Research On Data-driven Ontological Modeling Of Engineering Materials Based On Machine Learning

Machine learning methods can mine the mapping relationships within the data from a large amount of data and have outstanding advantages in the analysis of experimental data of engineering materials.Concrete is one of the most widely used engineering materials in defense military and civil engineering.In this paper,machine learning techniques are applied to the process of concrete Split Hopkinson Pressure Bar(SHPB)test data analysis to solve the problems of inertia effects and electromagnetic signal interference and other effects on waveforms during the SHPB test.By conducting high-temperature static tests and high-temperature SHPB tests,data acquisition is carried out,and the accuracy of the test stress-strain curve is improved by accurately processing the SHPB test data based on BP neural network;then the high-temperature dynamic mechanical properties of concrete under different strain rates at different temperatures are analyzed in terms of peak stress,peak strain,and strain rate effects;based on the experimental data,genetic algorithm is used to Based on the experimental data,the parameter optimization of the material damage evolution equation was carried out,and the material instantiation equation applicable to the combined effect of high temperature and impact loading was established.Firstly,the SHPB experiment of concrete was carried out,and BP neural network was introduced into the SHPB test data processing process.Based on the BP neural network,the incident,transmitted and reflected waves collected from the test were processed,and the interference of transverse inertia effects and electromagnetic wave signals were separated to obtain the intrinsic waveform that truly reflects the material properties,and the accuracy of the stress-strain curve was improved.Secondly,the stress-strain curves obtained after BP neural network processing were compared and analyzed to verify the advantages of BP neural network in SHPB experimental data processing,and the following conclusions were obtained: for the concrete with the same amount of steel fiber admixture,the peak stress showed an obvious strain rate hardening effect under approximately the same temperature condition;for the concrete with the same amount of steel fiber admixture,the ultimate The results show that K is not a simple dot product of strain rate hardening and temperature softening results under a single test condition(i.e.,high temperature quasi-static compression and ambient temperature dynamic compression),but a simple dot product of strain rate hardening and temperature softening results under a single test condition(i.e.,high temperature quasi-static compression and ambient temperature dynamic compression).Rather,it is the result of competing interactions between temperature and strain rate within the concrete material,and thus the effects of temperature and strain rate on material strength are mutually coupled.Finally,the damage evolution equation parameters were optimized based on the genetic algorithm,and the principal structure equation was established;since the hightemperature softening effect and the strain-rate hardening effect are not simply dotproduct relationships but are coupled together,the damage evolution equation was established by representing the coupled effects of temperature and strain rate through the damage variable D.The parameters in the damage evolution equation were optimized using genetic algorithm,and a concrete principal structure model was established.The stress-strain curves obtained from the tests were compared and analyzed with those calculated using the established principal structure equation,and it was found that they were in good agreement,which further proved that the principal structure equation can effectively describe the high temperature dynamic mechanical properties of concrete materials.