Digitalization makes production systems transparent: hidden potential is just as easily identified as emerging malfunctions at the earliest stage. Efficiency and plant availability can thus be increased to a degree that seemed unattainable just a few years ago. At the same time, the amount of data collected is constantly increasing.

In modern systems, only a fraction of this data has been specifically evaluated to date. Older systems are usually not digitized and data cannot even be recorded in some cases. However, with the right measures, existing systems can also be retrofitted economically, as the example of electric motors shows. These are regularly checked and maintained in order to prevent unplanned system downtimes - often purely manually, sometimes also via condition monitoring at control level. With u-sense, Weidmüller now offers a retrofit-capable solution for integrating motor monitoring directly into the IIoT in the shortest possible time. In addition, the data from the old systems can be integrated into the development of machine learning models (ML models).

Rotating devices such as motors, pumps and gearboxes can be found in millions of production plants. Depending on the load, they wear out at different rates. If costly failures are to be avoided, the devices must be carefully monitored - which also incurs costs. This applies in particular to manual inspections. It would be more efficient to integrate the data into an analysis tool.

Weidmüller relies on its own machine learning tool AutoML and the retrofit-capable u-sense solution for the ventilation system of its electroplating plant. Smart fan monitoring and data analysis with the Industrial AutoML tool ensure clean air in electroplating processes. Several fans are located on the roof of the electroplating plant for this purpose. Previously, the data and status of the fans were not recorded digitally. For maintenance and inspection, you had to literally climb onto the roof of the electroplating plant. In order to integrate the fans into the digital process, they were retrofitted with the universal u-sense energy drives current sensor and the u-sense vibration sensor. Both devices contain sensors and are suitable for installation close to the machine thanks to their robust housing and industrial-grade connection and communication technology.

Acquisition of all relevant electricity data

The u-sense energy drives are installed in the supply line and record all relevant electrical states of the motor. Industry-standard sensors are used to record the current and voltage curve with an accuracy of 3% at a sampling rate of 1 kHz. Additional sensors can be integrated as required via digital and analog inputs. A control module processes the recorded states and digitizes them so that both directly measured and calculated electrical variables are digitized. This means that not only measured values for voltage and currents of all phases are available, but also, for example, active and reactive power, switching cycles or operating hours. The recorded current values of the fan flow into the Industrial AutoML software as training data.

Vibrations also reveal a lot about the condition of a drive, which is why the u-sense vibration smart sensor was installed. Transmission takes place via Bluetooth Low Energy (BLE) 5.0, with a replaceable AA battery serving as the power supply. As the transmission is via BLE and therefore energy-saving, a service life of up to two years is possible.

The exhaust air blowers for electroplating in particular have many wearing parts, such as the motors or the drive belts that ultimately drive the impeller. The blower draws the air out of the electroplating process via a special filter system, thus ensuring a regular exchange of air. The filter technology used here meets the latest environmental criteria. The exhaust air blower, in turn, is connected to the building with a sleeve, which must guarantee appropriate tightness.

From reactive to predictive maintenance

A closer look at the ventilation systems revealed three key monitoring points: the drive belt, the impeller and the sleeve. Previously, these components were monitored manually or visually. By integrating the data into the Industrial AutoML software, service calls can now be planned in a targeted manner or only take place when required. This is accompanied by a shift from reactive to predictive maintenance.

But how can the data be analyzed? How do you create a machine learning model to analyze the extraction fans? The methods and tools of machine learning (ML) not only enable digital access to the data of the ventilation system, but also make it possible to identify relevant correlations.

Weidmüller wants to enable machine builders and operators to create ML models independently and thus convert the collected data into added value. To this end, the company has simplified the application of machine learning to such an extent that domain experts can use their knowledge of the machine or production process to independently implement ML solutions - without expert knowledge in the field of data science. The software helps to translate and archive the application knowledge into a reliable machine learning application. The expert focuses on their knowledge of machine and process behavior and links this with the ML steps running in the background.

With the Industrial AutoML tool, an ML model could be built and operated in four simple main steps: Data import, data enrichment, automated model creation and model deployment.