当前位置:首页 >> 机械/仪表 >>

Design of a PCB plant with expert system and simulation approach


Expert Systems with Applications 28 (2005) 409–423 www.elsevier.com/locate/eswa

Design of a PCB plant with expert system and simulation approach
Felix T.S. Chan, H.K. Chan*

/>Department of Industrial and Manufacturing Systems Engineering, University of Hong Kong, 8F, Haking Wong Building, Pokfulam Road, Hong Kong

Abstract This paper presents a case study of using simulation in planning a new Printed Circuit Board (PCB) manufacturing system. The process simulation technique and expert system are proposed to evaluate the design alternatives and logistics management. Simulation results, such as machine utilisation, waiting time, and throughput of system are generated for evaluation. Simulation models were developed using software package SIMPROCESS, which is based on the concept of visual interactive simulation. Besides, the formation of the models is assisted by the expert system package VP-EXPERT. Each approach was found to accurately model the PCB manufacturing plant layout, provide operational and logistics performance in terms of throughput, waiting time, and resource utilisation, and provide a basis for suggesting the best modes of logistics operation. q 2004 Elsevier Ltd. All rights reserved.
Keywords: Expert System; Manufacturing System Design; Printed Circuit Board (PCB); Simulation

1. Introduction Computer simulation is being increasingly accepted by industry as a valuable and practical technique. The technique of simulation in most cases requires the building of a dynamic computer model of a situation or system, which can then be studied giving a realistic view of the proposed real world. Alternative versions of the model can be tested in order to discover, by trial and error approach, which conguration suits the purpose best. Brooks and Tobias (2000) proposed an eight-stage procedure for doing the simplication in the simulation of manufacturing systems. Simulation is a relatively new tool, having evolved over the last 30 years (Smith, 1987). It has been developed rapidly since the early 1970s with the advent of digital computers. Simulation provides a means of testing and enables alternative courses of action to be tested on a model of the real life situation, thereby minimising risks. Szymankiewicz, McDonald, and Turner (1988) described that ‘Simulation involves the construction of a replica or
* Corresponding author. Tel.: C852 2859 7967; fax: C852 2858 6535. E-mail address: hkchan@ieee.org (H.K. Chan). 0957-4174/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2004.12.002

model of the problem on which we experiment and test alternative courses of action. This gives us a greater insight into the problem and places us in a better position from which to seek a solution’. Simulation is commonly employed as most computeraided tools, which can have varying levels of complexity. Their work can carry out with a simple 2D package or equally with a full-colour graphic modelling package. In this connection, simulation can mean anything from a simple mathematical model, taking a matter of hours to build, hugely complex model requiring as much as two man-years to create. The justication for a particular level of detail could be made in purely nancial terms: the savings as a result of the simulation should be equal to or even higher than the cost of carrying out the simulation. However, it is not always possible to quantify savings directly, especially when you are uncertain in what form they are going to be until you have completed the simulation. Experience has shown that the costs of simulation for rst-stage planning are rarely greater than 1% of the capital cost of the system (Hurrion, 1986). This of course is not a rule, but gives a good indication of the magnitude of cost likely to be incurred in getting to a rststage model that has conrmed the feasibility of the proposed system. The model will then be suitable for additional

410

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

detailing in order to test and optimise the system in greater depth. The basic types of simulation are discrete event simulation and continuous variable simulation. The latter type concentrates on a continuous stream of product such as in the process industry or in the ows of information, whereas the former type is more applicable to the manufacture of components, subassemblies and the nal product. There are well over 100 different packages available for discrete event simulation. Davis and Williams (1994) surveyed 14, all of which were obtainable and supported in the UK. During the survey and subsequent research works, they found that the degree of modelling effort and exibility offered were obtainable at varying costs. A useful attempt at grouping the packages had been achieved by Chaharbaghi (1990). Clearly the potential user faces the problem of selecting, from among the packages available, the one which best matches the requirements at some preferred cost. Simulation software selection needs some clearly dened guidelines culminating in a relatively unambiguous decision about which software to purchase. For example, discrete simulation packages such as WITNESS, SIMFACTORY/SIMPROCESS (CACI, 1996), PROMOD, STARCELL, are used to examine material ows, process scheduling, product assembly strategies, warehouse requirements, production ergonomics, logistics strategies and so on (Petropoulakis & Giacomini, 1997). Such tools can provide useful information, which could help the realistic evaluation of production and logistics costs, and hence assist in nalising the product design and the production process requirements thereafter. The software packages employed in this study are SIMPROCESS and VP-EXPERT. SIMPROCESS is a hierarchical and integrated process simulation tool that radically improves productivity for process modelling and analysis. SIMPROCESS is designed for Business Process Reengineering (BPR) and IT professionals of industrial and service enterprises who need to reduce the time and risk it takes to service customers, full demand, and develop new products. Unlike other tools, SIMPROCESS integrates process mapping, hierarchical event-driven simulation, and activity-based costing (ABC) into a single tool. The architecture of SIMPROCESS provides an integrating framework for ABC. ABC embodies the concept that a business is a series of inter-related processes, and that these processes consist of activities that convert inputs to outputs. VP-Expert is basically an inference engine. It can be used to reason about many different knowledge bases. Just as a database manager has built-in templates and procedures for managing large amounts of data, VP-Expert already has built into it the necessary search and managing routines for reasoning with the facts and rules, which can be used in the knowledge base. This paper attempts to use simulation to forecast the performance of a new PCB manufacturing plant. Besides, expert system plays an important role in the simulation

