Development of an Approximation Model of Selected Properties of Model Materials Used for Simulations of Bulk Metal Plastic Forming Processes Using Induction of Decision Trees

Abstract

The article discusses the development of an approximation model of selected plastic and mechanical properties obtained from compression tests of model materials used in physical modeling. The use of physical modeling with the use of soft model materials such as a synthetic wax branch with various modifiers is a popular tool used as an alternative or verification of numerical modeling of bulk metal forming processes. In order to develop an algorithm to facilitate the choice of material model to simulate the behavior of real-metallic materials used in industrial production processes the induction of decision trees was used. First of all, the Statistica program was used for data mining, which made it possible to determine / find the relationship between the percentage of particular constituents of the model material (base material and modifiers) and yield strength, critical and maximum strain, and provide the opportunity to indicate the most important variables determining the shape of the stress -strain curve. Next, using the induction of decision trees, an approximation model was developed, which allowed to create an algorithm facilitating the selection of individual modifying components. The last stage of the research was verification of the correctness of the developed algorithm. The obtained research results indicate the possibility of using decision tree induction to approximate selected properties of modeling materials simulating the behavior of real materials, thus eliminating the need for costly and time-consuming experiments carried out on metallic material.

Recognition of Hybrid PQ Disturbances Based on Multi-Resolution S-Transform and Decision Tree

“…The construction idea of decision tree (DT) is as follows: if all the samples in the training sample set are of the same kind, they will be regarded as Leaf nodes; otherwise, according to certain branch division rules, the sample set will be subdivided successively until Leaf nodes. For the same sample set, many decision trees can be generated, and branching rule is crucial to obtain an "optimal" tree [24] Two methods are commonly used, such as Gain ratio standard and Gini index.…”

“…Literature [28] used MST for detection, decision tree for classification, and 5 feature statistics to extract 16 PQDS features, with an accuracy of 99.97%. Literature [24] used S-transformation for detection, chaos integrated decision tree for classification, and 9 feature statistics to extract 23 PQDS features, with an accuracy of 91.9%. In this paper, GST transform is used for detection, and five feature statistics are used to extract the characteristics of the composite disturbance signal, which reduces the amount of calculation and the calculation cost is relatively small compared with the literature [24].…”

“…Literature [24] used S-transformation for detection, chaos integrated decision tree for classification, and 9 feature statistics to extract 23 PQDS features, with an accuracy of 91.9%. In this paper, GST transform is used for detection, and five feature statistics are used to extract the characteristics of the composite disturbance signal, which reduces the amount of calculation and the calculation cost is relatively small compared with the literature [24]. Compared with literature [28], more composite disturbance signals are detected by using the same characteristic statistics.…”

“…As a combination of WT and STFT, S transform not only has good time-frequency analysis ability, but also the transform result is directly related to its Fourier transform, which does not need to meet the wavelet admissibility condition, so it is very suitable for the extraction of power quality disturbance signal features, but it requires a large amount of calculation. The main classification and recognition methods of power quality disturbance signal feature vectors are Fuzzy Logic (FL) [13], support vector machine (SVM) [14,15], K-nearest neighbor (KNN) [16][17][18][19][20][21][22] and decision tree (DT) [23][24][25][26]. Both FL and KNN methods have the advantages of simple principle and easy to understand, but with the increase of disturbance signal types, the fault tolerance and complexity of the system will greatly limit its classification ability [13].…”