Identifying the scope for the use of statistics to improve the outcomes of manufacturing processes, the School of Mathematics and Statistics approached the machining group to work on a project with the aim to create time and cost efficiencies by automating the selection of cutting parameters for machining titanium components.
“The challenge is to do this by creating a process that is robust against the variations in material properties,” says University of Sheffield professor of statistics, Jeremy Oakley.
“The problem manufacturers have when machining material such as titanium, is that the material properties can vary from one batch to the next and require new cutting parameters.
“However you would not necessarily know they have changed until identified in the quality checks of finished components,” he adds.
AMRC machining group project engineer, Hatim Laalej, says: “The variation in material batches not only affects dimensional accuracy and surface quality of a finished component, but also tool life during machining, all which contribute to waste and scrappage.
“At the moment a machine operator observes the cutting process at pre-determined times, manually stopping the machine to check on the cutting tool, but this can be a costly process and relies on the experience of the machine operator.”
The AMRC Machining Group conducted physical cutting trials on batches of titanium alloys with different properties, and used an orthogonal peripheral climb milling operation to collect data such as temperature, cutting forces and vibration. A finite element (FE) model which replicated the machining process was also used to extract the same data through simulations of the process.
University statisticians used the output data from the cutting trials and FE model to identify robust optimal cutting parameters to use during the manufacturing process, which allow for the uncertainty of the material properties changing between batches.
Project research associate Dr Keith Harris from the University of Sheffield says it was fairly new to use this kind of statistical modelling within manufacturing: “The challenge here is in how to summarise large amounts of data from multiple sensors and integrate the data with the FE model predictions to get useful, usable results. One aim is to identify correlations in the data to predict the average lifetime of a machine tool.”
Following the identification of optimal cutting parameters, the second stage of the project involved tool wear tests successfully completed at the AMRC. Sensor data from these experiments was used to develop a statistical process control strategy to automate the decision of when to replace the cutting tool.
A feedback adjustment method is now being developed for taking corrective action to prolong the life of the tool, adds Oakley. He says: “This will allow the tool piece and machine to react to the properties of the material and automate the decision to adjust the cutting parameters independently; without the operator having to stop the process.”
Hatim Laalej says: “A fully automated system could be applied to any manufacturing processes outside of the titanium milling process. This will ensure the quality of components is standardised, no matter what the variations in the properties of the material and will save manufacturers time, cut waste and minimise the financial cost of producing any component.”