Infrared Temperature Mutation Precursors And Warning Model Based On SVM Of Saturated Siltstone Failure

The development of rock fracture with mutability may lead to the corresponding field of infrared temperature mutations,but there is little research about the transient variation characteristic of Infrared Temperature Field(ITF)during rock failure.In order to explore infrared temperature mutation precursors of rock instability,the paper took the infrared and acoustic emission monitoring experiment of siltstone samples saturated by 50° C water in the process of uniaxial compression as an example,featured the small time-stepping difference processing method of thermal image sequence,and deeply discussed the transient variation characteristic of ITF before rock failure through the extraction of ITF mutation information and statistical analysis of ITVF.Finally,an early warning model of rock instability based on SVM theory was built,with merging multi-source monitoring data.The main research content and results were as follows:(1)Through extracting infrared temperature mutations information directly based on threshold theory,the infrared temperature mutation abnormality of rock destabilization was researched.The result shows that the small time-stepping difference processing method could inhibit the interference of environmental factors and highlight the infrared temperature mutation information caused by rock micro fracture development;that approaching failure siltstone infrared temperature mutation abnormal precursory exists,including abrupt rise abnormality,abrupt fall abnormality,mutation extreme abnormality and mutation point amount abnormality;that the relationship of abrupt fall and abrupt rise abnormal omens of infrared temperature is generally negative correlation,and there is a positive correlation between mutation extreme abnormal omen and mutation point amount abnormal omen;that the four forms of infrared temperature mutation abnormal omen can complement and verify each other,which has a certain theoretical significance to improve the reliability and accuracy of rock catastrophic prediction.(2)As infrared temperature mutations may cause statistical distribution anomalies of ITVF,the paper put forward the concept of Infrared Temperature Variation Field(ITVF)and explored infrared temperature mutation precursors of rock instability from the perspective of statistical distribution.The research comes to some conclusions.Firstly,the abnormal jump of ITVF characteristic parameters,including Range(R),Kurtosis(K),Mutation Ratio(MR),etc,exists in the process of saturated siltstone failure.Secondly,the anomalies mainly focus on theplastic and post-peaking phase,especially tending to appear at the moment of obvious pressure drop occurring.Thirdly,the exception can be interpreted as,when the deformation enters into plastic and post-peaking phase,due to the intensified development of specimen surface fracture ITF mutations emerged,which,on one hand,led the infrared temperature variation rate to moving to the direction of the maximum or minimum significantly and resulted in the abnormal jump of R,on the other hand,made a sudden increase in the number of ITVF abnormal points,thus caused the sudden jump of K and MR.The study provided new ideas,new methods to the exploration of rock catastrophic infrared anomaly.(3)On the basis of exploring their infrared anomalous precursor for rock failure in multiple perspectives,the paper selected two characteristic parameters of ITVF— range(R)and kurtosis(K),took RBF as kernel function,adopt grid-search and cross-validation method to determine the optimal parameter combination(C,sigma),and constructed an early warning model of rock instability based on SVM theory,with merging multi-source monitoring data.The result shows that averaging transformation method is suited for the pretreatment of ITVF characteristic parameters sequence;that the training sample set established in terms of infrared monitoring data belongs to uneven training set,and under-sampling method can make the uneven set equalization and optimize classifier performance;that the early warning model has strong learning ability and generalization ability by inspection,and it is appropriate for the recognition of infrared anomaly signal during rock failure.The study has an important theoretical guiding significance to improving the pressure warning precision and protecting the safety of mine production.