As usual, CEO Hans Beckhoff begins the press conference at the Hannover Messe 2019 with the announcement of the latest business figures. With a realized turnover of 916 million euros, the specialist for IPC-based automation grew by 13% last year compared to the previous year and thus only just missed its own target of annual growth of 15%. When asked about business development in the current year, Beckhoff replies: "We are assuming a stabilization at a high level." The billion euros in turnover that has long been targeted for 2020 is therefore well within the realms of possibility.

In line with the motto of this year's Hannover Messe - 'Integrated Industry - Industrial Intelligence' - the company from Verler focused on the topic of artificial intelligence in Hannover and demonstrated the concrete advantages of this technology at its stand in Hall 9 using exemplary application scenarios based on the new 'Twincat Machine Learning'. This solution for ML, which is seamlessly integrated into the control technology, provides machine manufacturers with the optimal basis for increasing machine performance - for example through predictive maintenance, self-optimization of process sequences or autonomous detection of process anomalies.

The basic idea of machine learning is to no longer develop solutions for certain tasks using traditional engineering and transfer them into an algorithm. Instead, the desired algorithm should be learned using exemplary process data. In this way, powerful models can be trained. For automation technology, this should open up new possibilities and optimization potential in areas such as predictive maintenance and process control, anomaly detection, collaborative robots, automated quality control and machine optimization.

The respective model is trained within one of the common ML frameworks, such as Matlab or TensorFlow, and then imported into the Twincat runtime via the standardized exchange format ONNX (Open Neural Network Exchange) to describe trained models. This offers the following new functions:

• Machine Learning Inference Engine: for classic ML algorithms such as Support Vector Machine and Principal Component Analysis
• Neural Network Inference Engine: for deep learning and neural networks such as multilayer perceptrons and convolutional neural networks

Inference, i.e. the execution of a trained ML model, is possible directly in real time as a Twincat TcCOM object, and with small networks with a system response time of less than 100 µs (Twincat cycle time 50 µs), promises Beckhoff. The models can be called up via the PLC, C/C++-TcCOM interfaces as well as via a cyclic task.

Drive technology without control cabinet

In addition to the topic of AI, Beckhoff also brought other innovations to the Hannover Messe, including in the field of drive technology - such as the AMP8620 supply module with IP65 protection. While the connection of the AMP8000 servo drives from Beckhoff to the control cabinet with the coupling modules was previously reduced to just one cable, this is completely eliminated when using the new supply module. The elimination of the control cabinet further reduces the footprint and cabling requirements for the entire machine. In addition, there is no need for the otherwise required control cabinet air conditioning.

This module is connected directly to the supply network. It contains all the necessary circuit components such as mains filter, rectifier and charging circuit for the integrated DC link capacitors. Additional distribution modules or decentralized servo drives can be connected as required. There are also two Ethercat P outputs, which can either be used to supply Ethercat P modules or to contact additional supply modules required for system expansion. Integrated DC link capacitors store the regenerative energy of the entire system and then make it available again for acceleration processes.