Multi-criteria process planning for assembly of printed circuit boards | |
| Author | Nguyen Van Hop |
| Call Number | AIT Diss. no. ISE-00-02 |
| Subject(s) | Production planning Printed circuits industry |
| Note | A disse1iation submitted in partial fulfillment of the requirements for the degree of Doctor of Engineering, School of Advanced Technologies |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. ISE-00-02 |
| Abstract | This research deals with process planning for printed circuit board (PCB) assembly. A line of machines and a group of PCB types are taken into account. Each PCB requires a number of components. Components are loaded and allocated on the feeder rack of machine. Assembly arm assembles components on the board, which is fixed on the machine table. Assembly arm, feeder rack and machine table moves independently together with different speeds. Three main issues are commonly emphasized in the assembly of PCBs, namely: (i) Grouping: form machine groups, form component and PCBs families. (ii) Allocation: allocation of components and PCBs families to machines. (iii) Assignment and sequencing: assignment of components to the feeder slots on the feeder rack and sequencing of placement operations for each PCB on each machine (feeder assignment and assembly sequence). This research focuses on the grouping aspect in terms of component grouping only. The issue of machine grouping is given in the form of an assembly line. The board grouping aspect and allocation of components, PCBs families to machines in the form of component loading and board sequencing are considered in the setup problem. Before tackling the third issue, an additional issue of point specification is taken into account. In this problem, there is a need to determine the optimal picking and placement points when the feeder assignment and assembly sequence have been known Finally, the issue of feeder assignment and assembly is handled. First, components are clustered into disjoint groups based on their multiple attributes. Each group is classified at a certain level of attributes. The difficulty lies with the conflict of criteria when selecting a certain component into a group. In order to overcome this difficulty, the proposed approach uses the concept of fuzziness, which expresses the degree of concordance on the decision. These fuzzy sets are pairwisely compared to select the best one when selecting a component into a group. The final decision as to which component belongs to what group is made on the basis of the highest dergree of concordance among these fuzzy values of decisions. Thus, the conflict is resolved and groups of components are formed. The loading and board sequencing strategies in setup problem are also studied in a multiple criteria fashion. Normally, the setup problem is often solved for the case of one machine. Several criteria should be included when the problem is considered for the whole assembly line. There are the requirements of load balancing, setup time saving and ยท maximization of machine utility. The new approach of combining multi-attribute decision making (MADM) and multi-objective decision making (MODM) is proposed to solve this problem. Two MODM models are used to describe the loading criteria for each board type position in the assembly queue of PCBs. First, the boards are selected based on the capacity of the line. Then, The board sequence is determined based on the solution of the model, which describes the first position in the assembly queue, and the trade-off between objective criteria value of the solved model and the attribute of the board type by a MADM method. When the assembly board sequence is determined, the optimal loading plan is tackled by the solution of both MODM models. In addition, a TSP-based heuristic approach is also developed to find the solution for the board sequence problem in order to minimize the number of component changeovers for the case of one machine. First, the board sequence is determined by a TSP-based formulation and algoritlun. Then the loading plan is found based on the well-known rule of Keep Tool Needed Soon (KTNS). The numerical results show that the new approach improves the previous one by around 10% of total number of component changeovers. Furthermore, the machine-oriented criteria have arisen for this kind of scheduling problem as PCB sequencing in this case. When two previous issues had been tackled, the next question is how to determine the optimal picking points (or placement points). A new model extends the point specification problem based on the dynamic pick and place (DPP) model, namely, extended dynamic pick and place (EDPP) model. In DPP model, both magazine and board move along a certain axis while the robot arm moves between dynamic 'pick' and ' place' points. The EDPP determines the picking/placement point coordinates from consideration of point relationship in the whole system. The principle of EDPP is the willingness to pay extra cost for robot motion in terms of travelling time if in the next cycle the feeder rack or board travelling time can be saved. EDPP considers all objects' motion in the system (robot motion, feeder rack motion and board motion) rather than only robot motion as DPP. The system is analyzed from the simple machine configurations (fixed board, dynamic feeder rack, robot arm and fixed feeder rack, dynamic board, robot arm) to the complex one (dynamic board, feeder rack and robot arm) for the Euclidean robot motion in order to establish the mathematical model for the system. The network is built to handle the problem. It is shown that for the general case of machine setting (dynamic board, feeder rack and robot arm), the resulting graph is a binary tree. Hence, it is intractable to give the optimal solution. A heuristic approach is proposed. The numerical results prove that the EDPP is better than DPP in terms of total assembly time. In addition, other cases of rectilinear and Tchebyshev robot motions are also studied. For the case of fixed board, dynamic feeder rack and Tchebyshev robot motion, a polynomial algorithm is proved to give the optimal solution in O(n4 ) time. Finally, the feeder assignment and assembly sequence problem in PCB assembly is taken into consideration. The problem is decomposed into subproblems of feeder assig1unent and assembly sequence separately and the combined one under diffirent scenarios of machine settings and robot motions. The analysis of the problem leads to a number of nonlinear MODM models. These models are linearized and solved by heuristics which are based on network theory and interval graph. These heuristics are tested and found to give the closed optimal solution. In addition, for the general case of dynamic board, feeder rack and Euclidean robot motion, another heuristic is developed with the twin objectives of minimizing feeder rack travel time and minimizing board travel time. The problem uses Dynamic Pick-and-Place (DPP) model where robot arm, board and magazine move together with different speeds based on the relative coordinates between consecutive assembled points. The difficulty of the problem is that the feeder assignment depends on assembly sequence and vise versa. A new approach is proposed to trade-off between two strategies, namely assembly by area and assembly by component types. Assembly sequence and feeder assignment problems are also studied under different operating scenarios. Since most of the problems' nature is NP-complete problems, these heuristic algorithms are appropriate. Overall, the contribution of this disse11ation is on the development of methodologies to solve the PCB assembly planning under a hierarchical structure of sub-problems. In these hierarchical sub-problems, a grouping method using fuzzy set theory resolves the conflict between decisions to form groups of components. A setup problem tackles the issues of board sequence and component loading in the form of MCDM models and some heuristics. The special features in these procedures are not only the use of both MODM and MADM to handle the problem but also the achievement of better results and the open machine-oriented scheduling consideration. A contribution of this research is also in the development of a new model for picking and placement point specification problem, namely EDPP model which is better than the previous model, DPP model, in terms of total assembly time saving. This research also develops, analyzes and proposes models and heuristics to solve the feeder assginment and assembly sequence un der different scenamios of operation settings. The effectiveness of these procedures have been illustrated by numerical examples. |
| Year | 2000 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. ISE-00-02 |
| Type | Dissertation |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Industrial Systems Engineering (DISE) |
| Academic Program/FoS | Industrial Systems Engineering (ISE) |
| Chairperson(s) | Tabucanon, M.T.; |
| Examination Committee(s) | Batanov, D.N.;Bohez, E. L.J.;Nagarur, N.N.;Egbelu, P. J.; |
| Scholarship Donor(s) | Government of Austria; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2000 |