You are on page 1of 15

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

THE DATA USED IN AGILE FACILITY LAYOUT DESIGN PROCEDURE Agnieszka Stachowiak and Marek Fertsch Faculty of Computing and Management Poznan University of Technology, Poznan, Poland
a_stachowiak@wp.pl, marek.fertsch@put.poznan.pl

ABSTRACT Agile manufacturing is one of the latest conceptions in manufacturing management. It is the way to deal with globalisation and growing pace of changes in technologies, techniques, customers demands and competitors activities. Agility is the quality that can be achieved thanks to virtual organisation of processes and the extended enterprise conception along with proper facility layout. To generate facility layout suitable for agile manufacturing the Agile Facility Layout Design Procedure was developed. The procedure is based on evolutionary algorithm, as biologically inspired systems and methods are believed to be effective in solving Facility Layout Problem. The application of evolution-based methods to FLP solving is common due to its characteristics, but applying them to agile FLP brings forth one issue: the data to use. Traditionally facility layout is based on products the facility manufactures and processes it performs and this approach is useless in agility terms, as agility is about changes in products as well as in processes. Coping with these changes is possible only thanks to reliable information and data. This paper presents the idea of collecting data and the data used in developed procedure. Key words: production management, optimisation, AI

1. INTRODUCTION The idea of the paper is to present the idea of agile facility and the design procedure developed to generate agile facility layout. The most important and difficult stage of the procedure is the data to use. Thus the aim of this paper is to present the way of collecting and using data in AFLDP. The paper comprises five sections: section one is the introduction to the subject, section two introduces the conception of agile manufacturing, the most popular definitions of agility and agile manufacturing, points out agilitys characteristics and shows the idea of agile facility, the assumptions made and the model developed. Section three presents Agile Facility Layout Design Procedure (AFLDP), its stages and their brief characteristics. Next section discusses the data used in the procedure and gives an example of using the procedure described in the previous part. The last section is the concluding section. 2. AGILE MANUFACTURING 2.1. Definitions Increasing pace of changes in customer demands, technologies and market environment has made competing based on cost and quality factors inefficient way of gaining satisfying market position. Therefore, a new strategy, new model of behaviour, has become indispensable for
6.5.1

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

each company, especially for the manufacturers. Essentially, manufacturing needed to become a service (Emigh, 1999),because coping with volatility of the market is possible only thanks to individualising companys offer, which means turning production into craftsmanship again, or even the art the art of agility the will, knowledge and skill to continually reconfigure and integrate the process of making things (Mutch, 1996). Agility is the approach that can be applied to deal with challenges of todays market and difficult conditions successfully as it no longer employs the idea of product-focused production organisation. Being agile means having a quick resourceful and adaptable character, so it is basically being adaptive, able to adjust to changing situation. The definitions and interpretations of agile manufacturing are usually based on unavoidable changes on the market (changes concerning various fields of companys activity) and the customer centric way of thinking about manufacturing. Thus, for a company agility is the ability to respond efficiently (doing it at a profit) to demand volatility (Fraser, 1995) as the volatility of todays market is the main reason for creating agile manufacturing conception. According to Iacocca Institute (Fraser, 1995) agile manufacturing can be defined as the ability to thrive and prosper in a competitive environment of continuous and unanticipated change and to respond quickly to rapidly changing markets driven by customer-based valuing of products. Speaking Steven Ashleys words agile manufacturing is the idea that a successful company should be adaptable and flexible enough to respond rapidly to changes in demand and bring highly customer targeted products to market faster (Ashley, 1998). These definitions are precisely showing the essence of agile manufacturing, stressing the customer needs importance and volatility factor. To achieve expected, agility-driven level of responsiveness, flexibility and adaptability it is essential for the company to understand and use all the information coming from the customers, competitors, partners, suppliers and, last but not least, information flows in the company. 2.2. Agility Characteristics According to agility and agile manufacturing definitions, agility characteristics are following (Stachowiak, 2004): - Flexibility; - Co-operation (with customers, suppliers, competitors, partners and company employees); - Communication; - Customer-focused strategy; - Adaptability; - Responsiveness. Possessing these characteristics makes company agile, which means being competitive, efficient and successful. Thereupon the crucial issue for modern company is to find a way to achieve qualities mentioned above in order to be agile and, as a result, to become and stay successful. Some of agilitys characteristics can be achieved thanks to proper facility management: communication, co-operation and customer-focused strategy are the matter of proper operational and strategic management which usually involves organising virtual teams, employing so called agile (well qualified, multi-skilled) staff and co-ordinating information flows. Virtual teams are the perfect organisational structures , as nowadays products are not necessarily developed in one place, which makes solid, formal structures inefficient. Organising virtual teams enables collaborative work among geographically dispersed product
6.5.2