process. All the criteria are pre-set in the expert system and a set of systematic procedures is formed in the determination of plant design. The simulation results of machine utilisation, waiting time, and throughput in each process are used to analyse the performance of the system. This is a very useful and economical method to predict the feasibility of the design. The rest of this paper is organised as follows: Section 2 reviews related literature on simulation and expert systems. Section 3 describes the case study company and problem to be overcome. Section 4 explains the simulation method employed here, and Section 5 concludes the results and discussion.

2. Literature review Many authors performed simulation study in plant and system design (Ashayeri & Gelders, 1988; Brancaleoni, Bugno, & Cavalloro, 1988; Doerr & Magazine, 2000), and logistics ow in supply chain management (Chan, Humphreys, & Lu, 2001), but not many of them can make use of intelligent techniques like expert system. Expert system (ES) solves problems that are normally solved by human ‘experts’ (Jayaraman & Srivastava, 1996). An ES is a computer program that attempts to capture the knowledge and experience of one or more human experts in order to make such expertise available on demand to the user of the program (Friederich & Gargano, 1989). On the other hand, an ES is a problem-solving package that mimics a human expert in a specialised area. An ES can also be dened as a computer program that exhibits, within a specic domain, a degree of expertise in problem solving that is comparable with that of a human expert. ESs need to exploit one or more reasoning mechanisms to apply their knowledge to the problems that are given. Then they need a mechanism for explaining what they have done to the users who rely on them. ESs can be grouped into three major categories depending on their area of applications. The rst category deals with problems of diagnosis. Medical diagnosis has been a widely explored area under this category. The second category deals with systems concerned with design problems. The third category comprises ESs designed for decision-support systems. There are three key types of information transfer in an ES, which differentiate it from other decision-support systems (Mertens & Kanet, 1986). An ES requests relevant information from the user about the problem area. It also offers a recommendation based on the data given by the user and, if a non-expert requests information, it provides the justication for the decision. Gupta and Chin (1989) summarised the four main advantages of ESs as follows: (1) an apparent intelligent behaviour in performing complex tasks;

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

411

(2) the ability to give reasonable justication and explanation to conclusions; (3) the ability to exploit knowledge in an opportunistic way; and (4) the ability to deal with incomplete and imprecise knowledge/data. ESs are one of the most commercially successful branches of articial intelligence. Articial intelligence is an attempt to teach machines the characteristics of human intelligence. Intelligence is associated with the ability to recognise patterns, apply experience and expertise to patterns to solve problems, learn from new experiences and apply judgement when data are incomplete or unavailable. ESs capture the strengths of, and integrate well with, other branches of AI such as natural language processing, pattern recognition and programming environment. Waterman (1985) dened ES as follows: ‘an expert system contains a knowledge base, a dialog structure and an inference engine, which consists of an interpreter and a scheduler. The knowledge base is the collection of the domain knowledge. The dialog structure provides communications and interaction with the user during the operation and processing of the expert system. The inference engine contains the general problem solving knowledge. The interpreter makes decisions on how to apply the rules for inferring new knowledge and the scheduler prioritises the rules in the appropriate order’. ` ` ` ` Kovacs, Mezgar, Kopacsi, Gavalcova, and Nacsa (1994) described the application of expert systems to assist quality control and to help the control of Flexible Manufacturing Systems (FMSs) via simulation. They believed that a closeto-optimal operation of complex, real-time, and stochastic systems such as FMSs could not be achieved by the application of traditional programming. They strongly advocated the use of expert system and articial intelligence techniques in conjunction with sophisticated modelling and simulation. Fu, Occena, Ho, Chang, and Chen (2000) developed a computer-based expert system that can be used as a consulting system by assembly machine builders in order to reduce time deciding what type of machine system and control system to select and what kinds of devices to use in material handling. This expert system collects expert assembly machine knowledge and experiences. Then, an evaluation is performed to ensure accuracy and reliability, and performance is evaluated by building a practical product. The authors concluded that the expert system is found to be a valuable tool, providing signicant information for machine builders. Rao and Gu (1997) proposed a new design methodology and an integrated approach for the design of manufacturing systems. The methodology discussed the steps leading to design of manufacturing systems; the integrated approach suggested ways of integrating the different stages of manufacturing system design using genetic algorithms, another intelligent tools like expert system.

The methodology and approach has been used for an industrial case study and the result has shown that the approach is effective. Raoot and Rakshit (1991) formulated a fuzzy set theory based layout planning approach using the linguistic variables representing ow, control, process, organisation personnel and environment relationships between facility pairs. In this approach distance of facilities are expressed by if–then rules. The authors stated that the possible utility of the proposed approach is promising. Dweiri and Meier (1996) also investigated the application of fuzzy sets theory in facilities layout planning. They suggested that some input variables such as material ow, information ow, and equipment ow can be fuzzied using linguistic variables like high, low, etc. Then they use membership functions for these factors and set fuzzy rules to imitate designer’s decisions. An example of such rules is as follows: IF the Material Flow is Very Low AND the Weight Factor is High, THEN the Rating is lower medium. The authors found that when a fuzzy decision making approach is used to evaluate the traditional facilities layout planning and fuzzy facilities layout planning, the results are better than the other approaches.

