Data is the foundation of digitalization, but it is becoming increasingly complex to handle and evaluate due to the growing networks and increased security requirements. Data-driven analysis of production systems enables scalable and flexible support for production management and maintenance work. For example, to detect anomalies in the factory in good time, the MLnext solutions make it possible for you to evaluate your data through machine learning and to make optimizations quickly, easily, and precisely.
Production data at a glance at all times
Nowadays, huge amounts of data can be found in the machines and systems of a production facility. This provides the basis for the implementation of all digitalization solutions. When it comes to evaluation, the data provides insights into the consumption of various resources such as (compressed) air or water, for example. The data analysis based on machine learning offers a variety of advantages compared to manual data analysis. In addition to automated data evaluation, users of this approach can benefit from scalability for entire production processes. In addition to automated and therefore much faster data evaluation, users benefit from the scalability of individual machines, complete systems, and even entire production processes.
With the MLnext smart solutions, Phoenix Contact offers an easy way to create and deploy machine learning models for production. For example, this means that neural networks can be used to automatically detect problems in production processes and link to recommended actions. In Phoenix Contact’s PLCnext Factory, the use of MLnext-based solutions has already boosted productivity by 10% in a very short time. It has also been possible to realize shorter implementation cycles for new solutions, resulting in a faster return on investment (ROI).
With MLnext, you benefit from various advantages:
The model learns which optimization measures can be used based on past consumption activities. Current changes, such as a drop in ambient temperature in a system, are automatically taken into account and included in the evaluation.
In practice, there are no limits to the use of MLnext solutions. Possible areas of application could include predictive maintenance or process optimization of production plants, for example.
With the MLnext solutions, Phoenix Contact’s electronics production in the PLCnext Factory was the first to implement the machine-based use of artificial intelligence across the board. Since then, it has been possible to record the states of the components within the machine more easily. Based on this result, further data analyses for condition-based maintenance (predictive maintenance) will be carried out in the future.
MLnext Creation enables the creation and parameterization of neural networks. Since the application works on the basis of a configuration file, no programming knowledge is required. Through automatic logging, all processing steps are recorded individually and generate standardized reporting. To choose the ideal neural network for the application, several models can also be quickly compared. The configuration file can then be customized based on the reports.
MLnext Execution is a platform-independent software solution for neural network execution. The solution can be used flexibly – whether at the control level, on a local IT server, or in the cloud. Developed solutions can be tested, improved, and compared in the preferred environment. Here, the necessary data flow is created by the configuration files.
By default, the configuration includes data loading, preprocessing, execution of the machine learning model, postprocessing, and storage of the new information. Requests can either be repeated cyclically or set on demand of a REST interface.
In addition, new solutions can be developed easily on the same platform without interrupting the processes already integrated. Changes can be made flexibly by the user, service partner, or the experts from Phoenix Contact.
The ability to extend data flows during runtime provides users with a high degree of flexibility.
High-level language programmers can rely on the MLnext Framework for a quick and easy overview of the data. The Python library contains a variety of functions for the preprocessing and visualization of production data. Different models can also be compared with each other using evaluation functions.
The programming library is constantly being developed so that it always implements the latest approaches from research. This means that developers always have the latest state-of-the-art options for analyzing time series data from production in order to implement their ideas.
Whether anomaly detection, locating leaks, or optimizing processes, MLnext provides a simple and targeted way of identifying hidden potential. Find out more in our Infopaper.
