Andrew W. H. Ip

A Hybrid Genetic Algorithm for a Type of Nonlinear Programming Problem

Abstract Based on the introduction of some new concepts of semifeasible direction, Feasible Degree (FD1) of semifeasible direction, feasible degree (FD2) of illegal points ‘belonging to’ feasible domain, etc., this paper proposed a new fuzzy method for formulating and evaluating illegal points and three new kinds of evaluation functions and developed a special Hybrid Genetic Algorithm (HGA) with penalty function and gradient direction search for nonlinear programming problems. It uses mutation along the weighted gradient direction as its main operator and uses arithmetic combinatorial crossover only in the later generation process. Simulation of some examples show that this method is effective.

Parallel multi-population Particle Swarm Optimization Algorithm for the Uncapacitated Facility Location problem using OpenMP

Parallel multi-population particle swarm optimization (PSO) algorithm using OpenMP is presented for the uncapacitated facility location (UFL) problem. The parallel algorithm performed asynchronously by dividing the whole particle swarm into several sub-swarms and updated the particle velocity with a variety of local optima. Each sub-swarm changes its best position so far of to its neighbor swarm after certain generations. The parallel multi-population PSO (PMPSO) algorithm is applied to several benchmark suits collected from OR-library. And the results are presented and compared to the result of serial execution multi-population PSO. It is conducted that the parallel multi-population PSO is time saving, especially for large scale problem and generated more robust results.

Heuristics for Joint Decisions in Production, Transportation, and Order Quantity

An attempt is made to tackle joint decisions in assigning production, lot size, transportation, and order quantity for single and multiple products in a production-distribution network system with multiple suppliers and multiple destinations. The approach hinges on providing an optimized solution to the joint decision model (JDM) through a two-layer decomposition (TLD) method that combines several heuristics. By combining the Lagrange multipliers and introducing a number of artificial variables into the two-layer decomposition, a Lagrange relaxation decomposition (LRD) method with heuristics is developed to solve multiproduct joint decision problems (JDM-M). Using the LRD, the JDM-M model is solved by decomposing into two subproblems in two layers. The first layer is the joint decisions in assigning production, transportation flow, and lot size (APLS-TF) using the assignment heuristic AH-M. The second layer is the joint decisions in transportation and order quantity (TOQ-M) using a revised BH heuristic. Combined with Lagrange multipliers, the APLS-TF model takes into consideration the transportation costs together with production costs when it assigns annual production among suppliers. In essence, the algorithm assigns annual production simultaneously with annual transportation flows. Simulations on different sizes of problems and problems with large variances in data have shown that the LRD is effective, and in general more effective than the TLD.

Heuristics-based integrated decisionsfor logistics network systems

Abstract In the global manufacturing environment, decisions involvingproduction activities and distribution should be made in an integrative manner for substantial savings in the logistics costs. An integrated decision on production assignment, lot sizing, transportation, and order quantity for a multiple-suppliers/multiple-destinations logistics network in a global manufacturing system is discussed in this paper. The integrated decision process can be viewed as a two-layer decision. The first layer is the combined decisions of production assignment and lot sizing (APLS-M), while the second layer is the combined decisions of transportation and order quantity determination (TOO-M). A two-layer decomposition method combining two heuristics is hence developed to solve the integrated decision model (IDM-M). An assignment heuristic and a transportation heuristic are designed to solve the decision layers APLS-M and TOQ-M, respectively. Simulations were conducted on a practical example in an electronics manufacturing enterprise together with different sizes of problems. The results indicate that the two-layer decomposition method with heuristics is effective and provides a practical way to solve medium and large-scale integrated decision problems.

An application of fuzzy sets to process performance evaluation

The successfulness of total quality management is dependent on the initiative of employees to apply their problem solving techniques continuously. However, the performance evaluation is quite subjective since it relies on the individual judgements of supervisors who have different perceptions of the process performance. This paper presents a performance measurement system to determine the priority in execution of quality improvement according to the performances of processes. Using a model containing a set of strategy‐related performance measures, misdirection of implementation of improvement projects due to subjective evaluation of process performance can be reduced. A fuzzy set of quality measures is introduced to calculate numerical quality rating for processes. The model is simulated and experimented on with a conventional improvement implementation practice. It is expected that the proposed model is superior in the major competitive capabilities including quality, cost flexibility, delivery reliability and speed.

