A.I. Loves Food. Process Optimization of Food Production Lines
Data driven reduction of production cost for profitability and competitiveness
Food production industry is a highly competitive market, hence reducing the production cost is critical for the profitability and competitiveness of the product. Producing food can be a complex process. The product should not only taste and look good but usually other properties such as sensory, kinetic, or physico-chemical properties must be within its specifications.
Cost reductions can be achieved by optimizing the following attributes of the production line while keeping the properties mentioned above within its specifications:
- Production line equipment and its settings
- Environmental conditions
- Way of working / Operator productivity
- Formulations and recipes
In 2018, our customer decided to start optimizing their production line in a data driven way. This required the capability to aggregate the different data sources containing the production attributes listed above for optimization and to analyze which of these attributes are the most important ones for reducing the production cost.
DataStories AI Platform empowered our customer to analyze the production process data to reduce production cost
Effective analysis tools like DataStories AI Platform allow inspecting thousands of historic production variables in a holistic way and finding their optimal settings for reducing the production cost.
Before being able to find key variables and their optimal settings, the data needs to be combined in a single format. This was a challenging task as the data was stored within multiple sources (ERP, SAP, sensor data, etc.) and needed to be combined on production batch level. In total we had over 200 GB of data spanning across 4 years containing at least 350+ sensor signals. To find the most actionable drivers, we extracted more than 7000 potential drivers (features) out of the 350+ sensor signals.
With the holistic data, the DataStories AI Platform empowered the process engineers to create accurate & robust predictive models based on a handful of actionable key drivers selected out of these +7000 features. As the model behavior is explained via powerful ‘What-if analysis’ visualizations, the models provided more understanding of their process and helped the customer to determine the optimal settings of the key drivers of their production line.
The impact we created
Great insights for cost reduction were generated in just 3 months
This project delivered clear ROI on the formulated targets and contributed to future process excellence initiatives in the following ways:
- This project showed that data analytics projects, even on a new production line with a lot of data, can be executed successfully in a short period of 3 months.
- Obtained insights can also be deployed on other lines leading to additional cost reductions.
- New insights gave birth to new innovation projects.