By Dr Javier Dominguez-Caballero, technical fellow in AI and digital manufacturing for the AMRC’s machining group.

Artificial Intelligence (AI) and data science have become fundamental digital building blocks for driving efficiency and productivity in subtractive manufacturing. 

At the AMRC, we leverage AI as a core framework for data analysis and optimisation, enabling high-fidelity predictions across the machining lifecycle, from cycle times and tool wear to the residual stresses of finished components. 

By integrating machine learning (ML) into our workflows, we enable reducing programming times from hours to minutes, mitigating scrap through real-time anomaly detection and optimising the end-to-end machining process. 

Beyond generative AI: the role of data science in machining 

While generative AI such as ChatGPT and Gemini dominate public perception, industrial AI in a machining context is a distinct discipline. While we are exploring generative applications, our primary focus remains on established, traditional ML models and data science.

The relevance of AI in machining stems from the sheer complexity of the machining environment, where variables like extreme thermal gradients and mechanical stresses occur simultaneously. While many of these factors can be modelled via classical physics, the inherent variability of the machining process often introduces randomness that predictable models fail to capture. 

We believe in a hybrid approach, combining the use data science, AI and modelling;

• Physics-based models: handles the implementation of known mechanical laws. 
• AI models: act as a corrective layer, capturing real-world differences, such as machine and tool degradation, coolant fluctuations or material inconsistencies that physical models might overlook. 

The use of AI in subtractive manufacturing

There is still a lot for us to learn about AI which can be perceived as a black box. 

We have extensive knowledge regarding the physical aspect of machining; but our aim is to not replace this expertise with AI, but to augment it. 

By fusing together physics with machine learning, we can create explainable machine models that carry higher credibility and trust. We are currently in a pivotal transition period. Historically, machining has lagged in data availability compared to the vast datasets used to train large language models. To help bridge this gap, the AMRC is leading extensive work in connectivity and digitalisation, standardising data across our diverse fleet of machines to ensure we are ‘AI-ready’. 

From research to production: current and future applications 

Our current work around AI in the digital machining team within the machining group is heavily focused on tool wear prediction, a critical factor in part quality and process stability. 

By pairing AI models with digital twins, we have developed a framework that provides high-accuracy forecasting in production environments.

Our projects include:

• Factory+  ACS integration: we are deploying AI models into cloud environments via our Factory+ platform to demonstrate how a smart factory ecosystem functions in real time.
• Boundary-pushing research: through our links with the University of Sheffield and the Centre for Doctoral Training (CDT), we are mentoring PhD students on low-TRL (Technology Readiness Level) projects that will define the next generation of machining intelligence.
• Expert advice: we guide partners through the ‘plug-and-play’ myth. AI implementation is complex, requiring significant data trials and validation. We provide the advice and direction to ensure our partners select the right path, whether that is a pure AI route or a hybrid physical-digital approach.

The AMRC is uniquely positioned to lead this evolution. With our vast infrastructure and niche technical expertise, we are moving beyond the groundwork toward a future where AI-driven machining is the industry standard.

Interested in optimising your subtractive manufacturing processes through AI, speak to us today.