Production Data Analytics - to identify productivity potentials

Manufacturing industries are always under constant pressure to improve the productivity. Many manufacturing companies started to capture the shop floor data to the time scale of seconds. Consequently, the challenge is to harness the value from the data and identify the ways in which the value extracted could improve the productivity.
In this thesis, two sets of Manufacturing Execution Data (MES) data consisting of shop floor data were used to identify the productivity potentials. The first set of data had the MES information was derived from a common data source of 23 industries consisting of 884 machines. The second data set was more specific to one manufacturing line. The methodology to analyse both data sets includes data cleaning, data preparation and data modelling.
The outcome of the analysis of the first data set was the impact of operator influenced loss times on Overall Equipment Efficiency (OEE). The outcomes of the analysis of the second data set were identification of static and momentary bottlenecks in the production line from the real time data and to develop algorithms for those. Also, the Key Performance Indicators (KPI) were modelled to determine the pattern and to predict their behaviour.
Identifying the productivity potentials (operator influenced loss times, bottlenecks detection and predicting the behaviour of the KPI) from the real time data is very useful to make fact based decisions which reduces the value at risk of making these decisions which in turn helps to improve productivity.