Mechanical Property Prediction of Strip Model Based on PSO-BP Neural Network

Mechanical property prediction of hot rolled strip is one of the hotspots in material processing research. To avoid the local infinitesimal defect and slow constringency in pure BP algorithm, a kind of global optimization algorithm-particle swarm optimization （PSO） is adopted. The algorithm is combined with the BP rapid training algorithm, and then, a kind of new neural network （NN） called PSO-BP NN is established. With the advantages of global optimization ability and the rapid constringency of the BP rapid training algorithm, the new algorithm fully shows the ability of nonlinear approach of multilayer feedforward network, improves the performance of NN, and provides a favorable basis for further online application of a comprehensive model.

PREDICTION OF MECHANICAL PROPERTY OF WHISKER REINFORCED METAL MATRIX COMPOSITE: PART-I. MODEL AND FORMULATION

Based on study of strain distribution in whisker reinforced metal matrix composites, an explicit precise stiffness tensor is derived. In the present theory, the effect of whisker orientation on the macro property of composites is considered, but the effect of random whisker position and the complicated strain field at whisker ends are averaged. The derived formula is able to predict the stiffness modulus of composites with arbitrary whisker orientation under any loading condition. Compared with the models of micro mechanics, the present theory is competent for modulus prediction of actual engineering composites. The verification and application of the present theory are given in a subsequent paper published in the same issue

Artificial neural network prediction of mechanical properties of hot rolled low carbon steel strip

Conventionally, direct tensile tests are employed to measure mechanical properties of industrially pro- duced products. In mass production, the cost of sampling and labor is high, which leads to an increase of total pro- duction cost and a decrease of production efficiency. The main purpose of this paper is to develop an intelligent pro- gram based on artificial neural network （ANN） to predict the mechanical properties of a commercial grade hot rolled low carbon steel strip, SPHC. A neural network model was developed by using 7 x 5 x 1 back-propagation （BP） neural network structure to determine the multiple relationships among chemical composition, product pro- cess and mechanical properties. Industrial on-line application of the model indicated that prediction results were in good agreement with measured values. It showed that 99.2 % of the products＇ tensile strength was accurately pre- dicted within an error margin of ~ 10 %, compared to measured values. Based on the model, the effects of chemical composition and hot rolling process on mechanical properties were derived and the relative importance of each in- put parameter was evaluated by sensitivity analysis. All the results demonstrate that the developed ANN models are capable of accurate predictions under real-time industrial conditions. The developed model can be used to sub- stitute mechanical property measurement and therefore reduce cost of production. It can also be used to control and optimize mechanical properties of the investigated steel.

Analysis of Hot Rolling Routes of AZ31B Magnesium Alloy and Prediction of Tensile Property of Hot-rolled Sheets

At the initial rolling temperature of 250 to 400 ℃, AZ31B magnesium alloy sheets were hot rolled by four different rolling routes. Microstructures and mechanical properties of the hot-rolled magnesium alloy sheets were analyzed by optical microscope and tensile tests respectively. Based on the Hall-Petch relation, considering the average grain size and grain size distribution, the nonlinear fitting analysis between the tensile strength and average grain size was carried on, and then the prediction model of tensile strength of hot-rolled AZ31B magnesium alloy sheet was established. The results indicate that, by rolling with multi-pass cross rolling, uniform, fine and equiaxial grain microstructures can be produced, the anisotropy of hot-rolled magnesium sheet can also be effectively weakened. Strong correlation was observed between the average grain size and tensile property of the hot-rolled magnesium alloy sheet. Grain size distribution coefficient d（CV） was introduced to reflect the dispersion degree about a set of grain size data, and then the Hall-Petch relation was perfected. Ultimately, the prediction accuracy of tensile strength of multi-pass hot-rolled AZ31B magnesium alloy was improved, and the prediction of tensile property can be performed by the model.

