梯度下降优化模糊系统的接触电阻预测方法

根据接触电阻的特点,将结合模糊逻辑的预测方法引入电气领域,提出基于模糊系统的接触电阻预测新方法.根据接触电阻与各影响因素之间的关系及研究目的进行试验,得到足量试验数据,将所有试验数据分成两部分,训练数据和测试数据.通过训练数据运用梯度下降算法训练模糊系统,调整模糊系统参数,建立相应的接触电阻模型,利用训练数据建立接触电阻的回归分析模型.通过测试数据对两种模型进行测试,基于模糊系统的接触电阻模型的测试结果优于回归分析.预测与比较结果表明:若能得到足量训练数据,用梯度下降算法训练模糊系统建立的接触电阻模型精确可靠.

Method for Predicting Contact Resistance of Optimizing Fuzzy System by Gradient Descent Algorithm

Based on the characteristics of the contact resistance, the predicting method combined fuzzy logic is introduced to electric field, the new method for predicting the contact resistance based on fuzzy system is developed. According to relationship between contact resistance and influence factors, as well as the research target, the test is processed. The enough data is obtained and all data is divided into two parts, namely, training data and testing data. Fuzzy system is trained by gradient descent algorithm through training data, the systemic parameter is adjusted, the corresponding model of contact resistance is found. The regression analysis model of contact resistance is built by training data. The two models are tested through testing data, the prediction effect of the model based on fuzzy system is better than that of regression analysis. The prediction results show that if enough training data is obtained, the fuzzy system trained by gradient descent algorithm is reliable to predict the contact resistance.