[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Abnormal I_DDQ (Quiescent Power Supply Current) is a signal to indicate the existence of physical damage in a circuit. This signal is detected as abnormal leakage current at supply voltage terminal,pass through the faulty portion. The paper presents the trend of testing, analysis and diagnosis technology of CMOS-LSI using this phenomenon. Testing way is the evaluation way to select effectively the abnormal LSI with excessive current by LSI tester. Analysis technique makes it possible to detect rapidly physical abnormality by physical phenomenon with unusual leakage current. Diagnosis method is the CAD-based way to narrow progressively to the faulty area by logic information and test vector number with abnormal I_DDQ. This diagnosis method is applied additionally to detect the fatal defect fragments in logic manufacturing line.

View full abstract

Automatic diagnosis is one of the important functions of programmable electronic (PE) systems that give us the most availability of PE safety-related systems (SRSs). Thus, it is often regarded as the automatic diagnosis reduces the hazardous event frequencies (HEFs), which are caused by detected failures of PE SRS, to a negligible level as compared with those by undetected failures. However, this is not always realistic. This study shows there are cases where HEFs caused by detected failures exceed those by undetected failures especially in the range of higher demand rates and/or more expanded diagnostic coverages.

View full abstract

The present paper introduces a new approach of advanced-FTA (a-FTA), which is applicable to sequential failure logic (SFL), to study how to combine risk analysis and allocation of safety-integrity levels (SILs) to safety-related systems taking a batch-process plant as an example. Fault tree diagrams for the hazardous event resulting from the loss of the safety functions are developed taking common cause failures into account. The relationships among failures of the subsystems, demands to activate the safety functions and the hazardous event are studied. It is found that the SFL governs the system failures. The top event frequency is estimated using conventional FTA (c-FTA) and a-FTA. The former might overestimate the top event frequency nearly ten times higher than those obtained with aids of a-FTA. Furthermore, it is demonstrated that the SILs obtained through a-FTA conform to those resulting from the architectural constraints on hardware safety integrity required by IEC 61508.

View full abstract

It examined about failure phenomena of the chip components, used for electric components occurred by the sulfuration gas atmosphere. The chip components of test targets have the termination(inner) of silver. As the result, on the chip resistor arrays and the chip resistors, it found that sulfuration silver creap and sulfuration silver whisker expanded between the exterior and the termination(outer). Their phenomena cause short mode by ion migration and open mode by silver exhaustion.

View full abstract

This paper is described on the effect of leadless and high soldering temperature for PWB. The result indicated that a amount of cupper leaching is varied by solder-material, and a failure appeared in lower-level manufacturing process

View full abstract

In recently, a lot of BGA (ball grid array) packages, which are one of the semiconductor packages, are used to the reduction of the mounting area of the printed wiring board. The ball share strength measurement is generally used as a method of evaluating the joint quality between ball terminal of this BGA package and BGA package substrate. However, there was a problem of not detecting an abnormal connection of the interface easily because the ball share strength measurement was a method of measuring share strength in parallel for the substrate of the BGA package. We investigated the new ball share strength measurement method, which can detect the abnormal connection of the interface.

View full abstract

Authors proposed the method of evaluating the characteristic of a cell which composed a photovoltaic module in un-destroying from a module terminal. In this time, the proposed method was simplified so that it could apply to degradation evaluation of the cell. Isc of cell which composed a photovoltaic module is specifically presumed and how to investigate the distribution of Isc is shown.

View full abstract

Semiconductor packages are, in general, subjected to both thermo-mechanical stress and moisture effect in an actual use environment, and therefore, it is a matter of concern that the delamination and micro crack induced by periodic temperature change would degrade moisture resistance. The effect of temperature cycles on the moisture resistance was evaluated, using two types of packages (EBGA and PBGA), by the Sequential Stress test, comprised of temperature cycling followed by moisture-proof test (HAST or PCT). In EBGA package, it was confirmed that the temperature cycling stress would increase moisture absorptivity and would change the package micro-structure, although little electrical failure was observed. Considering the packaging technology future trend of further miniaturization, such as narrower pitch design of wirings on BGA laminate, etc., it might be crucial to perform the Sequential Stress test to assess moisture resistance, because of its capability of earlier detection of potential failure.