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

development team members (the dispersion is the reason why agile manufacturing often takes advantage of strengths in communication and information technologies (Emigh, 1999) the only tool than can provide fast, continuous and easy communication is the internet and that is why agile manufacturing is sometimes called e-manufacturing). Virtual teams usually employ people of different professions and experiencesin order to improve products and processes. Without people and their skills performing agile processes would be impossible, as agility is more like a craftsmanship with unique products manufactured.. Undoubtedly wellqualified, multi-skilled workers are expensive (high salaries/ wages, cost of training etc.) but they are worth the money as multi-skilled virtual teams are able to develop new product in very short time, which makes manufacturing flexible, innovative and responsive. The other qualities of agility can be achieved thanks to proper facility layout. Flexibility, adaptability and responsiveness of the manufacturing process are dependent not only on people, but they are often linked with flexible automation of a production process and FMS (Flexible Manufacturing Systems). Hence flexibility is not only the issue of adaptable management, but it is also dependent on machines performing the production process, especially on the facility layout. The ideas most often used in making facility layout agile are: flexible manufacturing units, u-shaped lines and making facilitys equipment as universal as possible. Putting these elements together, combining them to benefit from all the ideas presented, is the crucial issue when designing agile facility. 2.3. Agile Facility The facility layout has a considerable influence on production process characteristics (Azadivar et al. 2000) and the agile layout is supposed to provide agility and, in consequence, customers satisfaction. It is impossible without co-operation between the company and the customers. Synchronising manufacturing process and market requirements is the main characteristic of productfocused production based on JustinTime system, but applying JIT (and consequently implementing synchronised production line, which is typical in this conception) to agile manufacturing is impossible, because of on-product orientation of this system. Agile facility is not product oriented, as the basic assumption of agility is uncertainty about products to manufacture. On the other hand flexibility (crucial for gaining agility) is characteristic for traditional cellular manufacturing system, which is a processfocused way of production. Agile manufacturing is, according to its definition, also typical processfocused manufacturing conception, as products manufactured are various and usually unique. These two conceptions (JIT and cellular manufacturing) are traditionally believed to be contradictory and conflicting, but they coexist in agile manufacturing conception, what is more agile manufacturing takes advantages from both of them. The dualism of agile manufacturing conception is reflected in agile facility layout assumptions. Combining characteristics of process-oriented manufacturing with synchronisation and flexibility leads to introducing following model of agile facility layout (Stachowiak, 2004): 1) Agile facility comprises two kinds of units: - Flexible manufacturing units, u-shaped (or multi-u-shaped which is a consequence of design procedure employed), technologically specialised, performing the most labourconsuming stages of a production process, - Universal machines to perform the other stages of a production process. 2) Flexible manufacturing units are allocated in the central part of a facility, 3) Universal units are allocated on its perimeter,

6.5.3

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

4) Input and output to the manufacturing process are located on the opposite sides of agile facility. The scheme of the agile facility is shown in the figure 1. The shape of facility in the scheme is rectangular, which is the most common instance. The idea of agile machine placement can also be used in facilities with the irregular floor shape, as the most important is the relation between the external, universal machines and the internal, specialised ones.

