Abstract:
Traditional machine learning algorithms usually adopt iterating method to achieve the parameter optimization, which may lead to slow learning speed and easy falling into local minimum. The paper proposes a soft measurement modeling method based on extreme learning machine with wavelet kernel function (KEML). By applying the idea of the kernel function in support vector machine (SVM) to the extreme learning machine (EML), the proposed algorithm can overcome the slow training speed of SVM and the unstability of ELM algorithm. The experimental results by applying this method to the acetic acid distillation of soft measurement model show that the learning speed of KEML is 92 times as SVM, and the accuracy and generalization ability of the model is also improved.