Open_TUoS.png

© The University of Sheffield

Jonathan Walsh

Project Title: Using sensor fusion to identify machine tool wear

Academic Supervisor: Dr Tom Slatter

AMRC Supervisor: Dr Tom McLeay

During my physics undergraduate, my Masters project was in conjunction with the AMRC. After seeing the incredible facilities of the AMRC mixed with my own passion for both the application and development of technology, I knew I had to pursue an EngD position.

 

Condition monitoring is a broad and extensively investigated research topic relying on monitoring certain process parameters and variables to infer or predict the operational state of the machine/structure under observation. Tool condition monitoring (TCM) is a branch of condition monitoring that uses the same principles to classify tool wear or breakage. Considering the size and complexity of some parts in the aerospace and nuclear industry, excessive wear and tool breakage can cause significant productivity and financial losses, therefore a more dependable system is necessary. A TCM system needs to improve upon the knowledge, experience and pattern recognition abilities of human operators in order to reliably predict tool wear and therefore replace them in an industrial environment. To do this, a combination of high-quality data obtained from sensors and reliable artificial intelligence decision-making constructs are required to recognize and respond to process abnormalities.

 

It is estimated that downtime due to tool failure is between 6.8% and 20%. The installation of a TCM system will reduce the downtime attributed to tool wear and is predicted to result in savings of between 10-40%. A fully realized TCM system will also be able to alter cutting conditions based on sensor data to maximize tool life and optimize production.

 

Throughout this project, a low-cost solution that can be applied to milling will be sought. Exploring several sensor types and combinations in conjunction with attempted optimization of feature selection and decision making paradigms. The current approach I am working on will use 2 microphones and an acoustic emission sensor placed around a manual vertical milling machine. Once cutting data has been acquired, various statistical analyses will be carried out, ultimately resulting in the output of ranked features to input into a machine learning algorithm to classify the tool wear.

Publications: