Prediction of residual stress in precision milling of AISI H13 steel

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Surface integrity describes the attributes of a surface and it influences the functional performance of a work piece significantly. Residual stress is one of the major characterization parameters of surface integrity. Non-favorable residual stresses on a machined surface can reduce the fatigue life and performance of the machined part. It therefore requires a prediction model for residual stress in order to establish machining strategy to obtain favorable residual stress for prolonged fatigue life. Hardened tool steels have been widely used to make molds and dies by precision milling in aerospace and automotive industries. Knowledge of the relationship between residual stress on the machined molds and machining conditions is very important for process control. In this work, a prediction model for residual stress was developed by using a model-based approach on an Artificial Neural Network. This model is expected to predict the residual stress based on cutting parameters such as cutting speed, feed rate, depth of cut and tool lead angle. Several precision milling trials were carried out using a central composite design method. The networks have been trained and validated by experimental results. The performance of a feed forward neural network model with backpropagation was assessed and compared with a radial basis function network model by criterion of least mean squared error. Furthermore, the neural network prediction model was supported by the finite element simulation of the milling process to understand the formation mechanism of the residual stress in the machined surface. It was found, that the predicted values by the neural network model matched well with the experimental results. The radial basis function network showed better results than the feed forward network and was therefore chosen to take forward in the analysis. The feed rate was in this case the most influential factor, because it contributes significantly to heat and deformation on the work piece. The model could be used to optimize machining processes to obtain machining strategy for generating favorable residual stress and increasing fatigue life performance of the machined parts.
LanguageEnglish
Pages329-334
Number of pages6
JournalProcedia CIRP
Volume71
Early online date6 Jun 2018
DOIs
Publication statusE-pub ahead of print - 6 Jun 2018

Fingerprint

Residual stresses
Steel
Machining
Radial basis function networks
Molds
Fatigue of materials
Neural networks
Milling (machining)
Tool steel
Aerospace industry
Feedforward neural networks
Backpropagation
Automotive industry
Process control
Lead
Composite materials

Keywords

  • residual stress prediction
  • artificial neural network
  • tool steel
  • HSM

Cite this

@article{64fd0794083c4e8193ba967fe4529f5e,
title = "Prediction of residual stress in precision milling of AISI H13 steel",
abstract = "Surface integrity describes the attributes of a surface and it influences the functional performance of a work piece significantly. Residual stress is one of the major characterization parameters of surface integrity. Non-favorable residual stresses on a machined surface can reduce the fatigue life and performance of the machined part. It therefore requires a prediction model for residual stress in order to establish machining strategy to obtain favorable residual stress for prolonged fatigue life. Hardened tool steels have been widely used to make molds and dies by precision milling in aerospace and automotive industries. Knowledge of the relationship between residual stress on the machined molds and machining conditions is very important for process control. In this work, a prediction model for residual stress was developed by using a model-based approach on an Artificial Neural Network. This model is expected to predict the residual stress based on cutting parameters such as cutting speed, feed rate, depth of cut and tool lead angle. Several precision milling trials were carried out using a central composite design method. The networks have been trained and validated by experimental results. The performance of a feed forward neural network model with backpropagation was assessed and compared with a radial basis function network model by criterion of least mean squared error. Furthermore, the neural network prediction model was supported by the finite element simulation of the milling process to understand the formation mechanism of the residual stress in the machined surface. It was found, that the predicted values by the neural network model matched well with the experimental results. The radial basis function network showed better results than the feed forward network and was therefore chosen to take forward in the analysis. The feed rate was in this case the most influential factor, because it contributes significantly to heat and deformation on the work piece. The model could be used to optimize machining processes to obtain machining strategy for generating favorable residual stress and increasing fatigue life performance of the machined parts.",
keywords = "residual stress prediction, artificial neural network, tool steel, HSM",
author = "Andreas Reimer and Xichun Luo",
year = "2018",
month = "6",
day = "6",
doi = "10.1016/j.procir.2018.05.036",
language = "English",
volume = "71",
pages = "329--334",
journal = "Procedia CIRP",
issn = "2212-8271",

}

Prediction of residual stress in precision milling of AISI H13 steel. / Reimer, Andreas; Luo, Xichun.

