We worked with a major Norwegian chocolate manufacturer to resolve a process issue causing production and quality problems
Chocolate production consists of several process steps where both raw material quality and storage conditions influence the final product quality. Although chocolate making today is based largely on science such as chemistry and food technology, the ‘human touch’ and flair based on many years of domain-specific knowledge remains an essential ingredient in the process.
Nidar, a major Norwegian chocolate producer, was experiencing a quality issue in one of their production lines, forcing them to regularly scrap batches. From a business perspective, this resulted in significant waste, downtime, energy usage and re-work costs.
When the quality issue arose, the team at Nidar realized that to fully understand the complex variables at play, multivariate process control methods and Design of Experiments (DoE) strategies were required, and consequently implemented the following 3-step process:
- Analyzed historical data with multivariate models
A number of batches, both with and without quality problems, were selected and the process data from each was analyzed with multivariate regression methods and Principal Component Analysis, giving deeper insights into possible causes of the quality issue.
- Applied Experimental Design in full scale production
Secondly, the client realized that designed experiments were necessary to fully understand the issues and isolate specific problems. This began by investigating the various steps in the process with fractional factorial designs to pinpoint the important variables. This required them to stretch the process settings in different directions, occasionally allowing some batches to be scrapped in order to find the operational envelope where the process was stable even when subjected to changes. The next phase was to implement experimental plans for important variables across the entire production chain.
- Implemented changes in the process settings
Based on the conclusions from the above steps, the team at Nidar implemented changes in the process settings, allowing them to bring production back in line and consistently produce a high quality product which was robust towards changes in the raw material and other factors which may vary without the ability to control them.
This analysis revealed that the process of making the chocolate could not be viewed as an isolated event. For example, the speed at which the process was running was important for the production volume, thus this variable could be regarded both as a critical process variable as well as a response variable in terms of efficiency. Furthermore, the process settings when filling and cooling the product had interactions with the storage conditions such as temperature, time and humidity.
Multivariate data analysis used in combination with Design of Experiments enabled the product quality department at Nidar to get a better understanding and view of the whole process which was used to resolve a difficult quality problem.
From a business perspective, this enabled the company to save $1M per year on one production line alone and transfer the knowledge gained to other production lines.