DSFT的建立及故障概率空间分布的确定

为利用工矿企业生产过程中已有的长期监测得到的元件故障数据,提出适合于处理这些数据的离散型空间故障树(DSFT).在DSFT基础上提出了确定元件故障概率空间分布的方法,即因素投影拟合法和基于神经网络(ANN)的方法.给出了这两种方法的具体实现过程,考虑了工作时间t和工作温度c对于元件故障的影响,研究了范围c:0~40℃,t:0~50天内的元件故障概率空间分布.并将这两种方法和CSFT得到的元件故障概率空间分布进行了对比分析.比较分析证明:使用DSFT的ANN预测法所得结果要比使用DSFT的因素投影拟合法更为接近CSFT的相对真实结果.在只有监测数据而不清楚系统结构的情况下这两种方法均适用,但ANN法更精确.

Discrete space fault tree construction and failure probability space distribution determination

In order to make full use of existing long-term monitoring of component fault recording data, in the production process of industry and mining, the discrete space fault tree (DSFT) is put forward for processing the data. The method for failure probability space distribution is proposed based on DSFT, namely the factors projection fitting method and ANN method. Implementations of the two methods are put forward. Considering the component failure impact because of the working time t and working temperature c, on a range of c: 0~40℃ and t: 0~50 day, the failure probability space distributions are studied. The results are compared with that of CSFT. Comparative analysis shows that: using ANN in DSFT, the prediction results are more close to the relatively real results with CSFT than using DSFT factors projection fitting method. In the case of only the monitoring data without clear system structure, these two methods can be used, but the ANN method is more accurate.