高强韧硬质合金刀具焊接区热损伤动态识别方法OA

In the dynamic identification process of thermal damage in the welding area of alloy cutting tools,the non-linear response of the non-uniformity correction circuit firmware in the multispectral pulse thermal excitation system leads to the superposition of artifacts and local distortion in the thermal image.The high frame rate infrared focal plane array acquisition macro thermal imaging lens group introduces depth differences and thermal diffusion artifacts in near-field observation,further exacerbating the confusion and masking of the characteristics of the thermal damage area,making it difficult for traditional methods to achieve stable and accurate extraction and identification of the thermal damage area.Propose a dynamic identification method for thermal damage in the welding zone of high-strength and tough hard alloy cutting tools.Firstly,the thermal response differential method is used to process the infrared thermal image,effectively suppressing the superposition of artifacts and local distortions caused by the nonlinear response of the non-uniformity correction circuit,separating the surface thermal emissivity and temperature,and significantly enhancing the real thermal damage area;Secondly,an improved distance preserving level set evolution model(DRLSE)is introduced to achieve automatic and accurate extraction and contour localization of damaged areas under the interference of complex thermal diffusion artifacts and depth of field differences;Finally,principal component analysis(PCA)is combined to perform feature dimensionality reduction on high-dimensional dynamic thermal sequences,and support vector machine(SVM)classifiers are used to achieve robust recognition of thermal damage and non damage areas,effectively overcoming the problems of feature confusion and noise interference in high frame rate acquisition.The experimental results show that the proposed method still has excellent anti-interference and feature extraction capabilities in complex thermal excitation and imaging hardware environments.The recognition accuracy of four types of thermal damage,including microcracks and ablation,reaches 100%.The F1 value is superior to the comparative method at different degrees of damage(up to 0.93),and the computational complexity and memory usage are significantly reduced(GFLOPs 1.1-1.3,parameter size 2.4×10⁶-2.6×10⁶),achieving high-precision and high-efficiency dynamic thermal damage recognition.