At the ARC European Industry Forum 2017 in Sitges (Barcelona), ARC organized a workshop for Industrial Analytics and Edge Computing. In times of increasing amounts of data generated within plants and factories, the need to analyze them properly and give the right information to the right people also increases. Experts from different industries and areas shared their experiences and knowledge to show the challenges of analyzing the large amount of data generated and how companies can implement and benefit from industrial analytics.
May Roca from Dow Chemical stated that before building or implementing a solution for industrial analytics, it is important to bring in all the operators and discuss the needed analytics. In the end, they are the workforce who use the results and react accordingly, proactively plan working tasks, and organize maintenance to improve production processes and guarantee continuation and stability of the production. While this is not a one-time discussion, as needs might change over time, it is critical to continuously keep the communication with the operators alive.
For building a process organization for analytics, necessary steps are to include the operators, data scientists, and process specialists, said Edwin van Dijk from Trendminer. Everyone will have a different view on how to analyze and what to show to whom in the company. Edwin presented two different approaches for industrial analytics. The model-based approach takes the following steps.
- Train data scientist in business domain and set up multidisciplinary team
- Collect expert input to identify data required
- Work with automation / IT to extract the data
- Import data in analytics environment
- Data exploration and filtering
- Choose algorithm / model
- Train model
- Validate model
- Work with IT to deploy solution (model, alert)
He also showcased the self-service approach that drops some of the points of the model-based approach and puts more control over the demanded analytics into the hands of the people using the results.
Peter Guilfoyle from Northwest Analytics stated that the modern manufacturing ecosystem is complex and littered with legacy systems. It is often too expensive to replace and comprises multiple disparate databases, so that not one has ALL the data, and in most cases, real-time analytics is not well leveraged.
Marc Pijpers from Sitech, a company the provides services in maintenance, technological improvements and advanced process controls for sites located in the Chemelot site in the Netherlands, said that they are the number one company in data storage coming from manufacturing. But he also mentioned that they are number one in doing not much with the vast amount of data. Especially in manufacturing only 5% of big data is used in the production line, which not only applies for Sitech’s business, but is a common occurrence in manufacturing. While he claimed to provide better analytics, it remains important to get even more data, customers and end users asked would usually say what they don’t need more data, but to do something valuable with the data already generated. This is where Sitech is leveraging the knowledge and expertise of Trendminer in providing better industrial analytics solutions to their customers.
Fréderic Erben from Grenzebach, a German family owned machine builder, remarke on the topic of edge computing that if a vendor of automation equipment says “put everything in the cloud”, it is hard to realize for remote locations because of slow or missing internet connections, and it is also not accepted for conservative industries. Basic analysis on premise with applying edge computing to provide operational continuity is a key element here.