Flexible Manufacturing Units Universal Machines Figure 1. The agile facility layout. The agile facility model presented (figure 1.) shows a general conception of agile facility, it does not demonstrate any specific shop floor design, but it clearly illustrates the agile manufacturing unit characteristics. 3. THE FRAMEWORK OF AGILE FACILITY LAYOUT DESIGN (AFLD) PROCEDURE. The part of AFLDP presented is used to allocate machines in flexible units as the universal units are to be placed along the perimeter of the facility, so their allocation depends only on their quantity and size. The part responsible for allocating machines in flexible units is based on evolutionary algorithms as evolutionary optimisation is perfect for complex problems (where traditional, mathematical methods of optimisation are slow if not useless) and facility layout problem is usually big (facilities contain plenty of different units) and complex as there are always some constraints to abide by (Suresh et al., 1995), (Rajasekharan et al., 1998), (Tam et al., 1998). Evolutionary strategy (one of the evolution-based optimisation method chosen on account of it characteristics) applied to Agile Facility Layout Design (AFLD) procedure is performed on a population of solutions which are encoded, random facility layouts. It uses basic genetic operations, i.e. crossover and mutation to produce children - new layouts. The offsprings fitness is evaluated and solutions with the best fitness value are joined to the next generation. The scheme is repeated until the expected fitness is obtained for at least one solution as it is shown in figure 2 (Stachowiak et al., 2004). This solution (the one with expected fitness) is to be the final facility layout.

6.5.4

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Initial population Evaluating fitness value Crossover Repairing chromosomes Loop until the expected fitness is achieved Mutation Evaluating fitness value Selection Creating new population

Figure 2. AFL evolutionary strategy Encoding solutions. The way of encoding solutions differs on account of the problem to be solved. The problem analysed in this project comprises forty groups of machines, the quantity of machines in each group is predefined. These groups are to be divided into two sets: - the set of highly specialised machine groups; - the set of universal machine groups. The set of universal machine groups is to be placed along the perimeter of a facility and the set of highly specialised groups is to be allocated according to material and information flows inside the facility, on the basis of analysis of connections and relations between them. Units which are the elements of specialised groups are to be placed using the AFD procedure. Each of the units has its own code, which is a letter; each string of letters makes a chromosome (representation of group of machines). Decoding solutions. The order of letters in a chromosome determines the order of allocating machine groups through the facility area. The pattern of allocating objects used in our project is shown in the figure 3 (it is so-called yoscillatory scheme (Hamomamoto et al., 1999), (Islier, 1996)):

Figure 3. The scheme of allocating objects The number of machines to place in each row is dependent on the size of the facility floor and the quantity of machines to place, as well as their size. The procedure of decoding solutions is based on the predefined code, which is a table comprising following information: - The name of machine;

6.5.5

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

- The code letter; - The size of machine. There are also some further information that should be predefined: - The width of aisles; - The number of machines/ the length of chromosome; - The space available. The first stage of procedure is to determine the shape of unit (the length of each row, the number of rows, the size of unit), the second is to allocate machines respectively to the place they take in chromosome and their sizes. The layout should be brought into the coordinate system to enable calculating solutions characteristics. After the decoding solution is ready to be used and/or checked. Initial population. The very first step in performing an ES search is choosing the initial population of solutions (from a pool of potential solutions), evaluating their fitness and choosing the best individuals for reproduction. It is necessary to create an initial population before the first selection is performed, simply to have some chromosomes/solutions to choose from. After the first life cycle, the next population is selected from the individuals bred during the reproduction process. The problem we are working on comprises forty groups of machines used in the manufacturing process of tractors. The initial population in our project is one hundred random solutions derived from a permutation of initial sequence of machines, showed as a string of letters. The size of the population depends on a project and its complexity. We decided to settle the number of solutions to survive at the level of maximum quantity of individuals. Optimisation criteria. Traditionally the criteria employed in facility layout optimisation is the time, cost or distance. Minimisation of these factors is usually the object function in optimising facility layout. The optimisation criteria applied in the project are: the time, the cost, the distance, and the ratio derived from these factors and their weights. Weighing factors is based on their importance for achieving agility, the highest weight characterises the most important factor, the lowest - the least important one. According to the agility features presented, the least important issue is cost and the most important one is time. The designer using the method developed is able to choose which optimisation criteria is the most suitable for the problem to be optimised, as well as the weights for the factors chosen. Evaluating fitness value. Evaluating fitness value during the research phase we are in consists in evaluating cost, time, distance and mix of all these criteria for each solution. The essential problem in agile facility designing is what data to use. The procedure of gathering data and calculating fitness value is presented in the next section of present paper. Selection. Selection is based on a fitness value. Solutions with the best fitness value are to survive. The most popular mechanisms of selection are the tournament and the roulette wheel. Both are used in this project and the question is to check whether the method has any influence on the quality of the solution finally obtained. The stochastic mechanisms of selection are used before the first run of reproduction (choosing random solutions to be parents, all solutions with the same probability to be chosen). The deterministic mechanism of selection is used after the reproduction stage to create a new generation (choosing parents with probability dependent on their fitness).