View full abstract

The ESD testing method of the semiconductor device is classified broadly into Human Body Model (Machine Model as alternative test) and Charged Device Model. Regarding Charged Device Model, the discharge method using a mercury relay is standardized in Japan as JEITA standard. On the other hand, the aerial discharge method using a discharge plate is standardized in U.S. as JEDEC standard. Generally, charging voltage represents the ESD mununity for the Charged Device Model. In this paper, we clarified that ESD failure was strongly depended on the peak current (Ip) value of discharge waveform, not discharge method or charging voltage. And we indicated the future direction of standardization by applying this result.

View full abstract

The importance of product safety problems caused by customers' misuses and carelessness cannot be overstressed. Several methods of preventive measures for these problems have been proposed. This research improves a method for advance disclosure of potential human errors using FAQ on Web pages.

View full abstract

[in Japanese]

View full abstract

This research carried out a QFD educational procedure which considered knowledge management. Although there are already many references about a QFD procedure manual, there are no procedure documents related to the SECI model of knowledge management. A new QFD educational system was established and a manual was created. Next, a quality deployment of a QFD practice course was implemented as a case study. Its purpose was to exactly grasp the demands of the participants, to reflected it in the next practice course, and to connect it to new improvements. A questionnaire was performed with eight participants of the 2001 year practice course, and based on its information, the correspondence relation between the demanded quality and the quality elements was taken and the quality table was created. The questionnaire concerned to the demanded quality deployment table items was carried out, and the participant's degree of interest was investigated in order to get the demanded quality weight.

View full abstract

It is not effective to use a usually applied p-control chart when the variation between and within groups of multiple failure rates is large. For such a case, we present a new method of controlling failure rates properly by using logit transformation and inverse sine transformation so as to transform an attribute value to a measurable value and by applying an x - Rs control chart and an x^^--S control chart to a measurable value.

View full abstract

In some of our products, certain measurement data may not show normal distribution. We wanted to propose a new index, similar to Cpk, to reflect the process capacity correctly, even for the data distribution not having normal distribution, e.g. asymmetrical distribution. If only data smaller than MEDIAN is taken into account then the distribution sets close to normal distribution. By using this method, Cpk showed the analogous value to actual process capability, thus this method was proved to be effective.

View full abstract

When we try problem analysis with statistical technique, there may be the missing data in a sample by the various reason. However, in real data analysis in quality control, there hasn't been studied about an encounter with missing data. We report the results of research that considered missing data in quality control from the viewpoint of sample or selection of subpopulation in this report.

View full abstract

When we control a manufacturing process, a suitable statistical method has been adopted. Recently, the development of computer hardware and statistical software becomes to use the complex statistical models such as the mixed linear models and structural equation models. However, it can not easily analyze the multi-level structured data because the computational or model specification difficulty. Goldstein (1986) and Goldstein and McDonald (1988) developed mixture linear models and described the excellent algorithm to fit the model to real data. In this paper, we describe to use the extended model in multivariate multi-level structured data and propose an algorithm to use the model. Our applications to real data show the possibility of the use in the fields of the statistical process control or the statistical quality control.

View full abstract

Adopting the definition of probability as the degree of belief, an equation for the improved reliability lead by only information of a success was derived using Laplace's Rule of Succession. The improvement of reliability by the minimum information greatly depends on the prior reliability we have. If it is almost zero reliability where we cannot expect its success at all, then the information of only a success leads to us the improved reliability of more than 0.5 as the posterior reliability.

View full abstract

The aerospace systems, which are manufactured by our Works, are the complex and small-to-medium mass-produced systems, and require high dependability and safety under special environment. This report is explained Reliability Management of our Works for Aerospace Systems.

View full abstract

In the organization, there are various incidents and accidents induced by human errors. The specific human errors belonging to the organization are those to be produced in due to defects of communication. By the communication between persons, the individual human errors will be amplified and spread widely, and in some cases rescued and recovered to the normal state. In this paper, we focused to the medical accidents in the hospital caused by the communication errors between personnel. The three elements FMEA considering a person, another person and their environment to work was used for the analysis. Prior to produce the FMEA, we introduced the SCM (service category map) followed by the action map to cover all the necessary activities. Furthermore, we introduced the communication flow chart to analysis the actions sequentially. As the results, this paper proposes the necessity of additional mental counseling work during the operation process in the hospital.