In: Procedia CIRP, Vol. 71, 06.06.2018, p. 329-334.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Prediction of residual stress in precision milling of AISI H13 steel

AU - Reimer, Andreas

AU - Luo, Xichun

PY - 2018/6/6

Y1 - 2018/6/6

N2 - Surface integrity describes the attributes of a surface and it influences the functional performance of a work piece significantly. Residual stress is one of the major characterization parameters of surface integrity. Non-favorable residual stresses on a machined surface can reduce the fatigue life and performance of the machined part. It therefore requires a prediction model for residual stress in order to establish machining strategy to obtain favorable residual stress for prolonged fatigue life. Hardened tool steels have been widely used to make molds and dies by precision milling in aerospace and automotive industries. Knowledge of the relationship between residual stress on the machined molds and machining conditions is very important for process control. In this work, a prediction model for residual stress was developed by using a model-based approach on an Artificial Neural Network. This model is expected to predict the residual stress based on cutting parameters such as cutting speed, feed rate, depth of cut and tool lead angle. Several precision milling trials were carried out using a central composite design method. The networks have been trained and validated by experimental results. The performance of a feed forward neural network model with backpropagation was assessed and compared with a radial basis function network model by criterion of least mean squared error. Furthermore, the neural network prediction model was supported by the finite element simulation of the milling process to understand the formation mechanism of the residual stress in the machined surface. It was found, that the predicted values by the neural network model matched well with the experimental results. The radial basis function network showed better results than the feed forward network and was therefore chosen to take forward in the analysis. The feed rate was in this case the most influential factor, because it contributes significantly to heat and deformation on the work piece. The model could be used to optimize machining processes to obtain machining strategy for generating favorable residual stress and increasing fatigue life performance of the machined parts.

AB - Surface integrity describes the attributes of a surface and it influences the functional performance of a work piece significantly. Residual stress is one of the major characterization parameters of surface integrity. Non-favorable residual stresses on a machined surface can reduce the fatigue life and performance of the machined part. It therefore requires a prediction model for residual stress in order to establish machining strategy to obtain favorable residual stress for prolonged fatigue life. Hardened tool steels have been widely used to make molds and dies by precision milling in aerospace and automotive industries. Knowledge of the relationship between residual stress on the machined molds and machining conditions is very important for process control. In this work, a prediction model for residual stress was developed by using a model-based approach on an Artificial Neural Network. This model is expected to predict the residual stress based on cutting parameters such as cutting speed, feed rate, depth of cut and tool lead angle. Several precision milling trials were carried out using a central composite design method. The networks have been trained and validated by experimental results. The performance of a feed forward neural network model with backpropagation was assessed and compared with a radial basis function network model by criterion of least mean squared error. Furthermore, the neural network prediction model was supported by the finite element simulation of the milling process to understand the formation mechanism of the residual stress in the machined surface. It was found, that the predicted values by the neural network model matched well with the experimental results. The radial basis function network showed better results than the feed forward network and was therefore chosen to take forward in the analysis. The feed rate was in this case the most influential factor, because it contributes significantly to heat and deformation on the work piece. The model could be used to optimize machining processes to obtain machining strategy for generating favorable residual stress and increasing fatigue life performance of the machined parts.

KW - residual stress prediction

KW - artificial neural network

KW - tool steel

KW - HSM

UR - https://www.sciencedirect.com/journal/procedia-cirp

U2 - 10.1016/j.procir.2018.05.036

DO - 10.1016/j.procir.2018.05.036

M3 - Article

VL - 71

SP - 329

EP - 334

JO - Procedia CIRP

T2 - Procedia CIRP

JF - Procedia CIRP

SN - 2212-8271

ER -