3. Problem overview 3.1. Case study company All electronic components must be interconnected and assembled to form a logistics functional and operating system. The design and implementation of these interconnections have evolved into a separate discipline called electronic packaging. Since the early 1950s, the basic building block of electronic packaging is the Printed Circuit Board (PCB) and it will remain that into the foreseeable future (Clyde & Coombs, 1995). After that, PCB manufacturing develops rapidly from single side to double side, and then multi-layer. In these few years, micro-via and ne line technology is one of the major topics in PCB industry. As the rapid development of PCB technology, PCB manufacturing process becomes more and more complicated. As a result, computer simulation can help a lot in the PCB manufacturing system design. The company under study is a world class PCB manufacturer in Hong Kong. It has the international position within fty, and with sales turnover of US$160 million per year. The product types are the parts of hard disks, motherboards, mobile phones, and other electrical appliances. In order to improve the global competition, the development of high technology and the increase of production capacity have to be done. In this connection, the company is planning to establish a new plant that is physically closed to the existing one for these purposes. Since the existing plant is rental and not planned for PCB

412

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

3.3. The SIMPROCESS model In the case study, the SIMPROCESS model consists of ve main parts. The rst part is Inner Process. The second part is the Drilling and Outer Process while the third part is the Coating (Wet Film) process. The forth part is the Surface Treatment and Proling Process and the last part is the Quality Checking Process. In the model, it is assumed that throughput of each machine is generated consistently without sudden breakdown. Moreover, the model is divided into many different layers and animated during the simulation process. In the case study, two validation methods are used. The rst validation is carried out by inputting the target production of 300 kft2/month to each process and the minimum number of machines required. In PCB industry, the production capacity of a plant usually indicates as the area of boards produced in each month, since PCB has different size according to the need of customers. The production of 300 kft2/month means that the plant can produce total area of 300 kft2 of PCB in one month. After this simulation trial, the estimated bottleneck is compared with the simulated bottleneck of each machine types. This method is used to validate the individual parts of the model. The validation result is shown in Table 1, where the items that indicate as star mean the utilisation is greater than 90%. The term of estimated bottleneck is determined by the following equation while the simulated bottleneck is the result of SIMPROCESS model simulation Estimated Bottleneck Z Requested Capacity=Calculated Capacity 3.2. Performance measures The next step is to decide the numbers of each machine types for the manufacturing system design. In order to nd out the optimum numbers of machines, system performance of different combinations is predicted by using simulation technique. The system performances include utilisation of resources, product output, and total waiting time within each process. In fact, a good system design can help to minimise area for buffer, decrease work in progress and lead time, increase machines and operators utilisation, and possibly increase production rate and enhance the logistics ow. In the rst stage of the case study, the minimum number of resources is computed by the rst VP-EXPERT model. This system design is then simulated in the SIMPROCESS model. After the simulation, system performance is determined for the evaluation. The results will be used as input to the second VP-EXPERT model for the generation of next system design. The above steps are repeated until the expected system performance is achieved. Details are explained in the following sections. where ‘Requested Capacity’ is the minimum production capacity required for that machine to achieve 300 kft2 target and ‘Calculated Capacity’ is the total production capacity that can be produced by that machine numbers. As it can be observed that the results of estimated bottleneck are nearly equal to that of the simulated bottleneck, except the one of Automatic Optical Inspection (AOI) and auto-chamfering machines. Nevertheless, the utilisation of AOI and auto-chamfering machines are 86.90 and 83.01%, respectively. The second validation method is based on simulation of the existing plant design. The production capacity of the existing plant is 650 kft2/month. Certainly, the actual number of machines and the monthly materials usage will be inputted into the model. Not only simulated bottleneck of each machine types is compared with the actual one, but also simulated monthly throughput of each process is compared with the actual one. The validation result is summarised in Table 2. The simulated bottleneck of the present plant system is very similar to that of actual one. Regarding to the throughput of each process, the deviation is less than 5%.

Fig. 1. Schematic diagram of the new PCB plant.

manufacturing at the beginning, the logistics ow is not smooth and very inefcient. Certainly, this problem is not willing to be happened again in the new plant. The land area of the new plant is about 90!120 m2. The plant is planned to be eight oors as shown schematically in Fig. 1. PCB manufacturing can be simply divided into many sub-processes. The main processes consist of board cut, inner, pressing, drilling, outer, coating, surface treatment, proling, electronic testing, and visual inspection. In Fig. 1, it can be seen that drilling process is placed in the lowest level of the plant while quality checking (QC) process locates at the highest level. This arrangement is due to the weight of machines and the logistics ow. The drilling machines are the heaviest, so they are usually located at the lowest level. Similarly, electronic testers are the lightest, so they are placed at the highest level. Other processes are placed in the order of logistics ow from bottom to top. After the decision has been made for the process location, the number of each machine types has to be determined.

