Utilizing Machine Data to Avoid Failures

Utilizing Machine Data to Avoid Failures

Predictive machining can be a complex subject, with most manufacturers assuming such use cases are far beyond their means, meant only for the future of manufacturing.

The Data Science team from Machine Metrics has been working hard to identify and deploy predictive solutions at customer sites. Their work explains several different ways in which customers are predicting and preventing costly scenarios with machine data, including:

1. Roughing tool failures, leading to cascading tool failures on finishing tools
2. Endmill failures leading to poor finish and scrap parts
3. Ball bearing failures leading to damaged spindle housing
4. Incorrect offset, leading to guide bushing misalignment
5. Tool failure, leading to incomplete cutoff
6. Incorrect estimation of tool life, leading to overuse or underuse of tools

Using Machine Data to Prevent Cascading Tool Failures

Having an entire column of tools break can be a frustrating experience for any machinist. Unfortunately, this is an all-too common occurrence in the industry. What if we could see when the first tool broke, and then stop the machine immediately before the other tools get taken out?

Turns out, you can. After the first tool goes, the change in load signature on subsequent tools is all but handed to you on a platter.

In one instance of this, we saw the first tool break many part-cycles before the machine stopped itself when it detected something catastrophic. By the time the self-stoppage occurs, other tools had also been taken out, costing the shop hundreds of dollars worth of unnecessary tool replacements. The first tool likely broke because it was worn out and past end-of-life, so it probably couldn’t be saved anyway. But there’s no reason why other tools after it, some freshly replaced, had to go too.

Read the full post from Machine Metrics to learn more about Cascading Tool Failures and the failure types that can be prevented using machine data.

 

About the author

This article was written by Lou Zhang. Lou has extensive experience with both the manufacturing industry and with developing predictive algorithms for time-series data. Prior to MachineMetrics, Lou conducted research with NIST.