Manufacturing
Manufacturing
Early warning to identify and manage product defects
Candy-Haier is an important European multinational company that designs and manufactures household appliances. Since the production processes and the quality control of the batches in production are a priority for the company, we have developed an early warning algorithm that supports the company in the identification and early management of possible problems.
- Description and benefits
Identifying and managing defects in production batches is of great complexity. The process between the reporting of individual defects by maintenance technicians in various countries and the collection and analysis of a sufficiently large number of similar reports can take a very long time. It is only at the end of this process that the quality control personnel report any production problems to be investigated. Prior to our work, this time could take up to 24 months.
We have created a text mining tool for the automatic classification of anomalies, through the extraction of textual information from maintenance reports compiled by technicians. This information feeds an early warning algorithm. An app was then implemented to support quality control activities and the definition of strategies for product improvement.
-50% detection time:
Reduction of problem detection time to 4 months, compared to 8 to 24 months with standard software tools
Multilingual interpretation
The text mining algorithm provides automatic classification of fault types from texts in 4 different languages
Speeding up the root cause analysis of problems (from days to a few minutes)
The application allows to identify the main causes of malfunctioning through a 5/10-minute analysis. Previously, analyses took 5-10 days
Deepening case study
The application in detail
The whole application has general menus that allow users to perform in-depth analysis by choosing the production site, the product type, the production batch, the faulty component and the geographical market.
Unlike other standard software packages, our solution is tailored to the internal processes of the company and allows to detect problems only 4 months after the release on the market.
The developed solution lends itself to integration with service intervention analysis, to identify fraud attempts, such as replacement of non-faulty parts or interventions beyond the granted warranty period.
C.F. e P.IVA 07015910966 | Via Schiaffino 11, 20158 Milano