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423 Table 1 300 kft2 validation result Process Inner Scrubbing and roller coating Exposure Etching AOI Black oxide Pressing X-ray Drilling Drilling Outer Deburring PTH Scrubbing and lamination Exposure Develop PP Etching AOI Wet lm Scrubbing and auto-printing Pre-baking Exposure Develop CM printing Post baking Surface treatment Gold Finger Plating Water rinsing HASL Immersion Gold Proling Routing Punching Auto-chamfering V-cutting Water Rinsing Final QC E-testing Water rinsing Immersion Silver Entek *Resources utilisation O90%. Monthly unit capacity 300,000 120,000 300,000 50,000 400,000 180,000 400,000 23,000 400,000 400,000 300,000 120,000 350,000 300,000 400,000 50,000 300,000 250,000 120,000 350,000 300,000 300,000 80,000 400,000 420,000 150,000 28,000 40,000 50,000 70,000 400,000 50,000 400,000 75,000 75,000 Calculated no. of machines 2 5 2 4(30%) 2 2 1 14 1 1 1 3 1 1 1 2(30%) 1 2 3 1 1 1 2(30%) 1 1(70%) 1(20%) 8(70%) 3(30%) 2(30%) 2(35%) 1 6 1 1(5%) 1(5%) Calculated capacity 600,000 600,000 600,000 200,000 800,000 360,000 400,000 322,000 400,000 400,000 300,000 360,000 350,000 300,000 400,000 100,000 300,000 500,000 360,000 350,000 300,000 300,000 160,000 400,000 420,000 150,000 224,000 120,000 100,000 140,000 400,000 300,000 400,000 75,000 75,000 Requested capacity 600,000 600,000 600,000 180,000 600,000 300,000 300,000 300,000 300,000 300,000 300,000 300,000 300,000 300,000 300,000 90,000 300,000 300,000 300,000 300,000 300,000 300,000 90,000 300,000 210,000 60,000 210,000 90,000 90,000 105,000 300,000 300,000 300,000 15,000 15,000 * * * 83.01% Estimated bottleneck * * * * Simulated bottleneck * * * 86.90%

413

*

*

*

*

* * *

* * *

* *

* *

*

*

Based upon these two validation cases, it can be concluded that the developed model is totally acceptable for the system simulation. In order to nd the best system design, a set of systematic procedures of simulation process should be established (Fig. 2). The procedures start from the rst process—Inner Process. At the beginning, VP-EXERT model 1 (see the following VP-EXPERT model) is used to calculate the minimum number of machines required to meet the requested production capacity of the inner process. An empirical system plan is developed. For convenience, it is called ‘Calculated Plan’. After the Calculated Plan of Inner

Process has been developed, the rst simulation of SIMPROCESS model will be performed and the utilisation of each machine types can be determined. The next step is to see whether this system is able to meet the minimum capacity or not. Then it is necessary to check whether the machines’ utilisation falling into our target range (which is pre-determined from the business strategies) of 60–90% or not. If not, some modications of the system are essential for the next simulation. This can be assisted by the VP-EXPERT model 2 (see the following VP-EXPERT model). The VP-EXPERT model 2 can help to modify the calculated plan into a new system design, this design is

414

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

Table 2 650 kft2 existing plant validation result Process Inner Scrubbing and roller coating Exposure Etching AOI Black oxide Pressing X-ray Monthly unit capacity 300,000 120,000 300,000 50,000 400,000 180,000 400,000 No. of machines 5 11 5 5(10%) 4 4 2 Calculated capacity 1,500,000 1,320,000 1,500,000 250,000 1,600,000 720,000 800,000 Requested capacity Simulated bottleneck * Actual bottleneck *

1,200,000 1,200,000 1,200,000 120,000 1,200,000 600,000 600,000 Total processed in 10 days Actual output in 1 month 600,000 600,000 600,000 600,000 600,000 600,000 600,000 600,000 180,000 Total processed in 10 days Actual output in 1 month 600,000 600,000 600,000 600,000 600,000 600,000 Total processed in 10 days Actual output in 1 month 180,000 600,000 480,000 60,000 420,000 180,000 180,000 210,000 600,000 Total processed in 10 days Actual output in 1 month 600,000 600,000 30,000 30,000 Total processed in 10 days Actual output in 1 month

364,318.5 109,2955.5 * 89.65% 89.39%

– 1,050,000 * * *

Drilling Drilling Outer Deburring PTH Scrubbing and lamination Exposure Develop PP Etching AOI

23,000 400,000 400,000 300,000 120,000 350,000 300,000 400,000 50,000

36 2 2 3 7 3 3 2 5(30%)

828,000 800,000 800,000 900,000 840,000 1,050,000 900,000 800,000 250,000

87.32% 324,226 972,678 85.69% 82.01% 84.95% 84.45% 81.50% 81.39% 307,405 922,215 89.65%

* – 950,000 No Obvious bottleneck

Wet lm Scrubbing and auto-printing Pre-baking Exposure Develop CM printing Post baking

