In this paper, an idea of constructing standard spaces to identify faults is discussed using the case of HOPE-X as an example. The items of a standard space for detecting the HOPE-X component faults were assigned to an orthogonal array for the selection of contributing ones, which were supposed to construct a new standard space to identify each fault. But the standard spaces thus reconstructed could not identify the faults. In order to reduce the interaction of items, the original standard space was orthogonalized and its orthogonal axes were assigned to an orthogonal array and Mahalanobis distances were calculated by the selected axes. This time, it became possible to identify the faults. It remains to be seen how to improve the accuracy of fault identification. However, this paper shows an idea as a possible coulltermeasure to reduce the interaction of items in using an orthogonal array to select items.

Potentiometer is an electronic component part. Basically, it is a variable resistor. The type of product studied in this report is used as a sensor. It is demanded that the product have a high precision. One of the important functions is its linearity. It is ideal that voltage increases proportionally to axis rotating angle. The causes affecting non-linearity include the dimensional variation of elemen七resistance film, misalignment during assembling and deterioration by wear. In this experiment, it was aimed to improve the uniformity of resistance film. After optimization, the process capability was improved.

A new alloy our company recently developed is lightweight and has an excellent heat resisting property, but it is diffcult to process causing rough finishing surface,tool wear or tool breakage. Sometimes it was impossible to process. A drilling process was studied for improvement using electric power as a means to evaluate drilling function. Since the power consumption during the pre-drilling,mid-stage drilling and final drilling is different, analyses were separately conducted for each stage. A very good reproducibility of conclusions was found. The results were applied to mass production resulting a good stability for the process and a 50% speed increase.

The stiffness of a toothbrush is an important characteristic for the selection of a brush. But sometimes, the stiffness that the users feel during usage is different from the one indicated on the package of the brush. In this study, the MTS method was used to quantify the user's feeling during usage in relation to the physical properties of the brush itself as well as to various other factors such as the user's age, the presence of caries or periodontal diseases. As a result, the stiffness that the users feel was successfully quantified using the MTS method with various factors cited in the study.

Artificial neural networks (ANN) have been in use for pattern recognition,learning, classification, generalization and interpretation of noisy inputs. There is a relatively new methodology called Mahalanobis-Taguchi system (MTS) that is also being used for such applications. MTS is based on the principles of quality engineering and Mahalanobis distance (MD). There are some commonalities and differences between these methods. This paper provides a detailed comparison between these methods with the help of two case studies. The comparison is made in terms of the ability to recognize various patterns and dimensionality reduction. The paper also provides a set of guidelines that will help in choosing one of these methods depending on the situation.

A simple material removing process to smooth the roughness of already processed workpieces was studied. It was aimed to utilize the functions of chemical corrosion and physical collision by ultrasonic wave. After optimization, the linearity between processing time and the amount removed was improved, and the surface unevenness was no more visible. However, the gain was not reproduced due to the significant difference in surface roughness between the spots measured.