Fig. 10 Average Processing
Fig. 11 Scatter diagram of observed value and estimated value
Fig. 11 is a scatter diagram between the observed value and the estimated value using the data of 30 addition-averages. The estimated value at each time of observation is plotted with the mark □ with the estimated value on the Y-axis (the estimated running time) and the observed value (the observed running time) on the X-axis. In the figure, the linear relationship between the observed value and the estimated value is strong, and the accuracy of estimation is excellent, and it is found that the obtained multiple regression equation agrees well. The previously obtained multiple correlation coefficient R of the model is 0.928, also indicating the strong correlation from the numerical viewpoint. In reality, the estimated value is distributed around the regression line with the mean error of approximately 400 hours at any time of observation. Taking the above-mentioned points into consideration, the multiple regression model obtained in this chapter estimates the running time of the engine in a transition mode under a normal running condition from the vibration data of the engine. That means, the multiple regression model explains the regularity of the vibration data to be changed together with the running time of the engine under the normal condition of the engine, and the running time corresponding to the inputted vibration data can be estimated thereby. However, if the engine is not in a normal condition, an appropriate running time can not be estimated. Because the normal model in a transition mode is prepared, it can be understood whether or not the running condition is normal when the vibration data obtained at the engine side is obtained.
As the test ship is engaged in an actual service, no abnormality can be artificially generated. In this experiment, no abnormalities were generated in the engine, and no extraction of abnormalities of the engine could be detected. Thus, in this experiment, an examination was made what result is obtained in the estimated value in a different running condition from the normal one by inputting the vibration data of the engine at the engine speed of 30rpm which is different from the normal engine speed of 103rpm. The vibration data at different engine speed is the one where the running condition of the engine is changed, and the engine is not in any abnormal condition, but from the viewpoint that the vibration data is different from the normal one, it was regarded as the abnormal condition. Fig. 12 is a scatter diagram between the observed value and the estimated value where the estimated value of the above-mentioned data (hereinafter, referred to as abnormal data) is plotted with ■ (one sample at the lower left in the figure), and the estimated value of the normal data used in Fig. 11 are plotted with □ marks. As clearly shown in the figure, the estimated value of the abnormal data is estimated at the different position from that of the group of the estimated value of the normal data. The estimated value of the running time of the engine of the abnormal data was 1085 hours, and the error hour from the observed value was 1831 hours. Based on the negative estimated value, and the large error time, it can be diagnosed that the abnormal data shows the running condition different from the normal running condition. Based on the fact that even a slightest change in the running condition can be detected by the above-mentioned method, it is thus considered that a serious abnormality can be sufficiently detected by the present diagnostic method when it is generated in an actual engine. This point will further be clarified through the analysis of the changing data as the proceeds.