300,000 250,000 120,000 350,000 300,000 300,000

3 3 6 2 3 3

900,000 750,000 720,000 700,000 900,000 900,000

– 900,000 *

Surface treatment Gold nger plating Water rinsing HASL Immersion gold Proling Routing Punching Auto-chamfering V-cutting Water rinsing

80,000 400,000 420,000 150,000 28,000 40,000 50,000 70,000 400,000

3(30%) 2 2(80%) 1(10%) 16(70%) 6(30%) 4(30%) 4(35%) 2

240,000 800,000 840,000 150,000 448,000 240,000 200,000 280,000 800,000

87.49% * 88.44% 267,424 802,272

* * * – 780,000 No bottleneck

Final QC E-testing Water rinsing Immersion silver Entek

50,000 400,000 75,000 75,000

20 3 1(5%) 1(5%)

1,000,000 1,200,000 75,000 75,000

240,219 720,657

– 700,000

*Resources utilisation O90%.

called ‘Plan 1’. The steps are repeated for the next simulation. If Plan 1 still cannot achieve the resources utilisation target of 60–90%, Plan 2 will be suggested by VP-EXPERT model 2. Similarly, Plans 3–5 and so on will be developed according to the feedback from the simulation

results. The procedures will be stopped until the target is achieved or a same plan is repeated to test. If the plan is repeated, the waiting time, production output and cost of the machine are considered to nd out the best plan. Finally, the above steps are repeated for other processes.

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

415

These procedures will be stopped after all the processes have been considered. 3.4. The VP-EXPERT model There are two VP-EXPERT models for this case study. The rst model is used to determine the minimum number of machines required to meet the target capacity. The users are requested to input the throughput of the previous process, t (ft2) and the monthly production capacity, c (ft2) of each machine type. Besides, the percentage of products, p produced by each specic machine is necessary to be inputted. Then the model will apply the following equation to generate the minimum numbers, m of each machine type required. m Z integerpt=c C 1 where integer means to cut off all the decimal place.

The second model is used after the simulation of the rst SIMPROCESS model. The user needs to input the machine numbers, which is obtained from the previous simulation, into VP-EXPERT model 2. Then the program will ask the user whether the previous simulation throughput can meet the minimum capacity or not. Another important data for the model is the machine utilisation, it falls into the range of ‘O 90%’, ‘!60%’, or ‘60–90%’. According to the ow chart of Fig. 2, the model will help to predict the optimum numbers of machines for the next simulation.

4. Results and discussion For each approach, the output of work in progress (WIP) through the system and the utilisation of each resource were determined and recorded. The analysis data of all the processes are summarised in Tables 3–6.

Fig. 2. Flow chart of simulation process.

416

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

Fig. 2 (continued) Table 3 Allocation of resources in inner process Process Time for 300 ft2 (min) Time for different lot number (min) Monthly capacity of one machine Validation 300 kft2 6 (5000) 2 5 2 4(30%) 2 2 1 256,453 384679.5 43.2 (300) 650 kft2 Present plant 3 (7000) 6 17 7 4(10%) 5 6 3 728,637 1092955.5 11.9 (300) 300,000 ft2 new plant Calculated plan 6 (7000) 3 7 3 5(30%) 3 3 2 353,504 530,256 24.4 Plan 1 6 (7000) 4 8 4 5(30%) 3 3 1 358,498 537,747 24.1 Plan 2 6 (7000) 4 8 4 6(30%) 3 3 2 362,624 543,936 23.8 Plan 3 6 (7000) 4 8 4 6(30%) 3 3 1 358,410 537,615 24.1 (300)

Generation (h) Inner Scrubbing and roller coating Exposure Etching AOI Black oxide Pressing X-ray 43.2 108.0 43.2 259.2 32.4 240.0 32.4 10.8 (30 ft2, 12 panels, 1 rack) 300,000 120,000 300,000 50,000 400,000 180,000 400,000 Total processed in 10 days Actual processed in 1 month Next step generation (min)

(300) (300) (300) (Minimum capacity Z520,000)

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423 Table 4 Allocation of resources in drilling and outer processes Process Time for 300 ft2 (min) Time for different lot number (min) Monthly capacity of one machine Validation 300 kft2 650 kft2 present plant 52 3 3 5 11 4 5 3 8(30%) 324,226 972,678 13.3 (300) 300,000 ft2 new plant Calculated plan 24 2 2 2 5 2 2 2 4(30%) 160,397 481,191 Plan 1 Plan 2

417

Drilling Drilling Outer Deburring PTH Scrubbing and lamination Exposure Develop PP Etching AOI

563.5 32.4 32.4 43.2 108.0 37.0 43.2 32.4 259.2

84.5

23,000 400,000 400,000 300,000 120,000 350,000 300,000 400,000 50,000 Total processed in 10 days Actual processed in 1 month Next step generation (min)

14 1 1 1 3 1 1 1 2(30%) 88,418 265,254 43.2 (300)

25 2 2 2 5 2 2 2 4(30%) 161,855 485,565

26 2 2 2 5 2 2 2 4(30%) 161,855 485,565

26.9 26.7 26.7 (300) (300) (300) (Minimum capacity Z460,000)

