TwinCAT 3 Machine Learning Creator (MLC) from Beckhoff greatly simplifies the process of configuring, training, and deploying AI models into real-time automation systems using the standard workflow in TwinCAT 3 automation software. This allows users to handle the entire engineering process, from data collection to the trained model themselves – without any AI expertise of their own. The finished AI model is optimally adapted to real-time requirements in the control environment in terms of latency and accuracy.

A prime application for the versatile TwinCAT 3 MLC (TE3850) is AI-supported image processing for quality assurance. This is delivered as a no-code, web-based platform as a service (PaaS) where users can easily collaborate and manage projects without having to worry about file versioning or synchronization. It not only leverages open AI standards, interfaces, and best practices, but also provides the trained models in the ONNX open standard format. These latency-optimized AI models for control applications are specially adapted to run on Beckhoff IPCs and with TwinCAT software products, although they can also be used as ONNX models beyond the Beckhoff ecosystem.

TwinCAT 3 MLC helps automate the AI model training process, making the power of artificial intelligence for automation available to all – including smaller companies. This provides a competitive edge and much-needed solutions for the worsening shortage of skilled automation engineers. And for seasoned AI experts, this solution streamlines their workload considerably while minimizing the potential for errors.

The ability to speed up project development processes offers yet another clear advantage, particularly as the development tool provides extensive and transparent methods for displaying the behavior of the models created and comparing them with each other. Users can also benefit from automated report generation, which supports auditing processes for AI model creation. Another crucial aspect is that the required application-specific data remains protected, since it does not leave the company for a large language model (LLM).

Initially, the focus of TwinCAT 3 MLC was on image processing, but it will be expanded to include the analysis of time-based process signals. In addition to the TwinCAT 3 MLC Computer Vision extension package (TE3851), TwinCAT 3 MLC (TE3850) also offers the TwinCAT 3 MLC Signals and Time Series module (TE3852). This extends the range of functions to include time-based process signal analysis. This is often crucial for industrial applications, as current, temperature, and vibration curves provide valuable information about the state of processes, components, and tools. The models created with TwinCAT 3 MLC Signals and Time Series detect patterns and deviations in real time based on the relevant data, enabling predictive maintenance, process optimization, and anomaly detection directly in a standard controls environment.

TwinCAT 3 Machine Learning Creator is a purely web-based application. As the engineering takes place entirely in a browser, no local computing power is required. The process of creating AI models is therefore very simple, accessible, and convenient for all users.

Beckhoff Automation