Resistance Variant Fault Diagnosis Of Mine Ventilation System Based On Genetic Algorithms

Resistance variant fault(RVF)phenomenons such as damper switched or damaged,tunnel fell or deformed that caused permanent changes in wind resistance often occurs in mine ventilation systems,real-time diagnosis of RVF can detect the location and type of faults in the system in a timely manner,and formulate targeted troubleshooting measures can reduce the potential for safety hazards to develop into accidents.The current fault diagnosis method for mine ventilation system needs to collect fault samples.In the process of performing diagnosis,the diagnosis of fault location and fault volume needs to establish independent classification and regression diagnosis models respectively,which makes it difficult to realize real-time diagnosis of RVF.Aiming to solve the problem that the fault collection of mine ventilation system is difficult to collect and the fault location and fault quantity cannot be diagnosed at the same time,the hybrid coding rule based on genetic algorithm was used to transform the fault diagnosis problem of mine ventilation system into the minimum Euclidean distance about air volume before and after the fault.An unsupervised learning fault diagnosis method based on hybrid coding genetic algorithm without sample training is proposed.The unsupervised resistive fault real-time diagnosis of mine ventilation system is realized by integrating fault location classification diagnosis and fault volume regression diagnosis.The pulsation and randomness of mine ventilation flow cause large error in wind speed monitoring value,which is one of the important factors that directly affect the performance of RVF diagnosis.Through the research of LDV wind speed test and simulated mine test,it was found that the Adaptive Kalman filter can effectively eliminate the wild value existing in wind speed monitoring and reduce the monitoring error of instantaneous wind speed,and reduced the monitoring error of instantaneous wind speed to realize real-time online monitoring of mine ventilation wind speed,whicch provide basic conditions for realizing the unsupervised fault diagnosis of mine ventilation system.The T-type mine ventilation network was used as the experimental model for RVF diagnosis of mine ventilation system.The test is carried out by using the single air volume,wind pressure characteristics and air volume-wind pressure composite characteristics as the variables of the fitness function.The results show that compared with the single wind pressure feature and the wind volume-wind pressure composite feature,when a single air volume was used as the fitness function variable of the genetic algorithm,the fault location diagnosis with higher accuracy and the fault diagnosis with lower error can be ensured.Fault simulation and fault location and volume diagnosis analysis of T-type mine ventilation network test model by using air volume as an adaptive function variable,the results showed that the hybrid coding genetic algorithm was feasible and effective for the diagnosis of resistive fault location in mine ventilation system.The mean square error of branch fault diagnosis was 33.446 Ns2/m8,and the relative error was 0.019.The R-squared value was 0.977.At the same time as the effective diagnosis of the mine ventilation RVF’s location was satisfied,the fault quantity diagnosis with lower error was obtained.Therefore,it is feasible and effective to use the hybrid coded genetic algorithm to perform unsupervised real-time diagnosis of resistive faults in mine ventilation systems.The Banshi coal mine was used as an example of field test.The 120-time resistive fault diagnosis field test was conducted on the Banshi coal mine using the proposed fault diagnosis method.The results showed that the accuracy of the fault location diagnosis was 1,the accuracy and recall rate were both 0.86,the R2 value of the fault diagnosis was 0.995,and the mean square error(MSE)was 15.89 N·s2·m-8.The diagnostic time was approximately 9.7s.The cumulative diagnosis of cumulative time consumption was linearly positively correlated with the number of diagnoses.The feasibility,robustness and effectiveness of the resistive fault diagnosis model were proved.The genetic algorithm resistive fault real-time diagnosis method can effectively solve the resistance fault of the mine ventilation system which is difficult to check on site.The dissertation has 46 figures,8 tables and 122 references.