Smart Manufacturing Analysis

The main objective in Smart Manufacturing Analysis is to support the manufacturing industry to make use of IoT by optimizing their production and identifying problems early at their occurrence. Therefore, we contribute among others a self-analytics solution that enables operators and other non-data scientists to optimize production settings and identify problems and their causes early, and identify patterns that can warn operators in the future for repetitions of certain production faults.

The core idea was to integrate TrendMiner as an analysis solution for plant workers to analyze the machine and production data. For this purpose, a connection to the machines had to be developed beforehand using a specially developed pre-processor, which retrieves, parses, checks, corrects, standardizes, and normalizes the data and finally saves it in a database. TrendMiner also had to be given access to this database.

The self-analytics tool intends to empower process engineers, plant managers, and other non-data scientists to optimize operations all the time, anytime, using a new generation of model-free self-service industrial analytics, as a means to analyze operational performance without data modeling, as well as to recognize cause-effect-relationships and root causes, and create early warnings for. The self-analytics tool based on TrendMiner addresses the following purposes:(1) Visualizing for operational storytelling and fast decision making, to see which parts of the process require extra attention or analysis from energy/sustainability indicators point of view and to start investigating consequently process anomalies, production losses or equipment inefficiencies. (2) Analysing data to find root causes fast, to know the root cause of deviating behaviors to help avoid them in the future. By indexing all tags, data becomes instantly available to find and overlay behaviors of interest quickly, easily assess problem impact, and validate or invalidate hypotheses in a matter of minutes, as well as to create a rich knowledge base, and leverage results from analyses of historically similar events. (3) Predicting what’s likely to happen, with the target of comparing saved historical patterns with live process data to forecast unwanted events, or to see how far the process has evolved and predict how it will likely continue.