基于灰色神经网络高速永磁电机试验效率评估

为了打破传统电机检测技术分析效率低、同步性差的局限，提出基于多参数评价的高速永磁电机动态性能评估模型。采用了热卡填充填补缺失值完成预处理，设计灰色关联度模型（GRA）得到各类属性列之间关联度，利用了贪心并查集思想得到降维后的4列电机属性参数，建立了一个4-5-1的三层神经网络结构。通过改变贪心算法得到的期望属性组数到5组并增加神经网络的参数设置，实现了电机测试数据分析模型的优化改进。在允许相对误差0.05的范围内，永磁同步电机(TB-416G-30-5型)运行效率预测准确度从90%提高到94%，试验表明：优化的灰色BP神经网络模型能有效适用于预测电机运行效率，在电机制造的智慧生产及机器学习在电机评估方面的应用有重要意义。

Test efficiency evaluation of high-speed permanent magnet motor based on gray neural network

To break the the limitations as slow analysis efficiency and poor synchronization, a dynamic performance evaluation model for high-speed permanent magnet motors based on multi-parameter evaluation is proposed. The hot card is used to fill in missing values to complete preprocessing. The grey relational analysis model (GRA) is designed to obtain the correlation between various attribute columns. The four-column attribute parameters of the motors after dimensionality reduction are obtained through the set of thoughts. A 4-5-1 three-layer neural network structure was established. By changing the number of expected attribute groups obtained by the greedy algorithm to 5 groups and increasing the parameter settings of the neural network, the optimized motor test data analysis model was designed. Within the range of allowable relative error that is 0.05, The accuracy of the prediction of operating efficiency has been increased from 90% to 94%. The experiments show that the optimized gray BP neural network model can be effectively used to predict the operating efficiency of motors, which is beneficial to the intelligent production of motor manufacturing and the application of machine learning in motor evaluation.