Optimise efficiency with predictive analytics

Uncover savings potential and optimise efficiency with predictive analytics

Transparency is the elementary component of an efficient manufacturing environment. It illuminates past and current processes, reveals potentials and thus helps industrial companies to optimise their value creation. However, with the wave of digitalisation sweeping through manufacturing, the requirements for efficient processing of the almost unlimited amounts of data obtained from it are also increasing – and ideally in real time. These data volumes in combination with new insights around the topics of ‘artificial intelligence’ and ‘machine learning’ now also allow the much-cited view into the crystal ball with corresponding reliable predictions.

Deloitte: With data analytics to new business activities

A functionality that contemporary MES solutions have long offered takes on a whole new meaning in the course of the above-mentioned wave of digitalisation and the establishment of the IIoT: predictive maintenance. The consulting firm Deloitte refers to this as ‘data analytics’: the IIoT provides a very precise picture of the current state in terms of data technology. Algorithms, Big Data applications and AI recognise certain patterns in this information. This leads to predictions about expected conditions and future trends. According to Deloitte, data analytics “uses hypothesis-driven models to generate important insights. Business activities thus generate the data that, through analysis, can again lead to actionable insights, new decisions and perhaps even new business activities”.

Predictive Analytics not a new topic: approach crucial

The university professor involved in the research project, Dr. Alfred Taudes, Vienna University of Economics and Business Administration, Department of Information Processing and Process Management, Institute of Production Management, knows the strengths of predictive analytics only too well: “With predictive analytics, manufacturing companies can use the data generated by sensors for better planning. A more accurate forecast of the reject rate, for example, leads to improved capacity utilisation, adherence to schedules and lower stock levels.” Taudes describes how in an MES like cronetwork the existing data can be used meaningfully in the context of predictive analytics: “The records made in the past in the MES on scraps, machine failure, malfunctions and product quality in the respective environment (machine, personnel, environment, material, order and time) provide information about constellations in which these problems occur more frequently by using suitable methods. These patterns are applied in predicting quality metrics in future planning.”

Of course, Industrie Informatik co-founder and Head of Strategic Product Management, Thomas Krainz, knows that predictive analytics is not a new topic. For him, however, the approach is decisive: “Our goal was to develop an out-of-the-box solution with which our users can achieve results quickly, easily and, of course, affordably.

Efficiency advantages through rapid implementation of predictive analytics methods

The path to a marketable product led through a research project lasting several years, in which several bodies were significantly involved. The theoretical, scientific approach was provided by the Vienna University of Economics and Business Administration. RISC Software GmbH, an established and internationally recognised research company, was also involved. The jointly gained knowledge was then developed in case studies of selected customers. The focus was on the development of possible fields of application on the basis of existing data, as well as the elaboration, development, testing and evaluation of algorithms, procedures and technologies for forecasting purposes.

“The result of the research project is, on the one hand, an out-of-the-box technology stack that can be used both in the cloud and on-premises. In addition, we have developed a data preprocessing model that helps the user to cleanse and prepare data from cronetwork MES in the first step before an AI carries out the ongoing evaluation and interpretation of the information obtained from it,” says Thomas Krainz. The fact that all these processes are based on the standard data model of cronetwork MES creates great efficiency advantages for the user, which are reflected in the rapid implementation of predictive analytics methods. Added to this is the use of the Random Forest Model as a learning algorithm, which is known for fast turnaround times and very good interpretability of the results. “Most of the data is already available, all that is missing is a suitable analysis and user-friendly integration into the planning process,” says university professor Taudes. In addition to process data, textual or visual information could also be relevant in quality management. “Here we are only at the beginning of the analysis, especially the integration of heterogeneous data sets is an active research area.”

Predictions of scraps and workplace disruptions

Thomas Krainz adds, “In the end, the decisive factor for success is the adaptation of all these technologies and functions to the respective data situation and, above all, the expectations of the customers. Artificial intelligence and predictive analytics are not miracle cures. They are neither better nor more intelligent in their tasks than a human. Their advantage lies in replicating human know-how – at high speed and also around the clock. Many possibilities derive from this.” In concrete terms, this means forecasts on relative scraps and workplace disruptions in follow-up stories, as well as on quality statuses after production steps. According to Krainz, this information alone can reveal hidden savings potential and massively optimise efficiency on the shop floor.