基于凸优化理论的多传感器故障诊断技术

Multisensor fault diagnosis technology based on convex optimization theory

针对DS证据理论基本概率赋值函数难以获取、证据间要求相互独立等缺点,提出了利用凸优化理论来建立满足需要的多传感器故障诊断模型。本文首先分析了传感器故障诊断报告的形式,构造了多传感器故障融合的代价函数。通过分解代价函数,将多传感器故障诊断问题转换为凸优化问题。同时利用对数罚函数内点算法求解凸优化模型,该算法结构简单,计算量小,易于实现。理论分析和仿真结果表明,基于凸优化模型的多传感器故障诊断方法较之传统的DS证据方法具有更好的识别能力、更强的鲁棒性和更广的适用范围。

Aiming at the drawbacks of DS evidence theory that the basic probability assignments function are difficult to get and the evidence must be mutually exclusive, a multisensor fault diagnosis model using convex optimization theory satisfied requirements is presented. First, the form of multisensor fault diagnosis reports is analyzed in the paper. The problem of multisensor fault diagnosis is transformed to convex optimization problem by constructing and decomposing cost function. Meantime, the convex optimization model is solved by using logarithm penalty function interior-point algorithm which has simple structure, small computation and easy to realize. The theory analysis and the simulation result prove that the convex optimization model has better diagnosis ability and stronger robustness than traditional DS theory.