论文部分内容阅读
针对传统神经网络进行抽油机示功图识别诊断时受同步瞬时输入限制,不能有效体现连续输入信号的时间累积效应,诊断精度偏低的问题,提出一种极限学习离散过程元网络,模型内部通过三次样条数值积分处理离散样本和权值的时域的聚合运算.模型训练算法采用极限学习,将模型训练转化为最小二乘问题,通过利用Moore-Penrose广义逆和隐层输出权值矩阵来计算输出权值,提升模型学习速度.进行示功图识别时,直接将位移和载荷离散时间序列作为模型输入,对常见的5种示功图状态进行识别.实验结果表明,该方法具有较高的识别精度,同时相对于其它过程神经网络模型,学习速度较快.
In view of the problem that traditional neural network can not reflect the time cumulative effect of continuous input signal and low diagnostic precision when it is diagnosed by synchronous instantaneous input, it presents a limit learning discrete process meta network, The cubic spline numerical integration is used to deal with the time-domain aggregation of discrete samples and weights.The model training algorithm uses extreme learning to transform the model training into the least-squares problem. By using Moore-Penrose generalized inverse and hidden layer output weight matrix To calculate the output weight and improve the learning speed of the model.When the dynamometer is identified, the displacement and the load discrete time series are directly input to the model to identify the common five kinds of dynamometer states.The experimental results show that the method has the advantages of more High recognition accuracy, at the same time relative to other process neural network model, learning speed.