Predictive modeling of hard turning processes using neural networks

Tuǧrul Özel, Yiǧit Karpat

Research output: Contribution to conferencePaperpeer-review


In machining of hardened steel parts, surface quality is one of the most specified customer requirements. A principal indication of surface quality on machined parts is surface roughness. Finish dry hard turning process using Cubic Boron Nitride (CBN) tools allows manufacturers to simplify their operations and still achieve the desired surface roughness. In order for manufacturers to maximize their gains from utilizing finish dry hard turning, accurate predictive models for surface roughness and tool wear must be constructed. In this study, we present a predictive modeling approach by utilizing computational static neural networks. The objective is to develop models based on backpropagation neural networks to predict accurately both surface roughness and tool flank wear for finish dry turning of hardened steel parts. This study utilizes neural network modeling to predict surface roughness and tool flank wear over the machining time for a variety of cutting conditions in finish dry hard turning. Regression models are also constructed in order to capture process specific parameters. A set of sparse experimental data obtained from literature and additional data collected from hard turning tests. The data sets of measured surface roughness and tool flank wear for various cutting conditions are used to train neural network models. Trained neural network models are then used in predicting surface roughness and tool flank wear for diverse cutting conditions. A comparison of predictive neural network models with regression models is also condcuted. Predictive neural network models are found to be more accurate in predictions for surface roughness and tool flank wear within the range that they had been trained. In the design of backpropagation neural networks, our major concern was to obtain a good generalization capability. For this purpose, we used Bayesian regularization with the Levenberg-Marquardt training algorithm. This methodology is also utilized to overcome the problem of determining an optimum number of neurons in hidden layers. The results obtained after the simulations proved that this methodology is efficient. Neural network models that are utilizing cutting forces as inputs and providing a single output, either as surface roughness or tool wear, yielded more accurate predictions than the models that are providing two outputs as surface roughness and tool wear simultaneously. Predictive neural network modeling is also extended to predict tool wear trends and surface roughness patterns for finish dry hard turning. A decrease in feed rate resulted in better surface roughness but slightly faster tool wear development. An increase in cutting speed resulted in a significant increase in tool wear development but resulted in better surface roughness. An increase in workpiece hardness resulted in better surface roughness but a higher tool wear. Implementation of these neural network models into a dynamic predictive system was left as future work.

Original languageEnglish (US)
Number of pages1
StatePublished - 2004
EventIIE Annual Conference and Exhibition 2004 - Houston, TX, United States
Duration: May 15 2004May 19 2004


OtherIIE Annual Conference and Exhibition 2004
Country/TerritoryUnited States
CityHouston, TX

All Science Journal Classification (ASJC) codes

  • Engineering(all)


  • Hard turning
  • Predictive neural networks
  • Surface roughness
  • Tool wear


Dive into the research topics of 'Predictive modeling of hard turning processes using neural networks'. Together they form a unique fingerprint.

Cite this