2.2.1. Systematic Validation from the bottom up 26

2.2.2. LMS Virtual.Lab Correlation. Features and Benefits 27

2.2.3. LMS Virtual.Lab Model Updating. Features and Benefits 29

3. Uses for Aircraft Design and Testing/Certificating Companies 31

3.1. Using MSC/NASTRAN and LMS/PRETEST to find an

3.1.1. MSC/NASTRAN and LMS/PRETEST 33

3.1.2. Target Mode Selection 33

3.1.3. Sensor Placement 37

3.1.4. Shaker Positioning 41

3.2. Modal Test of L-610G Aeroplane 43

1. State of the art. Antecedents. Review of existing methods

L = Number of matching pairs of modal vectors.

1.1. Historical Development of MAC

In the internal development of the modal assurance criterion at the University of

1.1.1. Modal Vector Orthogonality

1.1.2. Modal Vector Consistency

The function of the modal assurance criterion (MAC) is to provide a measure of

The modal scale factor is defined, according to this approach, as follows:

𝑀𝐴𝐶𝑐𝑑𝑟 = 𝑀𝑆𝐹𝑑𝑟 𝑀𝑆𝐹𝑑𝑟𝑐 (4e)

1.1.3. Modal Assurance Criterion (MAC) Zero

1.1.4. Modal Assurance Criterion (MAC) Unity

1.1.5. MAC Presentation Formats

Table 1.- Numerical presentation of MAC values.

Figure 1.- 2-D and 3-D presentation of MAC values.

1.2. Other Similar Assurance Criteria

Coordinate Modal Assurance Criterion (COMAC): An extension of the modal assurance

Inverse Modal Assurance Criterion (IMAC): An alternative approach to increasing the

Complex Correlation Coefficient (CCF): A significant variation in the frequency response

Frequency Domain Assurance Criterion (FDAC): A similar variation in the frequency

Coordinate Orthogonality Check (CORTHOG): The coordinate orthogonality check

1.3. Uses of the Modal Assurance Criterion

 Validation of experimental modal models.

 Correlation with analytical modal models (mode pairing).

 Correlation with operating response vectors.

 Mapping matrix between analytical and experimental modal models.

 Modal vector error analysis.

 Modal vector averaging.

 Experimental modal vector completion and/or expansion.

 Weighting for model updating algorithms.

 Modal vector consistency/stability in modal parameter estimation algorithms.

 Repeated and pseudo-repeated root detection.

 Structural fault/damage detection.

 Quality control evaluations.

 Optimal sensor placement.

1.4. Abuses of the Modal Assurance Criterion

 The modal analysis criterion is not an orthogonality check.