Predictive analytics has become essential for companies in all sectors, including the industrial sector, to leverage the potential of their data and to stay competitive. A frequent use case is predictive maintenance, i.e. predicting and preventing machine failures. This avoids unplanned production stops and maintenance, lowers costs, and improves planning and reliability. Another important use case is quality assurance. Production issues can be detected very early on, i.e. to optimize product design, or production processes can be optimized. Risk management is important for industry companies to identify potential risks that might impact future productions. Future developments can be predicted accurately and costs can be stabilized. Especially for industry firms, productions losses, i.e. caused by interruptions in the supply chain, lead to great issues and costs, because compliance can’t be fulfilled or production isn’t at full capacity.
The “Funded R&D” team of RapidMiner mainly works on research projects and innovative predictive analytics use cases, either funded by the European Union (e.g. within the “Seventh Framework Programme” and “Horizon 2020”) or the German government. Also the US government recognized the importance of predictive analytics and supports research projects that deal with predictive analytics in the industry. The advantage of research projects is that business and research come together and find solutions for relevant industry challenges using innovative technology and algorithms that are applied on real world scenarios at our industrial partners.
RapidMiner started as a research project at Technical University of Dortmund in 2001 and evolved to a widely used open-source code-free predictive analytics platform. Due to our academic roots, RapidMiner still keeps a strong interest in research and performs numerous joint R&D projects with universities. To gain experience in current topics and solve real issues of industry partners, we founded the “Funded R&D” team, which focusses on research projects with innovative predictive analytics use cases.
Within the ProMondi project we created a solution for predicting assembly plans and costs for new product designs using Data and Text Mining and thereby support product designers and assembly planners, enabling better and faster decision making in product design. This solution was successfully applied to product designs and assembly plans of truck engines, washing machines, and dish washers, making the partner companies more agile and reducing their costs. In the FEE project we work together with partners from academia, the chemical industry and industrial automation to early detect critical situations and to support plant operators in their decision making. We’ll create a solution integrated into existing software that helps plant operators and plant managers to detect and react proactively in critical situations. Therefore we use Big Data technology to analyze time series data from sensor readings and unstructured data from log entries (created automatically as well as manually). The change from reactive to proactive handling of critical situations saves costs that were caused by production losses.
With this presentation we share our experience and best practices with predictive analytics use cases in the manufacturing industry. Which use cases occur frequently? How can typical use cases in the industry best be solved? What role plays predictive analytics in those solutions and how can the solutions be integrated in your daily business?
David Arnu studied at University of Dortmund and holds a Master of Science degree in Computer Science. There he worked as a research assistant at the chair of Artificial Intelligence, later he joined the R&D team of RapidMiner as Software Engineer and Project Engineer. Now, he is a Data Scientist at RapidMiner. He is working on research projects focussing on Predictive Analytics and Big Data.