基于特征的机械设备故障预测系统

在机械设备的使用过程中,如果能准确地提前预测机械设备将要发生的故障并防患于未然,则可为企业节省大量的维修费用,降低企业成本。为此提出了一种依据特征基于神经网络算法的故障预测系统。把机械设备本身的运行特征和环境特征作为输入,把发生故障的概率作为输出,建立映射关系。以历史映射数据训练并建立神经网络,用以预测机械设备在某状态达成后发生故障的概率。通过某厂的机床对系统进行训练和验证,预测故障率与实际故障率的相对误差在15.3%以内。仿真结果表明,该系统可以有效的根据机械设备运行状态特征对其故障率进行准确预测。

Mechanical equipment failure forecasting system based on features

In the process of using mechanical equipment, if its failure can be forecasted before it happens, measures can be taken in advance and companies can save a lot of maintenance costs. To solve this problem, a Mechanical equipment failure forecasting system based on features and neural network is proposed. The machinery equipment operating characteristics are used as input and the probability of failures are used as output. Then mappings are established between the input and the output. Historical mapping data is used to train and build neural network which is used to predict the probability of failures. Features and probability of failures of a lot of Machine tools are used to train and validate this system. Simulation results show that the relative error is less than 15.3% and this system can effectively predict the sales of the probability of failures based on oroduct features.