基于乏信息失效数据的可靠性评估方法

基于乏信息失效数据,提出了机械产品可靠性的最大熵评估模型.根据可靠性经验值公式,获得失效数据的可靠性经验值向量,并逆推出离散失效频率向量即获得统计直方图;基于区间映射的牛顿迭代方法获得具有最大熵的概率密度函数,对其积分获得失效概率分布函数,进而得到可靠性估计真值函数.仿真案例和试验案例研究证明该方法可以很好地评估已知分布的可靠性并有效地解决只有失效数据而没有概率分布任何先验信息的可靠性评估问题.在寿命给定时,最大熵方法获得的可靠性取值与已知分布获得的可靠性取值之间的差值非常小仅为3.40%. 

Reliability evaluation method based on failure data with poor information

Under the condition of failure data with poor information, the maximum entropy model for evaluation of mechanical product reliability was put forward. According to the empirical value formula of reliability, the reliability empirical value vector of failure data was obtained, and discrete failure frequency vector of lifetime data could be inferred by the empirical value vector (i.e. statistical histogram was obtained); based on the interval-mapping Newton iteration method, the maximum entropy probability density function was established, the failure probability distribution function could be gained by integration, and then the estimated true value function of reliability was acquired. Studies on simulation cases and experimental cases have proved the method proposed is not only able to evaluate the reliability with a known distribution but effective and feasible in the reliability assessment under the condition of failure data without any priori information about the possibility distribution. At the given lifetime, the difference between the values of reliability obtained via the maximum entropy method and the known distributions is very small, which only reaches 3.40%.