4.1. Inner process The results obtained from Calculated Plan and Plan 1 of the Inner Process are displayed graphically in Figs. 3 and 4, respectively. Calculated Plan is the system design that has been calculated by VP-EXPERT model 1 in the rst simulation trial. From Fig. 3, only the utilisation of Black Oxide Line and Pressing Machine are within the target range of 60–90%. After the application of VP-EXPERT model 2, only the utilisation of Automatic Optical Inspection (AOI) and X-ray Drilling and Edge Bevelling Machines are out of the target range (Fig. 4). The steps discussed in the previous Section 3.3 of SIMPROCESS
Table 5 Allocation of resources in coating (wet lm) process Process Time for 300 ft2 (min) Time for different lot number (min) Monthly capacity of one machine

model are followed, and it is found that the simulation process stopped at Plan 3 according to the procedures of the simulation process. Considering the resources waiting time (Fig. 5), it is better to choose Plan 2 for inner process. The average total waiting times of Plans 2 and 3 are 1.7 and 2.9 h, respectively. The main difference between these two plans is that there are two X-ray Drilling and Edge Bevelling Machines in Plan 2, but only one machine in Plan 3. Another advantage of using Plan 2 is emergence planning. It is concerned that the whole system would be halted in case of a breakdown of X-ray Drilling and Edge Bevelling Machine in Plan 3. It would be lower risk to have two X-Drilling and Edge Bevelling Machines in Plan 2.

Validation 300 kft2 650 kft2 present plant 4 5 10 3 4 4 307,405 922,215 14.1 (300)

300,000 ft2 new plant Calculated plan 2 Plan 1

Wet lm Scrubbing and auto-printing Pre-baking Exposure Develop CM printing Post baking

43.2 51.8 108.0 37.0 43.2 43.2

300,000 250,000 120,000 350,000 300,000 300,000 Total processed in 10 days Actual processed in 1 month Next step generation (min)

1 2 3 1 1 1 89,472 268,416 43.2 (300)

2

2 3 5 5 2 2 2 2 2 2 148,578 153,046 445,734 459,138 29.1 28.2 (300) (300) (Minimum capacity Z390,000)

418

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

Table 6 Allocation of resources in surface treatment, proling, and nal QC processes Process Time for 300 ft2 (min) Time for different lot number (min) Monthly capacity of one machine Validation 300 kft2 650 kft2 present plant 4(30%) 1(30%) 2(80%) 1(10%) 22(60%) 12(40%) 300,000 ft2 new plant Calculated plan 2(30%) 1(30%) 1(70%) 1(20%) 12(70%) 4(30%) Plan 1 Plan 2

Surface treatment Gold nger plating Water rinsing HASL Immersion gold Proling Routing Punching

162.0 32.4 30.9 86.4 462.9 324.0 77.1 (50 ft2, 5 heads, 4 stacks)

80,000 400,000 420,000 150,000 28,000 40,000

2(30%) 1(30%) 1(70%) 1(20%) 8(70%) 3(30%)

2(30%) 1(30%) 1(70%) 1(20%) 13(70%) 4(30%)

3(30%) 1(30%) 1(70%) 1(20%) 13(70%) 4(30%)

Auto-chamfering V-cutting Water rinsing

259.2 185.1 32.4

50,000 70,000 400,000 Total processed in 10 days Actual processed in 1 month Next step generation (min)

2(30%) 2(35%) 1 84,975 254,925 43.2 (300)

6(30%) 5(35%) 3 267,424 802,272 16.2 (300)

3(30%) 3(30%) 3(30%) 3(35%) 3(35%) 3(35%) 2 1 2 130,931 116,527 132,754 392,793 349,581 398,262 33.0 37.1 32.5 (300) (300) (300) (Minimum capacity Z340,000) 8 9 9 1 2 1 1(5%) 1(5%) 1(5%) 1(5%) 1(5%) 1(5%) 119,580 119,504 119,504 358,740 358,512 358,512 (Minimum capacity Z300,000)

Final QC E-testing Water rinsing Immersion silver Entek

259.2 32.4 172.8 172.8

50,000 400,000 75,000 75,000 Total processed in 10 days Actual processed in 1 month

6 1 1(5%) 1(5%) 89,666 268,998

20 3 1(5%) 1(5%) 240,219 720,657

Fig. 3. Utilisation of inner process machines in calculated plan (By Calculation Only).

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

419

Fig. 4. Utilisation of inner process machines in plan 1.

4.2. Drilling and outer process For Drilling and Outer Process, the procedures are followed until Plan 2 achieves the target of machines’ utilisation 60–90% (Fig. 6). Besides, the average and maximum total waiting times of resources decrease to an acceptable level of 0.50 and 5.6 h, respectively (Figs. 7 and 8). Certainly, the throughput of this plan can achieve to the level of about 485,000 ft2/month while the minimum capacity of this process is 460,000 ft2/month (Table 4). 4.3. Coating process The part of Coating Process is nearly the same as that of Drilling and Outer Process. Initially, only the utilisation of pre-baking oven is over 90% (Fig. 9). After the recommendation of VP-EXPERT model 2, i.e. increases the number of pre-baking oven from 2–3 (Table 5), the utilisation drops back to 68%. Besides, the average and maximum waiting time are reduced from 4.6 and 14.2 h to 0.5 and 2.2 h, respectively (Figs. 10 and 11). 4.4. Surface treatment and proling process Regarding to the Surface Treatment and Proling Process (Fig. 12), the resources of gold plating line, rinsing line for Gold Plating, Immersion Gold Line, and rinsing line for proling in Plan 2 are outside the pre-determined range of utilisation. The utilisation of these machines is below 60%.

