Inverse Method For Analyzing Parameters Of Explosion Source Based On Trace Features At Blast Field And Its Application

The inverse problem of the feature parameters of explosive sources is of vitalimportance in the investigation of accidental explosion cases. In order to obtain significantinformation including the depth of a burst and the mass of the explosive used in the events,this paper draws on frontier technology of data-driven modeling(general regression neuralnetwork GRNN, particle swarm optimization PSO, support vector machine SVM, as well asnon-linear and non-stationary signal analysis algorithms HHT/EMD), then proposesinverse methods for analyzing parameters of explosion source based on trace features atexplosion scenes. Finally, these inverse methods are applied to the analysis module offeature parameters of explosion source which is the core module of the software―Theexplosion scene analysis supporting system‖. The main studies and results are brieflydescribed as follows:1.During the investigation of explosion events, it is significant to grasp the depth of aburst and the mass of the explosive used in attacks as soon as possible. Based on themathematical theory of nonlinear regression, we proposed the inverse analysis method ofthe feature parameters of explosive sources based on the crater sizes. Then in order tovalidate the inverse analysis method, we compared the modeling results with the resultscalculated with empirical equations and the experimental results respectively. The mass ofthe explosives obtained with GRNN was consistent with that used in the experiments in thesoils(clay, sand and sandy clay) which indicating that the inversion results by GRNN havethe higher accuracy than the results by the empirical equation. The relative error of the massof the explosive and the charge location depth under the ground obtained with GRNN is lessthan30%; the corresponding average errors are as follows respectively:15.41%,16.93%.2. Firstly, the function consisted of fractal box dimension and damped exponential ofseismic wave vibration speed is put forth and data processing method EMD/HHT whichwidely used in energy and spectral characteristics analysis of explosion seismic vibrationsignals is improved in the paper so as to reduce the end effects in the process ofdecomposition by combining SVM with PSO. Simulation results indicate that the extension method for data based on PSO-SVM method can restrain the end effects effectively andimprove the accuracy and reliability of the signal decomposition. Then, the paper presents amethod on inversion of the delay time interval in little-difference explosion based onimproved EMD so as to achieve inversion targets. Finally, we proposed the inverse analysismethod of explosive sources based on vibration velocity, distance between explosion sourceand monitoring sites, which are two characterizing parameters of surface ground vibrationcaused by explosion. Then inverse model training between explosive charge and (vibrationvelocity, distance) is undergone and implicitly nonlinear relation between the two sets isobtained so as to construct the data driven modeling of explosive charge based on vibrationparameters of explosion. Comparison between inversing results and those original explosivecharge in the experiments are also conducted and the inversion precision is relativelysatisfied. The method reduces the average error of the inversion mass from51.57%obtainedwith the sadov’s formula to13.32%computed by GRNN.3.Inverse method of explosive charge based on geometric characteristics (the length,width and thickness) of the plate glass under shockwave load and the distance between theexplosion source and the glass is proposed with the help of GRNN. Then experimentalverification with the GRNN model are conducted. The method reduces the average error ofthe inversion mass from51.15%obtained with the sadov’s formula to19.3%computed byGRNN. Meanwhile, for critical pressure data table involving different sizes of glass underoverpressure damage, a supplement to the table is the critical overpressure value which thesmall thickness glass can withstand by data extension method PSO-SVM, thus providingenough data for the inverse.4. Systematic inverse analysis of explosion source based on three main factors (cratersresulted from explosion in soil, characteristics of the glass under shockwave load, blastvibration parameters) is build up for the first time. That is: comprehensive inverse analysisof explosion source based on craters and characteristics of the glass under shockwave load,comprehensive inverse analysis of explosion source based on craters and blast vibrationparameters, comprehensive inverse analysis of explosion source based on craters,characteristics of the glass under shockwave load and blast vibration parameters.5.The three kinds of single factor inverse methods and these three kinds of comprehensive inverse methods are applied to the analysis module of feature parameters ofexplosion source which is the core module of the software―The explosion scene analysissupporting system‖. The software will always be an important means of explosion casesinvestigation, and provides the scene analysis support tools for the investigators and relatedexperts.The main academic contributions of the thesis is introducing data-driven philosophyand technology into analyzing parameters of explosion source based on traces of explosionscenes and proposing several inverse methods for analyzing parameters of explosion sourcewhich have the higher accuracy than the traditional equations after experimentalexamination and validation. The research for analysis on explosion scene has practicalsignificance, and provides technical support for the analysis and detection of explosivecases.