In the reliability engineering, it has been very important to estimate the lifetime distribution of the replacement parts of systems from the field data. But, the field data consist of times to failure on failed units intermixed with running times, called censoring times, on unfailed units. Such data are quite common and results from (1) removal of units from use before failure, (2) loss or failure of units due to extraneous causes, and (3) collection of data while units are still operating.
The field data of marine use mechanical units have also such characteristics as described above. Consequently, it is very difficult to obtain the true estimation of lifetime distribution from such censored data.
So far, the authors have qualitatively investigated MTBF, failure rate, repair distribution and so on of the repair units from the field data by using the shape parameter m of the Weibull distribution.
But, the graphical method by the Weibull probability paper is not strictly applicable to censored samples without modification.
In this paper, first, the shape parameter m of the Weibull model is theoretically discussed from the viewpoint of the mean and variance of this model. Next, the new estimation method of this parameter is proposed on the basis of the estimation principle of the maximum likelihood function introduced by A. C. Cohen.
Finally, the usefulness of the newly proposed method is illustrated by using the censored data.