Since there is only one Rinsing Line for Gold Plating and One Immersion Gold Line in Plan 2, there is no space for reduction in this case. From Table 6, it is found that the major difference between Plans 1 and 2 is the machine number of Gold Plating Line and Water Rinsing Line for Proling. Since Plan 2 can increase 14% of throughput per month, it is adopted for Surface Treatment and Proling Process. Actually, some of the production lines are tailor-made and the capacity can be adjusted according to customer requirement. However, in this case, the investment costs of plating line and water rinsing line cannot be greatly reduced

Fig. 5. Average waiting time of machines in inner process.

420

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

Fig. 6. Utilisation of drilling and outer process machines in plan 2.

if the capacity is decreased. The tanks’ size of these lines cannot be reduced because of the product size. 4.5. Final quality checking In the process of Final Quality Checking, the simulation process stops at Plan 2, but Plan 1 is adopted nally. The only difference between these two plans is the number of water rinsing line (see Table 6). The reason of choosing Plan 1 is same as that in inner process, i.e. prevent the system halt from the machine breakdown situation. Moreover, another important point found in this part is the number of immersion Silver Line and Entek Line. Although, there is

only one line in these kinds of machine, Fig. 13 shows a very low utilisation rate. It is because only 5% of products are required to pass through these machines. As a result, only one line can satisfy the need of production.

5. Conclusions This paper presents the results of a case study, which involved the use of computer simulation technique in the system design of a new PCB manufacturing plant. The study concentrated mainly on predicting the performance of different system designs. The results enabled the company

Fig. 7. Average waiting time of machines in drilling and outer process.

Fig. 8. Maximum waiting time of machines in drilling and outer process.

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

421

Fig. 9. Utilisation of coating process machines in calculated plan.

to gain a greater understanding of the behaviour of the systems, and hence the most suitable design for the new plant was chosen. With the assistance of expert system, a set of systematic procedures can be dened for the user. By using this set of procedures, a system design that utilises the resources optimally at the range of 60–90%, can be determined. In the study, output of the WIP, total waiting time, utilisation, and cost of resources were considered in the evaluation of the system. The simulation study was used to determine the most appropriate design for the new plant. Nevertheless, not all the requirements can be satised in the design. Plan 2 in inner process is used since the waiting time of resources

can be reduced from 2.9 to 1.7 h. Plan 2 of Drilling and Outer Process, and Plan 1 of Coating Process are adopted because they can satisfy all the targets. Plan 2 of Surface Treatment and Proling Process was chosen because the prot of increased throughput outweighs the investment expenditure. Furthermore, Plan 1 of the Final QC process was adopted in order to prevent system from ‘total blocking’. The simulation models in this case study can be easily modied for the use of other PCB manufacturing plant. Only the number of different machine types in SIMPROCESS model needs to be changed for the simulation of another plant. Certainly, different plants may have different

Fig. 10. Average waiting time of coating process machines.

Fig. 11. Maximum waiting time of coating process machines.

422

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

Fig. 12. Utilisation of surface treatment and proling process machine in plan 2.

product types. In order to apply the SIMPROCESS model to other system designs with different product types, the percentage of products, which passes through certain kinds of machine have to be modied. To conclude, the use of

simulation techniques offered the inexpensive analysis to choose the best manufacturing system in industry. Besides, it can help to propose and evaluate modications so as to improve the logistics performance.

Fig. 13. Utilisation of QC process machine in plan 1.

F.T.S. Chan, H.K. Chan / Expert Systems with Applications 28 (2005) 409–423

423

References
Ashayeri, J., & Gelders, L. (1988). Production planning evaluation of PCB assembly through simulation. Proceedings of fourth international conference simulation in manufacturing (pp. 63–70). Leuven, Belgium. Brancaleoni, C., Bugno, L., & Cavalloro, P. (1988). Italsim: A knowledgebased simulation of a PCB manufacturing plant. Proceedings of fourth international conference simulation in manufacturing (pp. 139–150). Italy. Brooks, R. J., & Tobias, A. M. (2000). Simplication in the simulation of manufacturing systems. International Journal of Production Research, 38(5), 1009–1027. CACI (1996). SIMPROCESS—User’s manual. CACI Products Company. Chaharbaghi, K. (1990). Using simulation to solve design and operational problems. International Journal of Operations and Production Management, 10(9), 89–105. Chan, F. T. S., Humphreys, P., & Lu, T. H. (2001). Order release mechanisms in supply chain management: a simulation approach. International Journal of Physical Distribution and Logistics Management, 31(2), 124–139. Clyde, F., & Coombs, J. R. (1995). Printed circuits handbook. New York: McGraw-Hill. Davis, L., & Williams, D. (1994). Evaluating and selecting simulation software using the analytic hierarchy process. International Journal of Manufacturing Technology Management: Integrated Manufacturing Systems, 5(1), 23–32. Doerr, K., & Magazine, M. J. (2000). Design, coordination and control of hybrid factories. International Journal of Operations and Production Management, 20(1), 85–102. Dweiri, F., & Meier, F. A. (1996). Application of fuzzy decision-making in facilities layout planning. International Journal of Production Research, 34(11), 3207–3225. Friederich, S., & Gargano, M. (1989). Expert systems design and development using VP-EXPERT. San Francisco, NY: Wiley. Fu, H. P., Occena, L. G., Ho, L. H., Chang, T. H., & Chen, K. L. (2000). Expert system for automated assembly machine design. International

