Abnormal Temperature Detection Of Roller Kiln Based On IWOA-SVDD

Kiln temperature is the most important index in the process of roller kiln ceramic production,which plays a decisive role in the quality of ceramics.Once the temperature of roller kiln is abnormal,it will have a serious impact on ceramic products.Therefore,whether the temperature of the kiln can be effectively monitored is the key to control the quality of ceramics.The existing detection of roller kiln temperature anomaly is judged by whether it exceeds the fixed threshold value.This method has low efficiency,poor detection accuracy,and is prone to false alarm and other defects,which is not in line with the trend of the increasingly intelligent roller kiln.Therefore,this paper proposes an improved whale algorithm to optimize support vector data description algorithm(iwoa SVDD).Because the temperature data of kiln is not strictly obeyed by Gaussian distribution,the traditional process monitoring algorithm is not suitable,so SVDD which requires low data distribution is chosen as the anomaly detection algorithm;Because the compromise factor C and kernel width of SVDD σ have significant influence on the detection results,the whale algorithm is used to find the optimal parameters,and the improved method is proposed for the shortcomings of low efficiency and weak global search ability of the standard whale algorithm;The experimental results show that iwoa SVDD is more accurate and the false alarm rate is lower,and the abnormal furnace temperature can be accurately located,which proves the superiority of the proposed algorithm.The detailed work is as follows:(1)The paper analyzes the research object roller kiln,analyzes the composition of roller kiln production system,the process of ceramic production,and the reasons why kiln temperature plays an important role in ceramic quality.The accuracy of SVDD abnormal recognition is greatly influenced by the quality of model training data.Therefore,the original data is processed by outliers,vacant values and standardized,which ensures the quality of SVDD training data.(2)The influence of key parameters c and σ are significant for the detection results.Therefore,the whale algorithm is selected to obtain the best parameters.In order to improve the optimization effect,adaptive reverse learning is proposed to improve the convergence accuracy of whale algorithm and the development ability of Gauss mutation operator in the later stage.Finally,the test functions of 8 common optimization algorithms are used to verify the improvement effect by comparing with particle swarm optimization,genetic algorithm and standard whale algorithm.(3)Based on the previous three chapters,the SVDD anomaly detection model is constructed.In order to measure the optimization effect of the improved whale algorithm on SVDD,UCI data set and historical kiln temperature data are compared with particle swarm optimization,genetic algorithm and standard whale algorithm.In order to compare the effect of SVDD anomaly detection model with the kernel principal component analysis and the kernel partial least square method,the comparison experiment is carried out.Because SVDD itself can not locate the abnormal,the contribution graph method of PCA is used to locate the abnormal temperature.(4)Combined with the above contents,the function requirements of temperature alarm module are analyzed.The system architecture of pretreatment module,abnormal detection module and alarm module is designed.The development system of ceramic kiln monitoring system is used.The energy module of temperature anomaly detection and location is developed by SSM architecture.Finally,the interface diagram of the developed functional module is shown.