Published 10/29/2024

From Process Monitoring to Quality Prediction Software

Even before inspection, Datanomix and Caron Engineering’s TMAC AI can extrapolate part quality from cut information. This information also helps shops push their tool life to its limits.

Evan Doran

Associate Editor, Modern Machine Shop

Real-Time Prediction

TMAC AI relies on data captured through Caron Engineering’s TMAC (Tool Monitoring Adaptive Control) system. Datanomix’s software system analyzes data about cuts from the TMAC sensors in real time, pairing them with a machine-learning model that predicts whether parts are in tolerance or out of tolerance, without waiting for inspection. From here, the system examines the data about whether parts meet tolerance to predict effective tool life, enabling users to push their tool life limits as far as they can while still creating a successful part.

The two companies first conceived of their current partnership just after IMTS 2022, planning to work hand-in-hand to bring advanced analytics to the common issues Caron’s customer base experiences. McHale says the software does this by automating the minimization of error (whether operator, tooling or even programming) and collection of tribal knowledge, ultimately improving tool life and reducing downstream issues that would be only discovered during inspection. He calls TMAC AI a way to “augment highly-skilled labor,” pointing to how the real-time predictions and notifications about process deviations act as faster substitutes for the data gathering that would go on during a lengthy process study. By automating these process studies, he says, skilled machinists can make decisions about process revisions faster.

Datanomix’s TMAC AI marks the boundaries for acceptable tool quality in blue, with a gold line indicating the average tool and a red line indicating the actual performance. Image courtesy of Datanomix.

The companies say that the software can build its initial model after 50 to 100 pieces recorded in TMAC. This can mean 50 to 100 pieces after TMAC and TMAC AI are installed together, or that the software can pull from the last 50 to 100 pieces recorded within TMAC if the AI add-on is installed later.

From the initial model, the software is usually able to judge whether parts are within tolerance with over 95% accuracy. McHale does acknowledge that some of this success can be attributed to the proven production pipelines of the OEM testers, with bad parts tending to exhibit “pretty significant physical noise” or other noticeable characteristics during machining, but says that refining this initial model is simple. When the model makes any errors, engineers can review the data and manually specify whether a part was conforming or non-conforming. The model updates with this data, further enhancing its accuracy.

Rob Caron, founder and president of Caron Engineering, predicts that this sort of software will become increasingly useful as robotics becomes more common across machine shop floors. With machinists spread across more machines, he continues, software powered by machine learning will be able to account for the everyday issues that machinists handle in a process, setting off an alert or finding other ways to automatically compensate and continue smooth production.

Repeat Performance

With the initial batch of data necessary for a model consisting of between 50 and 100 pieces, Datanomix and Caron Engineering recommend TMAC AI primarily to high-volume shops, but they can also see significant use cases for high-mix, low-volume shops’ critical repeat parts. While this software should work for any machine equipped with TMAC, Rob Caron notes that Swiss machining is a particularly good fit for the software.

Datanomix and Caron Engineering displayed the software at IMTS 2024 and are continuing to refine it with select customers through the end of 2024, with general availability in early 2025. Caron says that the software will not necessarily require Datanomix’s main software, as it will be an add-on to TMAC available through either company.