6.5.6

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Crossover. The crossover is performed to create new individuals . Pairs of parents used in the breeding procedure are chosen by the selection procedure. The only exception are parents from the initial population, as they are chosen stochastically. The crossover applied in our project is one point order crossover (there are many crossover patterns, usually they depend on problem to be solved). The point of crossing is chosen randomly for each solution. For a pair of parent chromosomes: ADGEFCHB ABCDEFGH And the crossing point i=3, the crossing operation runs as follows: ADGEEFGH ABCDFCHB The crossover is performed with predefined crossover probability. Mutation. Mutation is a genetic operator, which is a kind of self-crossover. It is an exchange of genes symbolised by letters and it takes place in one chromosome area. ADGEEFGH AGDEEFGH The mutation is performed with predefined mutation probability. Repairing Chromosomes. After the crossover and mutation some of the chromosomes may be damaged. The damage is multiplied letters, which means that some of the machine groups are to multiply, and some of them to disappear. The job of the repairing procedure is to detect multiplied letters and letters which disappeared, and to replace every extra letter with the missing one. The repairing procedure is shown in figure 4:
Analysing the chromosome Loop until all the letters in chromosome are used only once Replacing multiplied letters with missing ones

Correcting the analysis

Figure 4. The repairing procedure algorithm Stage one of the algorithm is counting all the letters in chromosome to check whether there are multiplied or missing letters. Stage two is performed for each multiplied letter. The idea is to find missing letter (the letter with number zero) and replacing additional letter with missing one. The order is alphabetical. Stage three is correcting the results achieved in stage one. The example of running the repairing procedure:

6.5.7

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

AGEBFCHBB Analysing stage: The number of As = 1 The number of Bs = 3 The number of Cs = 1 The number of Ds = 0 The number of Es = 1 The number of Fs = 1 The number of Gs = 1 The number of Hs = 1 The number of Is = 0 The conclusion is that two of Bs have to be replaced with D and I. First loop replaces first B with D (alphabetical order): AGEDFCHBB Following corrections are made: The number of Bs = 2 The number of Ds = 1 Second loop replaces second B with I (the only zero letter left): AGEDFCHIB Following corrections are made: The number of Bs = 1 The number of Is = 1 After running the second loop there are no more multiplied or missing letters so the repairing procedure is completed. Creating a new generation. This part of AFD procedure consist on adding new solutions to the population and removing the ones not good enough to survive. Children and parents with the best fitness values are to survive and make a new generation. Loop until the expected fitness is achieved. The expected fitness value for our project is half the time, the same cost and three fourths of distance achieved in a conventional solution. Of course, each of these ratios is likely to be changed during the research and the influence of changes made is to be analysed. In case of not achieving the presumed level of fitness factor values the procedure stops after a predefined number of runs. Summing up, the evolutionary strategy used in Agile Facility Design (AFD) procedure can be divided into two phases: the first one is an initial phase, in which the initial population is randomly created, its fitness is evaluated, then it is encoded and individuals for the reproduction are chosen. The second phase is growing the best solution which reproduces, evaluates and selects the individuals to the next generation. In this phase the selection is hidden in the creation of a new generation stage and it is based on a fitness value of the evaluated individuals. The second phase leads to reaching the best solution. 4. THE DATA TO USE IN AFLD PROCEDURE The procedure presented in the previous section needs following data connected with evolutionary operations it performs: - Probability of crossover;

