.png?width=250&name=Inveniam%20blue%20(1).png "Inveniam blue")
Public Initiative Networks (PIN) are key players in the French National Broadband Scheme. They provide homes and businesses in less densely populated areas with access to broadband services from Internet Service Providers.
To deploy fiber optic networks, Internet Service Providers operate on different types of infrastructure in the Public Initiative Networks in order to provide end users with broadband:
- Termination closets
- Optical connection points
- Optical branch points
The problem to be dealt with:
FTTH cabling operations follow a certain number of rules. The quality control of the equipment is done by sending a report containing a few pictures taken of the infrastructure. The reports are sent from the Internet Service Providers to the Building Operators in order to ensure the equipment is in a good state after the operation.
By the end of 2021, around 10 million cabling operations have been done on the Public Initiative Networks resulting in more than 70 million images
to be analyzed. How to control quality while dealing with large amounts of data? How to have a reliable solution with a 30% growth rate?
Problems of building operators
- Process an exponential volume of data
- Longer time allocated to quality check
- Corrective incidents in case of detection of anomalies
- Equipment in a bad state after operations
- Bad network quality
- Impact on relationships between Internet Service Providers and Building Operators
Targets of the project
Using the help of INVENIAM, the targets of Axione were:
- Ensure quality of the maintenance within its infrastructures
- Automate quality check
- Detect easily the failures and malfunctions within the equipment
- Optimize the processing of rework operations
- Ensure the good state of the infrastructure
What is Eygles?
Eygles is a SaaS based on Computer Vision and Machine Learning. It automates quality control of the FTTH cabling operations in Public Initiative Networks (PIN).
Eygles is based on a set of predictive and self-learning algorithms that can ingest and analyze an infinite amount of data 24/7.
By combining the knowledge of the field operators and the detection capacity of the machine, our solution allows identifying all types of failures in telecom infrastructures and therefore ensure quality control of the cabling operations.
Ressources allocated to the project
Human resources:
- Product Owner [1]
- Data-scientists [3]
- Tagging officers [5]
Technological resources:
- Analysis engine (processors GPU/CPI)
- FTP server to download images
- Web interface to visualize results
- Support for indicators and results of the analysis
Results of the cooperation
The cooperation with Inveniam has allowed Axione to reduce by 46% the number of incident failing tickets received.
Since Eygles has been implemented:
- +10 use cases have been developed (use case = a
type of failure) - 95% accuracy level on all the models
- Less disruption on the network
- Improvement of the connectivity of the end-users
Do you have issues related to quality control in your telecom infrastructures?
Do you have questions about the deployment of artificial intelligence in your organization?
