In correspondence analyses, the class of object is called the pattern. The aim of the analysis is to give numerical values to patterns and categories so that their correlation is high. The least absolute value estimation, as an estimation alternative of the conventional least-squares regression, has been studied in the robust regression literature. We formulate a new correspondence analysis based on least absolute value estimation with some generalizations and propose a solution algorithm by the interior point method of mathematical programming. The computational results by the regular technique and the new technique proposed are compared. In the numerical example, it is shown that the classification results by the proposed method are clearer than conventional results. Senior-simulation is for youngsters to experience the physical handicap of advance-aged persons. We analyzed how the psychological feelings and evaluations of youngsters varied to senior-simulation in terms of personalities and interests.

View full abstract

Tasks relying exclusively on expertised skill are considered one of the major factors impairing the efficiency of white-collar jobs. This paper suggests that even a novice is able to gain the same performance in so called "craft technology jobs" when provided with the proper information and well arranged business management structure. This hypothesis is examined by using an ordering task simulated from real-world operations, and setting experiment conditions controlled by novices and experts. The hypothesis can be verified based on the time consumed for decision making and resultant cost defined by inventory and shortage costs.

View full abstract

The relationship between skill learning, work duration, and workload is analyzed using a visual search task. The subject's task was to search for the same two-digit number on the panel as the one presented on the computer screen, and type in one of the four numbers attached to the target on the panel using a keyboard. The number to search and the location of attached number to answer were shown 5 seconds after the previous response. The task was repeated for 30 minutes without a rest, and the 30-minute trial block was repeated three times, with a short break for a Flicker Test and for filling up subjective scales. An analysis of reaction time for each subject showed how the learning process progressed and when it stopped. An increase in work strain was shown by CFF values and subejctive rating scales for fatigue. Subjective mental workload measured with NASA-TLX decreased once, and then increased along with the learning process, while subjective indices of physical/mental fatigue monotonously increased as a function of work duration.

View full abstract

In this paper, we propose a regional industry-social society model which considers the following factors : regional industrial structure, payment of taxes from the residents and companies and expenditures of the local government to ward local society. A linear regression anlaysis applying the above-mentioned factors is used to make clear the role of regional industry in regional society from the viewpoint of tax flow. Firstly, the features of the local industry are shown by the kind and added-value per employee of 22 manufacturing industrial categories. Secondly, as the revenue of the local government depends on the amount of payment of various taxes from the residents and companies, the amount of payment of residents' tax is shown by the log linear model of the number of employees in 3 or 10 industrial categories. Finally, as the amount of expenditures determined by the local government has a direct effect on social life, the amount of expenditures is shown by the log linear model of the payments of revenue classified into 4 categories. It is made clear that the proposed model can explain the tax flow in each prefecture clearly by the analysis of data of 9 prefectures in Chugoku and Shikoku areas.

View full abstract

Recently, Kanagawa et al.(1993) proposed a (x^^-, s) simultaneous control chart with the control limit based on the probability limit method. Further, the results derived by Arizono et al.(1995) enable the user to evaluate the performance of the (x^^-, s) simultaneous control chart. On the othe hand, some innovations have been proposed to improve chart performance in the case of monitoring the shift in the process mean. Reynolds et al.(1988) considered the properties of the Shewhart x^^- control chart with variable sampling intervals. A Shewhart x^^- control chart with variable sample size was also considered by Prabhu et al.(1993) and Costa (1994). Additionally, Prabhu et al.(1994) and Costa (1997) investigated the Shewhart x^^- control chart where both the sampling interval and sample size are varied depending on the position of the current observations on the Shewhart x^^- control chart. Furthermore, the CUSUM x^^- control chart and EWMA x^^- control chart with variable sampl sizes and/or sampling intervals were also considered for the purpose of improving chart performance. However, in a physical environment, it is usual to employ jointly the x^^- chart for monitoring the shift in process mean and the R chart for monitoring the change in process variance. Recently, Costa (1998) has considered joint x^^- and R control charts with variable sample size and variable sampling intervals with this view. The variable sample size and sampling interval x^^--R control charts presented by Costa (1998) have twelve design parameters, since the statistic x^^- for monitoring the shift in process mean and statistic R for monitoring the change in process variance are jointly employed. The feature of the (x^^-, s) simultaneous control chart is, however, that in this control chart we can monitor the process condition to the in-control state described by the process mean and variance by using a single statistic. In this paper, we consider the variable sample size and sampling interval (x^^-, s) simultaneous control chart for monitoring both the shift in process mean and the change in process variance for the purpose of improving chart performance. The performance of the variable sample size and sampling interval (x^^-, s) simultaneous control chart is compared with the performance of the standard (x^^-, s) simultaneous control chart.

View full abstract

In mixed-model assembly scheduling in JIT production system, keeping the consumption rate of each part used constant (parts usage leveling goal) is important. The goal (GC) chasing method and a number of papers have been proposed to achieve this goal. In such papers, however, the difference of assembly time among each product has not been considered. It has also never been described how the influence of the difference affects the parts usage leveling goal. Recently, a time-based goal chasing (TBGC) method that considers of the different assembly times among each product has been proposed. In this paper, we propose a new objective function. Finally, we show the efficiency of our proposed method and describe the influence of many conditions by numerical experiments.

View full abstract

Conveyors are used to transport or deliver material and also to store it, and the conveyor-serviced production systems (CSPSystems) generally consists of multiple generalized conveyor-serviced production stations (generalized CSPSs) (Matsui, 1993). This paper studies an ordered-entry array of unit stations, generalized CSPSs, called a simple flexible assembly system (simple FAS). In this station-centered approach, the main concern is a coordination problem or system balancing between generalized CSPs, and this design problem has not yet been discussed anywhere. From the view of performance evaluation, we first found that allowing for only an overflow from the last station is sufficient to minimize the total expected cost of simple FAS, and there is a limit point to decrease the overflow rate by increasing a coefficient of overflows. Also, the arrangement of heterogeneous servers gave a counter example that the best server's arrangement under the case of in-process inventory is established by arranging the ascending order of processing rates to decrease both the overflow rate and the total expected cost. Second, we propose a simple 2-stage design method for simple FAS, and discuss an optimal design procedure using the complex-type method or genetic algorithms. This basic idea (Matsui, 1994) originates in dividing some design variables into input/output variables like cycle time and buffer variables like look-ahead times and reserve capacities. In this method, the first stage is to set the parameters, the cycle time and the number of stations. The second stage is to design the buffers, the look-ahead times and the reserve capacities under the given traffic. Finally, a design example is given and discussed.

View full abstract