Loading...

Talk

Scroll

Previous Case Study QSL

Using artificial intelligence to automate your quality control

Labplas

Labplas is an SME with 125 employees that specializes in manufacturing sterile sampling products for lab analysis for the food, agricultural, environmental, veterinary and pharmaceutical industries.

With 15 production lines equipped with machine sensors (pressure sensors, thermocouples, proximity sensors, etc.) and environmental sensors (room temperatures, atmospheric pressures, etc.), Labplas generates a staggering amount of data.

Statistics

Sensors per machine:

Cameras, sensors, laser, etc.

38

Surveyed variables:

Humidity, temperature, pressure, etc.

+50

Challenge

We wanted to deploy AWS IoT Core, an Amazon service that transmits data from machines to the cloud in real time, then use this data to train artificial intelligence models.

SOLUTION

First, Vooban connected all the sensors to AWS IoT Core to send data to the cloud in real time as a time series (chronologically). The data was then saved in the Amazon DynamoDB database, generating a production history at the same time. In parallel, we used machine learning models to predict anomalies in the data flow.

Amazon DynamoDB can also generate dashboards in real time to monitor and detect anomalies. For example, if a temperature sensor detects an abnormally high temperature, the user will receive an alert.

The machine learning model was trained using Amazon SageMaker, a service that enables users to build, train and deploy AI models quickly on the cloud. In this particular case, the predictive algorithm identified potentially non-compliant bags based on 50 different criteria that were analyzed in real time.

Interface des tableaux de bord

Labplas’s machines are now able to independently identify the bags with the highest risk of non-compliance, eliminating the need for random product testing by an operator while maintaining the same level of quality.