基于多性能多故障映射有向图的数控装备健康评估和剩余寿命预测研究

Health management and maintenance are the main bottlenecks hindering the development of modern CNC equipment. The complexity of multi-failure modes makes it difficult to correctly manage and predict the condition of a CNC equipment. This proposal takes CNC equipment as the main research objective. The goal is to explore a new theory and method for health assessment and residual useful life prediction. The health assessment will be based on the performance characteristics and fault mechanism, and the residual useful life prediction will be based on relationship between multi-dimensional performance and multi-failure mode. The main contents of the project are: a performance selection algorithm by guided Wrapper under small sample size; a nonlinear degradation model involving transitional repetitive error; a mapping-digraph model representing the relationship between multi-dimensional performance and multi-fault mode; a new way of health status definition based on graph matrix and a flexible partition for health space by Technique for Order Preference by Similarity to Ideal Solution (TOPSIS);a probability prediction model based on failure-oriented performance threshold distribution and performance-based failure probability distribution; a new performance-driven failure assessment and prediction process including a two-layer fusion strategy for residual useful life prediction from "multi-performance → single failure" to "multi-failure → residual useful life"; a decision making model and an optimisation algorithm for Condition Based Maintenance (CBM）in the extension of machine service cycle based on previous health assessment and residual useful life prediction. This research aims to develop new fundamental theory and methods for CNC equipment's health assessment and residual useful life prediction. The research is of significant theoretical and realistic value to enhance the reliability and maintainability of CNC equipment.

健康管理和维修维护难题是阻碍当今数控装备发展的瓶颈，复杂的多故障模式使复杂装备的健康管理和寿命预测困难。本申请以数控装备为研究对象，从装备性能特性和故障形成机理的角度出发，探索一种基于多维性能和多故障模式的数控装备健康评估和剩余寿命预测新思路，包括：小样本条件下有指导打包(Wrapper)性能选取和基于重复误差变迁规律的非线性性能退化模型；数控装备多性能多故障映射有向图模型；基于图论矩阵运算的健康指数函数模型和健康状态空间的逼近理想解排序（TOPSIS）划分；面向故障模式的性能阈值分布拟合和基于性能的故障模式发生概率预测模型；一种从"多性能特征→故障模式"到"多故障模式→数控装备"的双层剩余寿命融合策略；基于健康评估和剩余寿命预测的视情维修决策模型和优化算法。本课题为数控装备健康评估和剩余寿命预测提供新的理论方法，对提高数控装备运行可靠性和维修性都有重要的理论意义和工程价值。

健康管理和维修维护难题是阻碍当今数控装备发展的瓶颈，复杂的多故障模式使复杂装备的健康管理和寿命预测困难。本课题以数控装备为研究对象，从装备性能特性和故障形成机理的角度出发，探索一种基于多维性能和多故障模式的数控装备健康评估和剩余寿命预测新思路，包括：提出面向数控装备的健康评估和寿命预测理论框架；建立基于FMECA的多性能多故障映射有向图模型，描述数控装备性能和故障之间的关联关系，并提出基于FMECA的数控装备故障模式的辨识方法，提出基于多性能多故障映射有向图模型的故障分析方法；基于ADAMS建立进给系统仿真模型，获得结构性参数和性能参数的数据，由此建立基于BP神经网络的故障-性能映射模型，并提出性能退化预测方法来预测进给系统定位误差；针对小样本、退化规律多变性和检测非周期性等精度退化特点，分析数控装备的精度退化规律，建立精度退化过程模型，提出小样本条件下数控装备性能非重复精度和重复精度的退化分析和寿命预测方法；利用逼近理想解法和马氏距离，根据相对贴近度划分出装备健康状态空间，并将性能退化轨迹转换到健康状态空间中，实现对装备健康状态的评估；依据失效原则，建立数控机床的性能阈值分布模型，然后建立基于阈值分布的剩余寿命预测模型得到剩余寿命概率密度函数，分析得到剩余寿命；提出多故障条件下的数控装备维修决策分析方法，提出一种基于风险二叉树的事后维修决策方法，利用扩展FEMCA辨识的故障风险度等构建风险二叉树，依据该风险二叉树的每一次检测能最大限度降低数控装备的风险程度，从而实现快速事后维修；数控装备健康评估和寿命预测平台设计与开发。本课题为数控装备健康评估和剩余寿命预测提供新的理论方法，对提高数控装备运行可靠性和维修性都有重要的理论意义和工程价值。