Application Of Extreme Learning Machine Based On Improved Crow Search Algorithm In Gas Explosion Risk Prediction

Coal mine disasters have occurred frequently in China for a long time.The main threat of coal mine safety in China is gas explosion.Scientific and accurate prediction of gas explosion risk level is the basis for effective control of gas explosion accidents.Therefore,it is necessary to study and predict gas explosion prediction methods to improve the prediction accuracy and efficiency of gas explosion risk.In order to improve the accuracy and speed of gas explosion risk prediction,this paper constructs an Improved Crow Search Algorithm-Extreme Learning Machine(ICSA-ELM)prediction model.In view of the slow convergence of the crow algorithm and the blindness of optimization,it is improved from three aspects: the Tent chaotic population initialization is introduced;in the global optimization,the alienation pollination strategy of the flower pollination algorithm is used to accelerate the convergence speed of the CSA algorithm;in the flight length,according to the increase of the number of iterations,the flight length is adaptively reduced.On this basis,An ICSA-ELM is constructed by using the Improved Crow Search Algorithm to optimize the initial input weights and thresholds of the extreme learning machine.Applying ICSA-ELM model to gas explosion risk prediction,the characteristics of gas explosion risk have high dimensional and nonlinear characteristics.Firstly,the Kernel Line discriminant analysis(KLDA)is used to extract the gas data with risk level.The KLDA algorithm,the kernel principal component analysis(KPCA)algorithm and the feature extraction model are not used for comparison experiments.It shows that the KLDA algorithm has obvious effects on feature extraction of gas explosion risk data.Secondly,the extracted feature data is compared with the CSA-ELM,PSO-ELM and ELM models,and the traditional classic SVM(Support Vector Machine)and BP(Back Propagation)prediction models.The test results show that ICSA-The accuracy,consistency and operational efficiency of the ELM model are significantly improved.To some extent,the proposed method has good predictive performance in gas explosion risk prediction.There are 19 pictures,12 tables,and 51 references in this paper.