Intelligent Tool Wear Monitoring

POCI/EME/59491/04. 30 months.

 

Summary

Manufacturing industries and their customers are now demanding substantial increase in flexibility, productivity, and reliability from process machines as well as increased quality and value of their products. For machine tools the systems that provide the required feedback are termed tool condition monitors (TCM). A tool condition monitoring system can be viewed as serving the following purposes: advanced fault detection system for cutting and machine tool; check and safeguard machining process stability; means by which machining tolerance is maintained on the workpiece to acceptable limits by providing a compensatory mechanism for tool wear offsets; and, machine tool damage avoidance system. Several factors have impeded advances in the development of TCMSs including inappropriate choice of sensor signals, and their utilisation, and their inability to perform robustly in noisy environments.
Artificial neural networks of sigmoidal and McCulloch-Pitts neurons have found increasing favour in industry research because of their most attractive features, abstraction of hardly accessible knowledge and generalisation from distorted sensor signals. Nevertheless, although working in certain conditions, most of the previous applications of neural networks have some limitations. However, in recent years experimental evidence has been accumulating to suggest that biological neural networks, which communicate through spikes, use the timing of these spikes to encode and compute information in a more efficient way. Also, in cutting tools, numerous sensed signals on the process have a rich temporal structure, and neural circuits must process these in real time. As suggested, these new computationally architectures reveal a greater computational power.
Thus, this project is committed to the development of a methodology for estimation and prediction system based on artificial spiking neuron networks (ASNN). The implications of such a system are evident and of crucial importance towards an efficient scheduling of tool replacements, therefore, playing an important role in industry automation.
The project encompasses research in 3 sub-areas which are thought to be important contributions towards the advance in Tool Condition Monitoring. These are: Feature analysis and sensor design:- the design of reliable and feasible sensors for industrial use; Neural model development:- the development of a spiky neuron network model for real-time information processing (to the full extent of our knowledge, it is a pioneer application in the engineering field) that lead to the separation and identification of sensed objects hidden in tool wear sensed signals that might contribute to a further understanding of wear mechanisms and their nature; and, System integration:- the development of a system to integrate, in a modular fashion, all components into a real-time flexible system.

 

Objectives

To the study and develop a real-time estimation and prediction tool wear monitoring methodology using Artificial Spiking Neural Networks (ASNN). It consists of six components: data collection; feature extraction; pattern recognition; multi-sensor integration; tool wear estimation; and, one-step-ahead tool flank wear prediction. For sensed information, a self-organizing map (SOM) will be developed to integrate the spikiness nature of neurons. This network will be employed to recognize and synthesize the extracted features as belonging to different tool wear degradation levels. Such a model will be used to learn from the experienced time evolution in order to forecast future levels of wear based on temporal representations. The results from applying spiking neural networks leads to the separation and identification of sensed objects that might contribute to a further understanding of mechanisms and its nature hidden in tool wear sensed signals. This system will be targeted, in terms of validation, for turning operations.

 

Publications

Silva, RG (2005) A Robust Methodology for Tool Condition Monitoring using Spiking Neuron Networks, Proceedings of the Ninth IASTED International Conference on Artificial Intelligence and Soft Computing. Setembro 12-14, Benidorm, Espanha. ISBN 0-88986-536-1. (artigo apresentado durante a conferência).

 

Silva, R.G., Wilcox, S.J. and Reuben, R.L. (2006) Development of a system for monitoring tool wear using artificial intelligence techniques. Proceedings of the Institution of Mechanical Engineers, Part B: J. Engineering Manufacture, 220(B8), 1333?1346.

 

Silva, R. Pereira, V. Oliveira, C. and Ferreira, J. (2006) Rede De Neurónios Impulsivos Aplicada À Monitorização De Desgaste De Ferramentas De Corte.
Tecnometal.

 

Silva, R.G. (2006) Tool condition monitoring of the turning process using spiking neuron networks. Lusíada. Engenharia. ISSN: 1645-8915. Nº 2/4, p.
5-17.

 

Silva, R.G., Wilcox, S.J. and Araújo, A. J. (2007) Multi-Sensor Condition Monitoring Using Spiking Neuron Networks. Proceedings of the International Conference on Applied Computing. February 18-20, Salamanca, Spain. Full Paper Presented.

 

Lima, R.M. and Silva, R.G. (2007) GRID e-Services for Multi-Layer SOM Neural Network Simulation 1st Iberian Grid Infrastructure Conference Procedings.
277-286, ISBN 978-84-611-6634-3. Full Paper Presented.

 

Silva, R. Gomes, P. and Cruz, J. (2007) Retrofitting de um Torno de Comando Numérico. Tecnometal.

 

Silva, R. (2008) Sensor Based Condition Monitoring Using A Self-Organizing Spiking Neural Networks Map. IADIS International Conference on Applied Computing 2008, Algarve, Portugal. 10-13 April.

 

Silva, R.G. and Wilcox, S.J. (2008) Sensor Based Condition Monitoring Feature Selection Using A Self-Organizing Map. The 2008 International Conference of Manufacturing Engineering and Engineering Management. Londres, U.K. 2-4 Julho.

 

Rui G. Silva, Carlos R. Oliveira e Pedro R. Gomes (submetido em 2007) Desenvolvimento De Um Sistema Integrado De Controlo E Monitorização De Um Torno De Comando Numérico. Robótica, Publindustria.

 

Silva, R.G (submetido) Condition Monitoring Of The Cutting Process Using A Self-Organizing Spiking Neural Network Map. Journal of Intelligent Manufacturing.

 

Araújo, A.J., Wilcox, S.J. and Silva, R.G. (submetido) An Investigation on the Impact of Metal Cutting Conditions on Cutting Forces, Vibration Signals, and Acoustic Emissions through the Use of Feed-Forward Neural Networks. Materials and Manufacturing Processes.

 

Research Team

Rui Gabriel Araujo de Azevedo Silva (Principal Investigator)
José Manuel dos Santos Cruz
José Francisco Ferreira
Carlos Alberto Rego de Oliveira
Rui Manuel Dias Ferreira Lima