View full abstract

In our former study, we tried to predict the final consequence of damages on the large-scale accident such as an earthquake disaster and an organizational technological accident from the news clips of newspapers during, early several days. We focused to the Hanshin-Awaji big earthquake disaster and the critical accident of the uranium processing facilities in JCO of Tokai-mura. In this paper, we studied to compare three prediction methods using keywords, numbers of words and news area how to explain the objective parameters of damages. This paper proposes that the news area method and the keyword method are suitable to explain the damages for the earthquake disaster. Furthermore, the keyword analysis for the local fire accidents given big damages was performed.

View full abstract

[in Japanese]

View full abstract

In this paper, we propose Bayesian hierachical models based on Poison distribution. Estimations of the Bayesian hierachical models are the maximum likelihood procedure based on marginal likelihood functions and the posterior mean using Gibbs sampler. We apply the information criterions ABIC and EIC to the failure models using the estimation methods for choice a fitted model. It is shown that the Bayesian analysis based on ABIC and EIC is useful for the reliability identification.

View full abstract

In this paper, we present method for applying time series analysis for autocorrelated process control data. Concretely, AR, MA, and ARMA models are used to fit the data. Parameters of the time series models are estimated based on the maximum likelihood using the Kalman filter. We use the information criterion AIC for the selection of the appropriate model for a set of the data. It is also shown that the comparison between the time series analysis and EWMA is possible based on prediction errors of two methods.

View full abstract

A new power normal family has been introduced by Isogai (1999) to transform a skewed and positive-valued distribution to a normal one. Its properties are shown compared to those of the Box and Cox's (1964) power normal family. Our new power normal family has many applications, because it includes not only log-normal and Weibull distributions but also the extreme value distribution of Type 2. However, the above two power normal families have a serious defect that they can treat positive-valued observations only. Thus, an alternative power normal family is constructed through our power normal family to deal with data comprised of positive- and negative-valued observations, and its properties are examined.

View full abstract

Cox's proportional hazards model (PHM) has been widely applied in the analysis of lifetime data. It involves covariates influencing the times to the failure of a system. In this article, the covariates are assumed to be discrete random variables, and a mixed proportional hazards model is derived. We consider three types of data; one type of experimental data and two types of field data. We investigate estimation in the mixed PHM with these three types of data. We develop maximum likelihood techniques for joint estimation of distribution parameters, mixing proportions, and a regression parameter using the Expectation-Maximization(EM) algorithm.

View full abstract

At present, software development environment has been changing into new development paradigm in such client/server systems and distributed development by using network computing technologies. Software systems produced in such distributed one tend to increase in size and complexity. If the size of the software system is large, the number of faults detected during the testing phase becomes large, and the change of the number of faults which are detected and removed through each debugging becomes sufficiently small compared with the initial fault content at the beginning of the testing. In this paper, we propose a software reliability growth model describing a fault-detection process during the system testing phase of the distributed development environment by applying a mathematical technique of stochastic differential equations of an Ito type. Considering the conventional stochastic differential equation model for distributed development environment, we derive a flexible stochastic differential equation model, and show numerical examples with sensitivity analyses on the weight parameter.

View full abstract

We propose an infinite server queueing model considering the time distribution of fault isolation process based on a concept of the delayed S-shaped SRGM (Software reliability Growth Model) which is one of well-known NHPP (nonhomogeneous Poisson process) models. A number of SRGMs have been proposed as the method of quantitative assessment of software reliability from 1970's. The delayed S-shaped SRGM has been developed for describing a test process consisting of failure detection and fault isolation processes. Based on the delayed S-shaped SRGM, we derive the SRGM considering a probability distribution for fault isolation time per a fault by developing an infinite server queueing model. We show that we can easily analyze a physical interpretation for fault detection phenomenon by using our model. And we also mention that our model is general approach for several SRGMs decribed as NHPPs.

View full abstract

In recent years, products such as household appliances will generally be warranted against their failures by the manufacturer for a specific period. It is important for the manufacturer of such products to estimate the cost for such warranty services and incorporate it in the budget. Under the general warranty, the manufacturer repairs or rectifies a filed system before its warranty period expires. However, in case the system fails several times during the warranty period, some manufacturers do not repairs systems for their second failures and their subsequent failures but replace them by a new one for fear of losing their credit. This study considers the cost analysis for such a warranty service where a manufacturer conducts a minimal repair for the first failure of a system and replaces it with a new one for its subsequent failures during the warranty period. We also compare this warranty service with a service where a minimal repair is carried out for the failed system only once and with a service under which minimal repairs are performed to the failed system before the warranty period expires.

View full abstract

[in Japanese]

View full abstract