Lean Manufacturing Soft Sensors for Automotive Industries

Lean and flexible manufacturing is a matter of necessity for the automotive industries today. Rising consumer expectations, higher raw material processing costs, dynamic market conditions are driving auto sector to become smarter agile. This paper presents machine learning-based soft sensor approach identification prediction lean (LM) levels based on their performances over multifarious flexibilities such as volume flexibility, routing product labour handling. study was database associated collected from 46 component enterprises located in Pune region Maharashtra State, India. As many 29 different learning models belonging seven architectures were explored develop sensors. These sensors trained classify firms into high, medium or low flexibilities. The included Decision Trees, Discriminants, Naive Bayes, Support Vector Machine (SVM), K-nearest neighbour (KNN), Ensembles, Neural Networks (NN). all compared basis respective training, validation, testing accuracies, computation timespans. Primary results indicate that neural network provided best predictions, followed by SVM, KNN, Discriminants. trilayered architecture attained highest accuracy 80%. fine, medium, coarse trees 60%, did quadratic cubic SVMs, wide narrow networks, ensemble RUSBoosted trees. Remaining obtained inferior accuracies. performing model further analysed scatter plots predicted LM classes versus flexibilities, validation confusion matrices, receiver operating characteristics (ROC) curves, parallel coordinate plot identifying flexibility trends levels. Thus, can be used create effective predict level an enterprise its