6.5.8

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Probability of mutation; Number of individuals in each population; Maximum number of generations to run; Expected fitness values which are (in the case presented) expected level of cost generated by process, time of the process to be performed and distance for products/modules to run during the process (the procedure developed includes also the weighted sum of all these factors); - Weights of optimisation criteria (time, cost and distance) to be used when calculating weighted sum. All the data mentioned are set when the procedure is initiated. They depend on designers experience and knowledge. As they have an influence on procedures performance i.e.: - Low mutation probability makes population less differential and often leads to local optimum too fast, while high mutation probability makes generation unstable and leads to degeneration of good solutions; - The number of generations is the emergency exit, stopping the procedure when no individual can meet expected fitness value that was set too high; - Expected fitness values set too low will not make satisfying layout and set too high will not let the procedure to be finished etc, it is usually recommended to run the procedure several times to find the best set of characteristics. The more important for achieving agility are the data connected with processes and products. As agility gives the designer no certainty about products to be manufactured and processes to be performed the designing procedure has no data to rest. That is why the following scheme of collecting and using data in Agile Facility Layout Design Procedure was developed (figure 5):
Stochastic determination of group of products. Analysing products and their manufacturing processes Evaluating fitness values of solutions based on products characteristics

Average fitness of a group of products = fitness of a chromosome (layout model)

Figure 5. Evaluation of a solutions fitness value. To follow the scheme presented in figure 5 there are some assumptions that have to be made: - First of all the domain of facilitys activity is know/ established (when designing new facility); - Data comprising technological processes of typical products from companys scope of interest are available. To complete the data to be used in AFLDP the designer is expected to: - Choose randomly predefined number of products from the branch considered; - Analyse technological of products chosen. Hence the data connected with products and processes comprise:

6.5.9

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

- Number of products chosen; - Number of operations in each products manufacturing process; - Machines to run each operation; - Time and cost of each operation; - Distance between each stage of manufacturing process. Thanks to the data presented above the designer/ design procedure is able to evaluate fitness value of chromosome generated with evolution process, and to make facility agile. 4.1. The Example Of AFLDP To present the idea developed in the previous section short and simple example is to be shown. The data used in our project comprise data connected with machines and products along with parts they are composed of. The problem to be solved in our project is a complex one and given its gravity we decided to show the idea of AFD by presenting a simplified example. As the part of AFD procedure, which performs genetic operations, is quite simple and universal it does not require more detailed explanations. However, we would like to show some details connected with the fitness evaluating part as this is the part where data connected with products and processes are used. The size and the length of the presented example make it easy to analyse it and comment on it. To evaluate fitness value of the solution generated by an evolutionary strategy, the data covering cost, time and distance are necessary. The calculations are based on technological processes of the products (we have decided to use the data covering technological processes of three products, typical for the branch considered, for each calculation). Fitness (cost, time, distance and mix of all these criteria) is evaluated for each product chosen and then the average is calculated. The average is obtained in all four aspects of optimisation: cost, time, distance and mix, the choice of which dimension is to be the most important one is the designers or the users of the solution. Data to use in the example are following: nm = 8 nm - number of machines to allocate np = 3 np - number of parts to evaluate fitness value One of the solutions generated with AFD procedure: ABGEDHFC Allocation of machine groups on the facility floor is shown in figure below (Figure 6):

G B A H
Figure 6. The layout of the facility obtained with AFD procedure (an example) Fitness of this solution is the result of calculations based on characteristics of random parts technological processes (the whole procedure: look back to the previous section). The data in the tables below are calculated for each part and solution.

E D

C F

6.5.10

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Manufacturing time is the time of each operation, the data is given in the description of manufacturing process of each part. The time of transport between operations is calculated with formula 1: ttxi-j = v * dxi-j (1) ttxi-j time of transporting part x between operations i and j v predefined speed of transportation dxi-j distance between machines performing operations i and j on part x Manufacturing cost is the cost of each operation, the data is given in the description of manufacturing process of each part (the data is based on cost of machines and peoples work). The cost of transport between operations is calculated with formula 2: ctxi-j = ctdx * dxi-j ctxi-j cost of transporting part x between operations i and j ctdx predefined cost of transporting part x on one distance unit dxi-j distance between machines performing operations i and j on part x The distance between machines performing operations is calculated with formula 3: dxi-j = ((xj - xi )2 + (yj - yi )2)1/2 dxi-j distance between machines performing operations i and j on part x (xi, yi) coordinates of machine performing operation i (xj, yj) coordinates of machine performing operation j The last column of the tables is the sum calculated respectively with formulas 4, 5, 6, 7, 8: tmx = tmxi
i=1..n

(2)

(3)

(4)

tmx total time of manufacturing operations on part x tmxi time of operation i on part x n the number of operations ttx = ttxi-j
i=1..n-1 j=2..n

(5)

ttx total time of transporting part x cmx = cmxi


i=1..n

(6)

cmx total cost of manufacturing operations on part x cmxi cost of operation i on part x ctx = ctxi-j
i=1..n-1 j=2..n

(7)

ttx total time of transporting part x dx = dxi-j


i=1..n-1 j=2..n

(8)

dx total distance for part x

6.5.11

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Table 1. Machine group Manufacturing Time Time of transport Manufacturing Cost Cost of transport Distance B 20 tu 5 tu 80cu 5cu 10du Table 2. Machine group Manufacturing Time Time of transport Manufacturing Cost Cost of transport Distance A 50tu 5tu 100cu 5cu 10du
Table 3.

