Construction of fatigue failure database and development of AI system for failure prevention

本课题编制并建立了机械结构故障数据库。事实数据库有258个实例。符合断裂类型为疲劳、静力破坏、腐蚀、腐蚀疲劳和热疲劳。每个实例包括电阻器编号、名称、产品名称、部件名称、材料、环境、主要损伤、次要损伤、原因、失效外观、对策、参考以及结构或断口图形。该搜索引擎采用Internet Java脚本编写，用户可以通过网络访问数据库，并开发了统计分析工具，对数据库中海量的疲劳强度和静态失效数据进行分析。该软件运行在个人计算机的Microsoft Windows上，使用商业数据库应用程序。“Microsoft Access”和电子表格应用程序“Microsoft Excel”。统计分析类型为正态分布、对数正态分布、威布尔分布、双指数分布，用户可选择。等级类型在平均等级和中位数等级类型中选择。软件名称为“STANAD_Win”，采用Microsoft Visual Basic语言编写，界面友好，用户可以更改数据检索和统计分析功能。在本软件中，参数估计分布等统计分析使用商业数据库进行，并显示统计分布或P-S-N图。参数估计有两种方法，即最小二乘法和相关系数法。

In this project, failure database of machines and structures was compiled and constructed. A factual database has 258 instances. Complied fracture type was fatigue, static failure, corrosion, corrosion fatigue and Thermal fatigue. Each instance consists of resister number, title, product name, parts name, material, environment, main damage, minor damage, cause, failure appearance, countermove, reference and figure of structure or fracture surface. The search engine was coded by the Internet Java script, user can access the database by using the network.Statistical analysis tool to analyze the voluminous data in the database on fatigue strength and static failure was developed. This software is running on Microsoft Windows of personal computer and use the commercial database application, . "Microsoft Access" and the spread sheet application "Microsoft Excel". The type of statistical analysis is normal distribution, logarithmic normal distribution, Wibull distribution, and double exponential distribution and user can select. The rank type was selected among Mean rank and Median rank type. The software name is "STANAD_Win" and coded by using the Microsoft Visual Basic language so that the software is user friendly and user can change the function of data search and statistical analysis. In this software, parameter estimation of distribution and other statistical analyzes carried out using a commercial database and display the statistical distribution or P-S-N diagram. Two kinds of methods, the least square method and the correlation coefficient method, are available for the estimation of parameter.