基于概率神经网络的机床健康评估方法研究

Study of Machine Tool Health Assessment Method Based on PNN

为有效地评估机床的健康状态,制定合理的维护与维修策略,提出一种基于概率神经网络的健康评估方法。采集不同加工状态下的主轴振动信号,进行特征提取与归一化处理,获取特征向量;基于PNN识别当前的加工状态,并将特征向量与该加工状态下的训练样本进行基于高斯核函数的相似度计算,评估机床的健康状态。验证实验表明:该方法能够有效地识别机床加工状态与健康状态。同时,引入Kafka与Storm技术,验证该方法对机床实时健康评估的可行性。

In order to evaluate effectively the health status of machine tools and formulate reasonable maintenance and repair strategies,a health evaluation method based on probabilistic neural network(PNN)is proposed.Spindle vibration signals are collected in different processing states to perform feature extraction and normalization processing,so as to obtain feature vectors.Based on PNN,the current processing state is identified,the similarity between the feature vector and the training sample in the processing state is calculated based on Gaussian kernel function,and the health status of the machine tool is evaluated.Validation experiments show that proposed method can effectively identify the machining status and health status of the machine tool,and whose feasibility in real-time health assessment of machine tools,meanwhile,is verified with the introduction of Kafka and Storm technologies.