Part A: manufacturing process G 20 tu 5 tu 60cu 5cu 10du E 40 tu 5 tu 40cu 5cu 10du G 20tu 5 tu 60cu 5cu 10du E 50 tu 50cu 150tu 20tu 290cu 20cu 40du

Operation1 Operation2 Operation3 Operation4 Operation5 Sum:

Part B: manufacturing process E 50tu 5tu 50cu 5cu 10du D 20tu 5tu 120cu 5cu 10du F 40tu 5tu 40cu 5cu 10du H 80tu 60cu

Operation1 Operation2 Operation3 Operation4 Operation5 Sum:

240tu 20tu 370cu 20cu 40du

Part C: manufacturing process

Operation1 Operation2 Operation3 Operation4 Operation5 Sum:

Machine group Manufacturing Time Time of transport Manufacturing Cost Cost of transport distance

B 20tu 5tu 80cu 5cu 10du

E 50cu 10tu 80cu 10cu 20du

D 40tu 5tu 40cu 5cu 10du

F 100tu 5tu 40cu 5cu 10du

C 45tu 90cu

355tu 25tu 330cu 25cu 50du

Note: Measures used to characterise cost, time and distance of production process of the analysed details are respectively: cost unit (cu), time unit (tu) and distance unit (du). Summing up the data from tables above total cost and total time of manufacturing single part are calculated with formulas 9 and 10: Tx = ttxi-j + tmx Tx total time of manufacturing part x ttx total time of transporting part x tmx total time of manufacturing operations on part x Cx = ctxi-j + cmx (10) Cx total cost of manufacturing part x ctx total cost of transporting part x cmx total cost of manufacturing operations on part x The average is calculated with formulas 11, 12 and 13. Applying the weighted average is not necessary, what is more, according to agility characteristics not recommended. (11) ac = Cx np ac average cost of manufacturing one part in agile facility (9)

6.5.12

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

np number of parts
at = Tx

np

(12)

at average time of manufacturing one part in agile facility ad = dx np ad average distance for one part to go in agile facility
Table 4. Cost, time and distance of final solution

(13)

Part A Part B Part C Average:

cost 310cu 390cu 355cu 351.67cu

time 170tu 260tu 380tu 270tu

distance 40du 40du 50du 43.32du

According to the agile manufacturing characteristics, manufacturing process characteristics, which is the most important, is time, followed by distance. The cost factor is the least important of the three considered factors. The procedure allows its user to choose whether the solutions obtained should be evaluated with time, distance or cost they generate. There is also a possibility to choose the combination of these characteristics as the fitness function. These values, calculated for random chromosome are respectively presented in table 4. To meet previous deliberations about agility the combination should include the importance of factors combined. The weight of factors depends on the designer and the problem optimised. In this example the importance weight of fitness compounds is the following (these values come of agility characteristics, which make time the most important and cost the least important factors to fulfil unique customers demands): wt = 0.5; wd = 0.3; wc = 0.2; wt - time importance wd - distance importance wc - cost importance Which makes the final value of mixed fitness function (formula 14): F = wt * at + wd * ad + wc * ac F mixed fitness value Which makes fitness value for the example presented: F = 0.5*270 + 0.3*43.3 + 0.2* 351.67 = 135 + 13 + 70.33 = 218.33fu The only problem with combining these factors is the unit the fitness is to be measured with and mixing the three unit types (time, cost, distance), but time, cost and distance can be converted into a fitness unit, of course prior to using them in the fitness counting formula. Combining time, cost and distance in one formula enables taking the manufacturing characteristics into consideration. (14)