Deep learning for predictive mechanical properties of hot-rolled strip in complex manufacturing systems

Higher requirements for the accuracy of relevant models are put throughout the transformation and upgrade of the iron and steel sector to intelligent production.It has been difficult to meet the needs of the field with the usual prediction model of mechanical properties of hotrolled strip.Insufficient data and difficult parameter adjustment limit deep learning models based on multi-layer networks in practical applications;besides,the limited discrete process parameters used make it impossible to effectively depict the actual strip processing process.In order to solve these problems,this research proposed a new sampling approach for mechanical characteristics input data of hot-rolled strip based on the multi-grained cascade forest(gcForest)framework.According to the characteristics of complex process flow and abnormal sensitivity of process path and parameters to product quality in the hot-rolled strip production,a three-dimensional continuous time series process data sampling method based on time-temperature-deformation was designed.The basic information of strip steel(chemical composition and typical process parameters)is fused with the local process information collected by multi-grained scanning,so that the next link’s input has both local and global features.Furthermore,in the multi-grained scanning structure,a sub sampling scheme with a variable window was designed,so that input data with different dimensions can get output characteristics of the same dimension after passing through the multi-grained scanning structure,allowing the cascade forest structure to be trained normally.Finally,actual production data of three steel grades was used to conduct the experimental evaluation.The results revealed that the gcForest-based mechanical property prediction model outperforms the competition in terms of comprehensive performance,ease of parameter adjustment,and ability to sustain high prediction accuracy with fewer samples.

Fast determination of meso-level mechanical parameters of PFC models

To solve the problems of blindness and inefficiency existing in the determination of meso-level mechanical parameters of particle flow code (PFC) models, we firstly designed and numerically carried out orthogonal tests on rock samples to investigate the correlations between macro-and meso-level mechanical parameters of rock-like bonded granular materials. Then based on the artificial intelligent technology, the intelligent prediction systems for nine meso-level mechanical parameters of PFC models were obtained by creating, training and testing the prediction models with the set of data got from the orthogonal tests. Lastly the prediction systems were used to predict the meso-level mechanical parameters of one kind of sandy mudstone, and according to the predicted results the macroscopic properties of the rock were obtained by numerical tests. The maximum relative error between the numerical test results and real rock properties is 3.28% which satisfies the precision requirement in engineering. It shows that this paper provides a fast and accurate method for the determination of meso-level mechanical parameters of PFC models.

Prediction of mechanical properties of cold rolled strip based on improved extreme random tree

Taking the 2130 cold rolling production line of a steel mill as the research object,feature dimensionality reduction and decoupling processing were realized by fusing random forest and factor analysis,which reduced the generation of weak decision trees while ensured its diversity.The base learner used a weighted voting mechanism to replace the traditional average method,which improved the prediction accuracy.Finally,the analysis method of the correlation between steel grades was proposed to solve the problem of unstable prediction accuracy of multiple steel grades.The experimental results show that the improved prediction model of mechanical properties has high accuracy:the prediction accuracy of yield strength and tensile strength within the error of±20 MPa reaches 93.20%and 97.62%,respectively,and that of the elongation rate under the error of±5%has reached 96.60%.

Prediction of Mechanical Properties of Hot Rolled Strip by Using Semi-Parametric Single-Index Model

A semi-parametric single-index model based approach was proposed for prediction of mechanical properties of hot rolled strip. Based on industrial production data, a semi-parametric single-index model was developed by choo-sing the appropriate kernel function and window width to predict the yield strength, tensile strength and elongation. When data samples are limited, compared with regression method and neural network method, the prediction results show that the semi-parametric single-index model based method is more adaptive and the prediction performance is superior to those by both regression and neural network methods.

Modeling hot strip rolling process under framework of generalized additive model

This research develops a new mathematical modeling method by combining industrial big data and process mechanism analysis under the framework of generalized additive models(GAM)to generate a practical model with generalization and precision.Specifically,the proposed modeling method includes the following steps.Firstly,the influence factors are screened using mechanism knowledge and data-mining methods.Secondly,the unary GAM without interactions including cleaning the data,building the sub-models,and verifying the sub-models.Subsequently,the interactions between the various factors are explored,and the binary GAM with interactions is constructed.The relationships among the sub-models are analyzed,and the integrated model is built.Finally,based on the proposed modeling method,two prediction models of mechanical property and deformation resistance for hot-rolled strips are established.Industrial actual data verification demonstrates that the new models have good prediction precision,and the mean absolute percentage errors of tensile strength,yield strength and deformation resistance are 2.54%,3.34%and 6.53%,respectively.And experimental results suggest that the proposed method offers a new approach to industrial process modeling.