Our customer is a tier-1 automotive equipment supplier, specialized in metal frames manufacturing (machining, stamping, welding). For tier-1 automotive suppliers, the assembly check of the final products before delivery to their customers is a priority. However, when performed by an operator, such quality control tasks come with several drawbacks: these tasks are unrewarding, visually tiring, and highly error-prone.
Furthermore, conventional machine vision solutions are not able to meet the specific needs of our customer:
- Assembly check of different types of components with the same hardware.
- Assembly check of black screws on a dark background. This is a particularly challenging task because there are no visual features which algorithms can rely on.
ADDI-DATA faced several technical challenges to meet these customer requirements:
A specific Deep Learning model has been trained in order to detect the presence of black screws, even on dark backgrounds.
ADDI-DATA manufactured a jig to ensure the correct placement of the part before the operator triggers the inspection. The system then generates the printout of a traceability label.
ADDI-DATA designed a solution that meets the different needs, according to the following process:
- The operator places the part on the jig, and then triggers the inspection
- The part reference is automatically detected, then the presence of the different elements is checked in 5 seconds
- At the end of the control, a visual and audible signal is emitted, a traceability label is printed and the results are sent to the Quality Management Software
The complete vision system for automation of metal frame assembly inspection developed by ADDI-DATA consists of:
- Our appliance optimized for computer vision
- A software licence including
- A customized skill developed by ADDI-DATA
- Data reporting to the Quality Management Software
- Connection to an existing industrial printer for the printout of a traceability label
- 1x high-resolution industrial camera
- A jig to ensure the correct placement of the parts to inspect
- A switch to triggers the inspection process
Our customer is one of the French market leaders in the field of orthopedic textiles. The control of the industrial knit fabrics is essential to ensure the quality of the final product. However, when performed by an operator, such quality control tasks suffer from several drawbacks:
- These tasks are unrewarding and generate a lot of eyestrain
- The reproducibility of the measure is complex to ensure
- It generally constitutes a bottleneck that slows down the production rate (20 to 30 seconds per knit fabric)
ADDI-DATA faced several technical challenges to to meet these customer requirements, especially to ensure the reproducibility of the process regardless the position, the orientation and the folds of the knit fabric. To speed up the quality control process, the solution needs to be integrated into the running process and carry out the measurement of a knit fabric in a few seconds.
A neural network has been trained to automatically determine the measurement points for each knit fabric reference.
ADDI-DATA has developed a method for geodetic distance measurement along the knit fabric surface. This method is based on the use of multiple 3D cameras, to avoid missing data in the points cloud.
ADDI-DATA designed a solution that meets the different needs, according to the following process:
- The operator clips all the knit fabrics to be measured on an overhead conveyor, then triggers the control of the whole batch
- Each knit fabric is automatically carried to the inspection area
- The measurement process of one knit fabric takes up to 5 seconds
- After the dimensional control, the knit fabrics are stored in the corresponding unloading areas (OK or NOK)
- At the end of the batch inspection, a visual and audible signal is emitted, and the results are reported on the dashboard and sent to the Quality Management System
The complete machine vision system for the automation of dimensional measurements of knit fabrics designed by ADDI-DATA consists of:
- Our appliance optimized for computer vision
- A software licence including
- A customized skill developed by ADDI-DATA
- Data reporting to a shopfloor dashboard
- Data reporting to the Quality Management System
- 4x high-resolution 3D cameras
- A dashboard for data reporting
- A conveying system to dispatch OK and NOK knit fabrics
- A switch to control the measuring process of a batch of knit fabrics
The MSX-AI-5000 manages your multi-point visual quality control of industrial components at any stage of your process:
- Check quality and manage the traceability of entering parts
- Perform a visual inspection of assembled or manufactured parts
Key features
- Non-intrusive measurement
- Information reading (QR code, barcode, labels)
- Defect detection and classification
- Missing components check (screws, labels, holes…)
Examples
- Control of components placement on PCBs
- Inspection of liquid transparency
- Detection of defects such as scratches, impacts, dust…
- Surface inspection
The MSX-AI-5000 provides a volumetric quality control system for cubic conveyed items at any stage of your process:
- Measure dimensions of entering parts such as parcels or metal bars
- Perform a dimensional quality control of machined parts
Key features
- Non-intrusive measurement
- No calibration required
- Volumetric measurement accuracy on request
Examples
- Volumetric measurement of metal parts entering a deburring machine
- Volumetric quality control of an embossing process
- Volumetric quality control of a bar turning process