6.5.13

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

All the values calculated in the analysed example are taken into consideration while evaluating the fitness and quality of solutions. 5. CONCLUSIONS Agility seems to be the perfect solution for companies as it allows to gain competitive advantage and to be productive. Becoming agile is not only the question of proper management but also proper facility layout. The evolutionary strategies are very useful in solving FLPs (Facility Layout Problems) because they are easy to use, cheap and fast. A solution obtained with evolutionary strategies is sufficient, though not optimal, but there cannot be any optimal solution when the characteristics of products to be manufactured are not known. Although the lack of data when designing agile facility is a vital problem, it can be solved by applying the scheme presented. The development of Agile Facility Layout Design Procedure, along with determining the way of collecting the data needed, was the first step of the research that is one of the authors phd thesis subject. The research is still in progress and the next stage is to evaluate the influence of using average values and random products to design agile facility.. The data used in the procedure is supposed to enable agile facility layout design and the quality of the data and usefulness of the collecting scheme applied can be evaluated only by measuring agility of the layout obtained with the AFLDP. Therefore characteristic to be measured and evaluated is the agility of the solution derived. The tool used for evaluating the agility is the Agility Index. Due to the agility features it is impossible to measure it with numbers and traditional ratings. Measures for communication, interaction etc. are described with linguistic terms which makes fuzzy logic indispensable in agility measuring [11). The Fuzzy Agility Index (FAI) developed by Ching-Torng Lin (Ching-Torng, 2003) is based on customer sensitivity, collaborative relationship, process and information integration. Attributes of these dimensions are assessed and weighed by experts, then fuzzy ratings and weights are aggregated into FAI. Analysing the results of assessment helps to measure the agility and shows the way to agility as well. The FAI used to measure the agility of a solution obtained is based on agility characteristics briefly presented in the paper. We believe that measuring agility of solution derived from AFLDP can show whether the project we are working at was successful or not. REFERENCES Ashley, Steven (1998): Rapid-response design. Mechanical Engineering, Dec97, 119.12, 72, 3p Azadivar, Farhad and Wang, John (Jian) (2000): Facility layout optimisation using simulation and genetic algorithms. International Journal of Production Research, 2000, 38.17, 4369-4383 Emigh, Jacqueline (1999): Agile manufacturing. Computerworld, 8/30/99, 33.35, 56, 1p Fraser, Julie (1995): Finite scheduling and manufacturing synchronisation: Tools for real productivity. IIE Solutions, Sep95, 27.9, 44, 2p,2c Hamammoto, S. Yih, Y. Salvendy G. (1999): Development and validation of genetic algorithm-based facility layout a case study in the pharmaceutical industry. International Journal of Production Research, 1999, 37.4, 749-768 Islier, A.A.(1998): A genetic algorithm approach for multiple criteria facility layout design. International Journal of Production Research, 1998, 36.6, 1549-1569 Mutch, David (1996): Making multiple products is a snap if company is agile. Christian Science Monitor, 3/29/96, 88.86, 8, 1c

6.5.14

Proceedings of the Fifth Asia Pacific Industrial Engineering and Management Systems Conference 2004

Rajasekharan, M. Peters, B.A. Yang, T. (1998): A genetic algorithm for facility layout design in flexible manufacturing systems. International Journal of Production Research, 1998, 36.1, 95-110 Suresh, G. Vinod V.V. Sahu, S. (1995): A genetic algorithm for facility layout. International Journal of Production Research, 1995, 33.12, 3411-3423 Stachowiak A. (2004): Zwinne jednostki wytwarzania. Komputerowo zintegrowane zarzdzanie. Zbir prac pod red. R.Knosali, Wydawnictwa Naukowo-Techniczne, Warszawa 2004, 2, p.476-483 Stachowiak A., Fertsch M.(2004): Evolutionary strategies in agile facility design. Proceedings Engineering of Intelligent Systems February 28 March2, 2004, Madeira, Portugal Tam, K.Y. Chan, S.K. (1998): Solving facility layout problems with geometric constraints using parallel genetic algorithms: experimentation and findings. International Journal of Production Research, 1998, 36.12, 3253-3272 Ching-Torng Lin (2003): Agility Index in Supply Chain. Proceedings 17th International Conference on Production Research August 3-7, 2003 -- Blacksburg, Virginia USA

6.5.15

You might also like