Artificial-neural-networks-based surface roughness Pokayoke system for end-milling operations
Introduction
Most machining centers used in industry have installed computer numerical controller (CNC) in applications such as CNC milling and turning. The wide implementation of CNC machines owes to the fact that they can not only increase productivity but also improve the accuracy of products [1]. However, product defects continue to result due to the uncontrollable machining factors, such as tool conditions and chatter effects.
The strategy of quality control was proposed to assist manufacturing companies in the area of quality assurance. However, the implementation of quality control procedures is time-consuming and labor-intensive, and hence increases the production cost. For this reason, an in-process monitoring of the machining process has received extensive research recently. For example, the detection of tool breakage [2], [3] and the monitoring of tool wear [4], [5] focused on reducing the time required to monitor tool conditions; online geometric error identification and compensation [6], [7], [8], [9] intended to improve machining accuracy. Another in-process monitoring system predicted the surface roughness within cutting processes [10], [11]. Surface roughness is a common quality characteristic which is affected by factors such as spindle speed, feed rate, and depth of cut [10]. In-process surface roughness monitoring has been used to not only predict the surface roughness, but also to diagnose the tool conditions [12], [13]. Therefore, successfully developing an in-process surface roughness monitoring system would be more useful and efficient among in-process monitoring systems since it can not only predict cutting surface roughness, but also respond to the cutting tool conditions.
However, current research concerning in-process surface roughness monitoring is limited in that the monitoring systems studied thus far can only predict the value of surface roughness [14], [15], [16]. The research has not found an effective means to change the machining conditions upon detection of a defect. This paper describes the extension of our previous work on in-process NN-based surface roughness prediction (INN-SRP) system [17] into a neural network-based surface roughness Pokayoke (NN-SRPo) system. Such an NN-SRPo system is able to adjust the machine parameters on-line while a defect of surface roughness is detected by the system.
Section snippets
The architecture of the NN-SRPo system
The NN-SRPo system is designed to provide real-time in-process surface roughness monitoring and controlling, which can adjust the machining parameters to create the desired surface roughness in end-milling operations. Fig. 1 shows the architecture of the NN-SRPo system. The NN-SRPo system consists of two sub-systems. One is the INN-SRP and the other is the NN-based adaptive machining parameter control (NN-AMPC) system. The output of the INN-SRP subsystem is the predicted surface roughness,
The development of the NN-SRPo system
The development of INN-SRP system was completed in the authors’ previous work [17]. This section introduces the NN-AMPC subsystem (the architecture is shown in Fig. 2) of the NN-SRPo system.
The procedures for developing an NN-AMPC subsystem are similar to those used to develop the INN-SRP system. The input parameters of this subsystem are the spindle speed (), feed rate (), average resultant peak force in the XY plane (), absolute average force in the Z direction () from the
The testing of the NN-SRPo system and testing results
The NN-SRPo system was tested to show its efficacy. The length of the workpiece was 2.5 inches shown in Fig. 5. The cutting tool moved to cut the workpiece in the X-axis for 1.25 inches and then stopped. Within this length, the program of the NN-SRPo system was activated to predict the surface roughness and provided the adaptive degree of feed rate when the predicted surface roughness was larger than the desired one. The feed rate was changed according to the adaptive degree of feed rate and
Conclusion
This paper presented the application of neural networks algorithm in an in-process Pokayoke system to perform the adaptive surface roughness control in end-milling operations. To control the quality of surface roughness, the NN-SRPo system can not only predictively evaluate the surface roughness within the cutting process, but also provide the adaptive function to adjust machining parameters to generate the desired surface roughness.
The INN-SRP subsystem and the NN-AMPC subsystem were
Dr. Bernie P. Huang is an assistant professor in the Department of Industrial Engineering at Chung Yuan Christian University. He received his Ph.D. degree in Industrial Technology at Iowa State University in 2002. His research and teaching interests are the CAD/CAM integration, design for manufacturability, lean cell manufacturing system, cost engineering, and application of fuzzy-neuron control and statistical tools in machining detection.
References (17)
- et al.
A fuzzy-nets in-process (FNIP) system for tool-breakage monitoring in end-milling operations
Int. J. Mach. Tool Manuf.
(1997) - et al.
On-line free form surface measurement via a fuzzy-logic controlled scanning probe
Int. J. Mach. Tool Manuf.
(1999) - et al.
Surface topography characterization in finish milling
Int. J. Mach. Tool Manuf.
(1994) - et al.
Prediction of surface roughness and dimensional deviation by measuring cutting forces and vibrations in turning
J. Mater. Process. Technol.
(2003) - et al.
Fuzzy adaptive networks in machining process modeling: surface roughness prediction for turning operations
Int. J. Mach. Tool Manuf.
(2004) - et al.
In-process surface roughness prediction using displacement signals from spindle motion
Int. J. Mach. Tool Manuf.
(2007) Computer Numerical Control-From Programming to Networking
(1994)- et al.
Tool condition monitoring in machining by fuzzy neural networks
J. Dyn. System Meas. Control
(1996)
Cited by (41)
Roughness maps to determine the optimum process window parameters in face milling
2022, International Journal of Mechanical SciencesCitation Excerpt :Finally, Kadirgama et al. [32] compared RSM, SVM and Potential Support Vector Machine (PSVM), concluding that the most accurate results are obtained with the last one. Other than to predict, Ozcelik et al. [33] and Oktem et al. [34] used ANN to optimise cutting conditions for minimising surface roughness and Huang et al. [35] to adapt the feed rate to a limit roughness. In addition, Oktem [36] used ANN for adapting the cutting conditions to achieve the desired roughness, Escamilla et al. [37] for optimising cutting speed, feed rate and depth of cut and Malghan et al. [38] for surface roughness, cutting forces and power consumption.
The development of an ANN surface roughness prediction system of multiple materials in CNC turning
2023, International Journal of Advanced Manufacturing TechnologyApplication of ML & Al for energy efficiency in manufacturing
2023, Sustainable Materials and Manufacturing TechnologiesPoka Yoke Meets Deep Learning: A Proof of Concept for an Assembly Line Application
2022, Applied Sciences (Switzerland)Poka yoke meets deep learning: A proof of concept for an assembly line application
2021, Research Square
Dr. Bernie P. Huang is an assistant professor in the Department of Industrial Engineering at Chung Yuan Christian University. He received his Ph.D. degree in Industrial Technology at Iowa State University in 2002. His research and teaching interests are the CAD/CAM integration, design for manufacturability, lean cell manufacturing system, cost engineering, and application of fuzzy-neuron control and statistical tools in machining detection.
Joseph C. Chen, Ph.D., PE is a Professor in the Department of Agriculture and Biosystems Engineering at Iowa State University. He received his M.S. and Ph.D. degrees from the Department of Industrial and System Engineering at Auburn University in 1990 and 1994, respectively. His teaching interests include: Lean manufacturing system design, automated manufacturing processes, facility design, Taguchi design in quality, etc. His research interests include: manufacturing system control, cellular manufacturing system design, smart CNC machining, Taguchi parameter design in manufacturing processes.
Ye Li is a Ph.D. candidate in the Department of Industrial and Manufacturing Systems Engineering at Iowa State University. He received his B.S. and M.S. degrees in Mechanical Engineering from Shanghai Jiao Tong University, PR China in 2000 and 2003, respectively. His research interests include: CAD/CAM theory with application in Manufacturing Engineering and Medical Planning; Manufacturing Processes and Rapid Prototyping; Life Cycle Design Performance Prediction.