Journal of Manufacturing Technology Management: Integrated Manufacturing Systems, 11(6), 405–416. Gupta, Y. P., & Chin, D. C. W. (1989). Expert systems and their applications in production and operations management. Computers Operations Research, 16(6), 567–582. Hurrion, R. D. (1986). Simulation: Applications in manufacturing. Kempston, Bedford: IFS (Publications). Jayaraman, V., & Srivastava, R. (1996). Expert systems in production and operations management: Current applications and future prospects. International Journal of Operations and Production Management, 16(12), 27–44. ` ` ` ` Kovacs, G. L., Mezgar, I., Kopacsi, S., Gavalcova, D., & Nacsa, J. (1994). Application of articial intelligence to problems in advanced manufacturing systems. International Journal of Computer Integrated Manufacturing Systems, 7(3), 153–160. Mertens, P., & Kanet, J. J. (1986). Expert systems in production management: An assessment. Journal of Operations Management, 6(4), 393–404. Petropoulakis, L., & Giacomini, L. (1997). A hybrid simulation system for manufacturing processes. International Journal of Manufacturing Technology Management: Integrated Manufacturing Systems, 8(4), 189–194. Rao, H. A., & Gu, P. (1997). Design methodology and integrated approach for design of manufacturing systems. International Journal of Manufacturing Technology Management: Integrated Manufacturing Systems, 8(3), 159–172. Raoot, A. D., & Rakshit, A. (1991). A fuzzy approach to facilities lay-out planning. International Journal of Production Research, 29(4), 835– 857. Smith, M. R. (1987). A case study on the application of simulation to a car assembly line. Proceedings of the third international conference on simulation in manufacturing (pp. 207–234). Italy. Szymankiewicz, J., McDonald, J., & Turner, K. (1988). Using simulation to solve problems. London: McGraw-Hill. Waterman, D. A. (1985). A guide to expert systems. Reading, MA: Addison-Wesley.


相关文章:
PCB数字地和模拟地
PCB数字地和模拟地_信息与通信_工程科技_专业资料。PCB数字地和模拟地数字地和模拟地★数字地与模拟地的区别 简单来说,数字地是数字信号的对地,模拟地是模拟...
Altium designer 仿真具体步骤
Altium designer 仿真具体步骤_电子/电路_工程科技_...PCB Project ,创建一个 PCB 工程并保存。 2) 在...“Examples\CircuitSimulation\Filter”文件夹中,复制...
毕业论文全文模板自动化专业2012
loopsimulation system, whichdesign based on LabVIEW...and simulationabout thecontrolledplanton the LabVIEW...界面截图、原理图、PCB 版图,状态机图、流程图等。...
PCB Design-General Guidelines
PCB Design-General Guidelines_能源/化工_工程科技... Identifying the major functions of a system/...and simulation may be performed on all or ...
毕业论文全文模板自动化专业2010
A method for modeling and simulation about the controlled plant on the ...PCB 版图 论文题目(五号字) 图 2-5 原理图 说明:在 Altium Designer 中,...
工业工程英语术语集合
systems engineering 系统工程 systems simulation 系统...PCB Printed Circuit Board 印刷电路板 CAR ...进口 Plant 装置 Facility 设施 Unit 装置/单元 ...
高速pcb设计与仿真内容
高速pcb设计与仿真内容_调查/报告_表格/模板_实用...Simulation in SQ Signal Explorer Expert Simulating ...Simulation SSN Waveform Analysis System-Level ...
Altium Designer学习笔记
使用的 时候,在右下角 System 菜单选择 Snippet ...另外还可以在右下 角添加 footprint、simulation 等...(altium designer 6.x)中设置 PCB 大小先在 ...
Allegro PCB SI - 一步一步学会前仿真
Allegro PCB SI: 一步一步学会前仿真 Learn Allegro PCB SI Pre-simulation ...//www.jedec.org/system/files/docs/design/DDR3/PC3-RDIMM_V072_RC_B0_...
AUTODESK 2013全系列产品密钥
Design 2013 424E1 Autodesk Quantity Takeoff 2013 426E1 AutoCAD Plant 3D ...Simulation Moldflow Adviser Standard 2013 571E1 Autodesk Simulation Moldflow ...
更多相关标签:
plant simulation | plant simulation下载 | plant simulation教程 | plant simulation 12 | plant simulation论坛 | plant simulation 9.0 | plant simulation v12 | plantsimulation |