Synchronized Production and Transportation Planning Using Subcontracted Vehicles in a Production-Distribution Network

Abstract Effective management of interfaces between procurement, supply, production and distribution for higher efficiency in the supply chain is an important issue in global manufacturing, where the synchronization of production and transportation planning represents important savings in operational costs. This paper focuses on the synchronization of production planning and transportation planning in a production distribution network, where transportation is subcontracted to a professional transportation enterprise (PTE) for vehicle-hiring. Dynamic and flexible numbers of vehicles are used to cater for fast changing market demands. Thus, the number of vehicles to be hired is viewed from the planning point of view as an operational decision considered simultaneously with production and transportation planning. A mathematical model – SPTP/MTDS – for synchronized production and transportation planning under multiple times and direct shipping strategy (MTDS) is discussed, and a Lagrange relaxation decomposition-based two layer decision procedure (LRD-TLDP) is developed. By introducing artificial decision variables and Lagrange multipliers, SPTP/MTDS is decomposed into a production decision sub-problem (SPTP-PD), and a distribution decision sub-problem (SPTP-DD). A priority-based assignment heuristic and a partial chain-based genetic algorithm are developed to solve SPTP-PD and SPTP-DD, respectively. An illustration of the application of the model in an electronic appliance manufacturing enterprise in China is presented.

A note on Efficient feasibility testing for dial-a-ride problems

Hunsaker and Savelsbergh [B. Hunsaker, M.W.P. Savelsbergh, Efficient feasibility testing for dial-a-ride problems, Operations Research Letters 30 (2002) 169-173.] developed a linear-time algorithm to verify the feasibility for dial-a-ride problems. However, this algorithm may incorrectly declare infeasibility due to ride time constraints in some cases. We propose a revised procedure to address this flaw, but in an O(n^2) worst-case time.

Fast multi-template matching using a particle swarm optimization algorithm for PCB inspection

Template matching is one of the image comparison techniques which is widely applied to determine the existence and location of a component within a captured image in the printed circuit board (PCB) industry. In this research, an efficient auto-detection method using a multi-template matching technique for PCB components detection is described. In many cases, the run time of template matching applications is dominated by repeating the similarity calculation, locating multi-templates, and exploring of the optimum result. A new approach using accelerated species based particle swarm optimization (SPSO) for multi-template matching (MTM) is proposed. To test its performance, our proposed SPSO-MTM algorithm is compared with other approaches by using the real captured PCB image. The SPSO-MTM method is proven to be superior to the others in both efficiency and effectiveness.

Petri Net-based workflow management systems for in-process control in a plastic processing plant

Abstract To manage a complex system, an operation model is proposed for evaluating the effectiveness and efficiency of each possible plan. During the planning process, an easy-to-use modeling system for operation modeling could be very useful for assisting the shop floor manager to understand and predict the performance of operation plans. In this paper, the Petri Net is used as a tool to realize the proposed modeling concept. Derived from the basic Colored Petri Net (CPN) language, a Modular Colored Petri Net (MCPN), composed of a number of CPN Modules (CPNMs), is proposed for this adaptive modeling approach. The CPN in each CPNM represents the discrete event logic of the corresponding operation. Despatching rules of the material and activities are modeled by the components and functions of the CPNs in these CPNMs, and the dynamics of the system can be modeled by the transition firing and connections amongst them. In this way, the time and cost for the development and maintenance of the shop floor operation model can be reduced. The application of the system in in-process control in a plastic processing plant is illustrated in the paper.

Professional Practice and Innovation: RF-Medisys - a Radio Frequency Identification-based Electronic Medical Record System for Improving Medical Information Accessibility and Services at Point of Care

This paper presents an innovative electronic medical records (EMR) system, RF-MediSys, which can perform medical information sharing and retrieval effectively and which is accessible via a ‘smart’ medical card. With such a system, medical diagnoses and treatment decisions can be significantly improved when compared with the conventional practice of using paper medical records systems. Furthermore, the entire healthcare delivery process, from registration to the dispensing or administration of medicines, can be visualised holistically to facilitate performance review. To examine the feasibility of implementing RF-MediSys and to determine its usefulness to users of the system, a survey was conducted within a multi-disciplinary medical service organisation that operates a network of medical clinics and paramedical service centres throughout Hong Kong Island, the Kowloon Peninsula and the New Territories. Questionnaires were distributed to 300 system users, including nurses, physicians and patients, to collect feedback on the operation and performance of RF-MediSys in comparison with conventional paper-based medical record systems. The response rate to the survey was 67%. Results showed a medium to high level of user satisfaction with the radiofrequency identification (RFID)-based EMR system. In particular, respondents provided high ratings on both ‘user-friendliness’ and ‘system performance’. Findings of the survey highlight the potential of RF-MediSys as a tool to enhance quality of medical services and patient safety.