论文部分内容阅读
提出一种用最小二乘支持向量机(least squares support vector machine,LS-SVM)构造函数链接型神经网络(functional link artificial neural networks,FLANN)的滚动轴承故障诊断系统。介绍了相关原理和具体算法,并给出了滚动轴承故障诊断系统模型。首先,采用LS-SVM模型核函数代替常规FLANN模型的扩展函数,避免了扩展函数选择的任意性;其次,利用LS-SVM学习模型得到FLANN权重系数,避免了BP方法多次迭代寻优存在的耗时长、局部极小及迭代设置初值依赖经验等不足;最后,构造了多层LS-SVM-FLANN结构,对多类滚动轴承故障进行诊断。具体实验表明,用LS-SVM构造FLANN的滚动轴承故障识别系统精度高、鲁棒性好、实现简单。
A rolling bearing fault diagnosis system based on least link support vector machine (LS-SVM) and functional link artificial neural networks (FLANN) is proposed. Introduced the relevant principle and the concrete algorithm, and gave the rolling bearing fault diagnosis system model. First, the LS-SVM model kernel function is used instead of the extended function of the conventional FLANN model to avoid arbitrariness of the extended function selection. Secondly, the weight coefficient of FLANN is obtained by using the LS-SVM learning model, which avoids the iterative optimization of the BP method Time-consuming, local minima and iterative set initial value dependent experience. Finally, a multi-layer LS-SVM-FLANN structure is constructed to diagnose the fault of many kinds of rolling bearing. The concrete experiment shows that the fault diagnosis system of rolling bearing with FLANN constructed by LS-SVM has high precision, good robustness and simple realization.