Search

Fluid dispensing is a popular process in the semiconductor manufacturing industry, commonly being used in die-bonding as well as microchip encapsulation of electronic packaging. Modelling the fluid dispensing process is important to understanding the process behaviour as well as determining the optimum operating conditions of the process for a high-yield, low-cost and robust operation. In this paper, an approach to integrating neural networks with a modified genetic algorithm is presented to model the fluid dispensing process for electronic packaging. The modified genetic algorithm is proposed by incorporating the crossover operator with an orthogonal array. We compare the modified genetic algorithm with the standard genetic algorithm. The results indicate that a better quality encapsulation can be obtained based on the modified genetic algorithm. [ABSTRACT FROM AUTHOR]

In apparel manufacturing, accurate upstream fabric-cutting planning is crucial for the smoothness of downstream sewing operations. Effective and reliable fabric-cutting schedules are difficult to obtain because the apparel manufacturing environment is fuzzy and dynamic. In this paper, genetic algorithms and fuzzy-set theory are used to generate fault-tolerant fabric-cutting schedules in a just-in-time production environment. The proposed method is demonstrated by two cases with production data collected from a Hong Kong-owned garment production plant in China. Results of the two cases preliminarily show that the genetically improved fault-tolerant schedules effectively satisfy the demand for downstream production units, guarantee consistent and reliable system performance, and also reduce production costs through reduced operator idle time. More cases will be conducted in order to further validate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

In Quality Function Deployment, determination of the target values for engineering requirements is the most difficult process. A fuzzy optimization model is presented for the determination of target values for engineering requirements in Quality Function Deployment. An inexact genetic algorithm approach was introduced to solve the model that takes the mutation along the weighted gradient direction as a genetic operator. Instead of obtaining one set of exact optimal target values, the approach can generate a family of inexact optimal target values setting within an acceptable satisfaction degree. Through an interactive approach, a design team can determine a combination of preferred solution sets from which a set of preferred target values of engineering requirements based on a specific design scenario can be obtained. An example of car door design is used to illustrate the approach. [ABSTRACT FROM AUTHOR]

Currently, determination of the initial process parameter settings for injection moulding is mainly performed by moulding personnel, and the effectiveness of the parameter setting is largely dependent on the experience of these personnel. In this paper, an intelligent hybrid system, called HSIM, is described, which is used to determine a set of initial process parameters for injection moulding based on the artificial intelligence (AI) techniques, case-based reasoning (CBR), and hybrid neural network (NN) and genetic algorithm (GA). HSIM can determine a set of initial process parameters for injection moulding quickly, without relying on expert moulding personnel, from which moulded parts free from major moulding defects can be produced. [ABSTRACT FROM AUTHOR]

In traditional product development process, supplier selection is performed in the production stage when the product design stage has finished. Recently, some scholars found that decoupling the supplier selection and product design may lead to infeasible or suboptimal solutions, and a number of research papers proposed methods integrating supplier selection into product family design. In this research, a one-step product family optimization model integrating supplier selection decision is established based on multinomial logit consumer choice rule. As a kind of widely used discrete choice model, multinomial logit consumer choice rule is regarded to be more flexible and suitable to formulate consumer choice behavior. Genetic algorithm is developed to solve the established nonlinear optimization model. The optimization result under multinomial logit rule is compared with that under the deterministic choice rule in the case study and sensitivity analysis of the model is performed. [ABSTRACT FROM PUBLISHER]

Product line design is a key decision area that a product development team has to deal with in the early stages of product development. Previous studies of product line design have focused on single-objective optimization. However, several optimization objectives may be simultaneously pursued, and the solutions that can address the objectives are required in many practical scenarios. In this research, we propose a one-step multiobjective optimization approach for product line design. The proposed optimization model has three objectives: 1) maximizing the market share of a company's products; 2) minimizing the total product development cost of a product line; and 3) minimizing the total product development cycle time. A curve-fitting method is introduced into the part-worth utility models so that the optimization model can be applied to products with level-based attributes and attributes that have continuous values. A multiobjective genetic algorithm is adopted to solve the optimization model, obtaining a set of nondominated solutions. With the solutions, a new product development team can select a preferred solution interactively in a 2-D graph. An example of the optimal design of a product line of digital cameras is used to illustrate the proposed approach. [ABSTRACT FROM PUBLISHER]