By: Dominic D. Nwagbaraocha
The risk of having a recall in trade or a consumer complaint which greatly affects our market share, should be avoided/eliminated. Think of a production line where before a defective product (no code, low weight) comes out of the line, you get an alarm or product is taking of the line.
Building a predictive quality system;
- Quality team skill upgrade: Team needs to take some machine learning or artificial intelligence courses.
- The use of digital system and the readiness to invest (this should be driven by cost benefit ratio from cost of quality point of view).
Based on historical defects and all its related process and production data, a defect prediction model can be developed in order to predict future quality defects.
Creating a predictive model for quality defects
- Data Gathering: The process starts with historical data acquisition.
- Data preparation: Format and align the raw data into a unified model, remove some outliers.
- Machine learning: Training of one or more prediction models.
- Models’ evaluation/selection: Prediction models are evaluated and compared by using KPIs.
- Defect Prediction: Best model is then used for the online prediction system, which feeds it with actual production data to predict defects.
About the Author
Dominic is a process & quality consultant with over 10 years’ experience leading various continuous improvement in world-class organizations. founder of Doruem Process Services and Co-Founder to Nodal Point Engineering, with core competence in manufacturing excellence and digital manufacturing.
You can reach him on LinkedIn here.