Recently, cellular manufacturing systems for intensive operations have been introduced into parts assembly workshops. In assembly operations, a worker plays the role as a parts supplier from shelves. Therefore, it is required to estimate the cycle time for his assembly operation including parts-supplying time. In this case, some parts shelves are set at higher positions than the height of the worker's shoulders or at a lower position than that of his waist, so that his hand or hands must be moved upward or downward. The motions of his hand (s), body and feet become complex, and in addition, must be done simultaneously. Consequently, the estimation of his move-motion time becomes rather complicated. This paper presents a method for calculating the upward or downward move-motion time during parts-supplying operations. Experiments were conducted on two types of parts boxes, that is, smaller ones to be carried by one hand and larger ones to be carried by both hands, placed on parts shelves of four heights above floor-level. In the experiments, upward move-motion time increases as the height of the parts shelf rises from floor-level and downward move-motion time decreases as the height of the parts shelf rises from floor-level. Based on the motion characteristics mentioned above, it is concluded that upward and downward move-motion time can be estimated by an equation with variables indicating the heights of the parts shelves above floor-level. As a result of the experiments, with due consideration of simultaneity, move-motion-experimental equations are proposed for estimating upward and downward move-motion time in standing-posture operations. It is found out, from the experimental equations proposed, that upward and downward move-motion time is least near the height of the worker's waist.

View full abstract

In a previous paper[1], an optimal control problem for a failure-prone multi-stage production line that minimizes expected total cost per unit time has been formulated as an undiscounted Markov decision process. It, however, is difficult to exactly solve the problem due to the curse of dimensionality. In recent years, several algorithms in the field of reinforcement learning, or neuro-dynamic programming (NDP), have been developed to overcome the curse of dimensionality. In [1], the simulation-based modified policy iteration method (SBMPIM) has been proposed as a new algorithm of NDP and the expected total costs per unit time of JIT production lines with optimal numbers of kanbans have been shown to rise by more than 7 percent in comparison with the expected total costs per unit time computed by the SBMPIM. This paper extends the failure-prone multi-stage production line to a failure-prone multi-stage production and distribution system and improves the SBMPIM algorithm in the convergence property and a stopping criterion. The expected total costs per unit time of pull production systems such as the kanban, base stock, CONWIP and hybrid systems, which are adjusted optimally, are numerically compared with the expected total costs per unit time computed by the improved SBMPIM. It is shown that the base stock system is best of the pull systems and that the expected total costs of the base stock (kanban) system can be reduced by 5〜24 (11〜30) percent by using near-optimal control policies computed by the improved SBMPIM.

View full abstract

Recently, the rapid growth of data-acquisition technology and the use of online computers for process monitoring have led to an increased interest in the simultaneous surveillance of several related quality characteristics or process variables. In general, related quality characteristics are assumed to be distributed as multivariate normal random variables. As a result, the multivariate control chart for the mean vector has been studied extensively in many works. Popularly, in the case that the quality characteristics are distributed as univariate normal random variables, the mean and variance are simultaneously treated as the objectives of surveillance. From this point, in the case that the quality characteristics are distributed as multivariate normal random variables, the mean vector and variance-covariance matrix should be simultaneously treated as the objectives of surveillance. Kanagawa et al. have proposed a (x^^-, s) simultaneous control chart that enables the user to monitor both changes in the mean and variance in a process simultaneously based on Kullback-Leibler information when the quality characteristics are distributed as univariate normal random variables. In this study, as an extension of the (x^^-, s) simultaneous control chart, we propose a multivariate (X^^-, S) simultaneous control chart that enables the user to monitor both changes in the mean vector and variance-covariance matrix simultaneously. Further, the evaluation of the power for the proposed multivariate (X^^-, S) simultaneous control chart is also considered.

View full abstract

This paper presents a strategic approach for effectively differentiating the quality of a product from its competing products. The proposed approach utilizes quality function deployment (QFD) to capture customer requirements, technical requirements and quality characteristics of the product. The newly introduced technical requirements are concerned with the feasibility of the target quality (i.e., the target values of the quality characteristics) to be determined. Then, the approach adopts fuzzy logic to quantify the relationships between the quality characteristic values and the satisfaction levels of the requirements, and to evaluate the technical feasibility of the target quality. As a result, it becomes possible to capture the target quality determination for the product as a multi-objective optimization problem of enhancing the satisfaction levels of the customer requirements while ensuring the technical feasibility of the target quality. Then, in order to deal with this problem in a strategic way, the condition required for differentiating the product quality from its competing products is given. Further, a function which measures the extent of the differentiation is defined according to the preferences among customer requirements and which customer requirements are chosen as the order winning criteria for the product. Accordingly, the multi-objective quality planning problem is translated into a problem of maximizing the extent of the differentiation under some feasibility constraints. Finally, a simple example illustrates how the proposed approach works.

View full abstract

A methodology for extracting and learning geometric features has been studied. At first, an algorithm was developed to create voxels (volume pixels) easily from three-dimensional CAD data. A voxel is a picture element in a three-dimensional coordinate system. Using this algorithm, CAD data such as STL-formatted data can be converted to very fine voxels in a few seconds. Then, distance values from the surface of each voxel are calculated. At the same time, a Ds (distance from surface) map can be obtained. In the next step, skeletons of shapes are extracted as polygons from the mapped data. Voxels with the maximum Ds among neighboring voxels are selected as skeleton candidates. After the selection, the candidates are converted to straight lines or circular rings. They are then represented by several vectors and stored as a tree structure. A standard tree involves, for example, four levels and each branch has four descendants. Each parent branch has the same number of descendants. The attributes include, for example, the scaled volume, connection strength, and scaled X, Y, Z coordinates. These vectors are the input to a skeleton classifier, which is constructed on the basis of the back propagation neural network model. The proposed approach was implemented on a personal computer. A viewer was developed to display the skeletons clearly in three-dimensional shapes. Several parts were selected to demonstrate the classification capability for this methodology. Here, the skeletons extracted directly from the three-dimensional shapes were used. The back propagation neural network model should be trained using some representative shapes. It was found that this method is practically useful through these experiments.

View full abstract

In spite of increasingly tough managerial conditions, improvements on the clerical operations at banks is not well advanced since bank operations are special in the sense that they handle money directly. It is difficult to apply a traditional work process analysis method because bank operations are incurred with unexpected correspondences with customers face-to-face or over the phone. On the other hand, a characteristic of bank operations is the number of stamps for each and every document and voucher. This fact implies the heavy burden on workers and the existence of wasted time that multiple stamps have for the same purposes, and so can be eliminated. This paper discusses 'stamping' and proposes a new method for analyzing clerical work taking the case of bank operations as an example. The paper applies a proposed method to an actual withdrawal operation at bank counters and describes that various checkpoints and ideas for improvement that can be derived in practice, thereby proving that the proposed method is effective.

View full abstract