Hempel
Implementation of AI models and a data platform for predictive maintenance
Delivery
Predictive Maintenance, Platform Architecture
Azure, back-end applikation, Machine Learning, MLOps
countries
Platform & AI architecture
seconds response time
THE COLLABORATION
Hempel
Hempel is a global specialist in paint and surface protection with more than 100 years of experience. The company’s solutions protect everything from ships and industrial infrastructure to buildings around the world. In the maritime sector in particular, the condition of surfaces plays a decisive role in fuel consumption, maintenance and operating economics. With more than 80 countries represented in its business, Hempel continuously works to develop new technologies and service solutions that can extend the lifetime of customers’ assets and make their operations more efficient.
Predictive Maintenance based on data-insights.
A ship’s surfaces are constantly exposed to changing conditions. Temperature, weather, load, routes and sailing patterns affect their condition and thus the need for maintenance. But actual wear is difficult to assess, and decisions are often made based on fixed service intervals and experience rather than precise insight. This can lead to unnecessary costs, premature maintenance or problems that are detected too late.
Foundation for predictable maintenance.
Codellent developed an advanced platform architecture in Azure that combines image analysis with measurement data from paint samples. Machine Learning models identify paint types, defects and wear, which provide input to a neural network where the models’ output is compared with sample data. Through a central backend and automated data integrations, the analyses were made available via a mobile app, so service technicians could quickly access relevant information directly in the field.
Reactive planning to data-driven decision-making.
Codellent succeeded in combining advanced image analysis and historical sample data in one solution that could deliver advanced results in under 10 seconds. The project thereby demonstrated how service technicians around the world can access specialised knowledge and AI-based assessments directly from a mobile device. In this way, the solution showed the potential for faster and better-informed inspections worldwide.
PERSPECTIVE
Hempel, Trust is earned
STARTING POINT
The Ambition
Hempel wanted to give its customers a more precise basis for planning the maintenance of their ships. With better insight into the actual condition of the surfaces, the company would be able to advise on when maintenance is necessary, and thereby help customers avoid unnecessary costs and operational disruptions. At the same time, the project was to explore the potential for new, data-driven service solutions that could strengthen Hempel’s business in the maritime sector.
The task
Developing a model that can predict maintenance needs required more than training an algorithm. Codellent had to create a reliable data foundation, translate knowledge of maritime reality into usable data models and develop an architecture where the AI models could be updated and further developed over time. At the same time, the solution had to fit into Hempel’s existing technology landscape, so the models would not just work in isolation but could form the basis for new applications.
IN DEPTH
2024
Case, Hempel
Advanced analytics and AI combined in one technical platform
An image of a painted surface can reveal a lot about a ship’s condition. But the real value emerges when the image can be analysed together with historical measurements, laboratory data and information about the material’s properties. That was the technical challenge Codellent had to solve for Hempel: making complex knowledge about paint and surfaces available to service technicians in a matter of seconds, regardless of their geographical location.
Codellent developed a unified platform architecture in Azure with Kubernetes, Docker and a central backend application built with REST APIs in Python and C#. Azure Event Hubs and Azure Functions handled integrations and data exchange, while Cosmos DB, Blob Storage and SQL Server supported different data types. The architecture made it possible to combine images, measurements and GPS data from mobile devices and process the information across systems.
The core of the solution was the interplay between several AI models and the backend application. A Machine Learning model for image classification was developed to identify paint types, defects and levels of wear. The model’s results were then combined with analysis data from paint samples in a neural network, which could produce structured assessments across the data sources.
An accompanying mobile app allowed technicians to take a photo and access the analysis results. Behind the simple user experience was an advanced infrastructure, where Codellent had combined data integration, image analysis and AI in one solution that could deliver information in under 10 seconds – even when the user was on the other side of the world.
THE COLLABORATION
Hempel
OVERVIEW
AI-based prediction of maintenance needs on ship surfaces.
Codellent
Distribute maintenance insights worldwide
Delivery
Predictive Maintenance, Platform Architecture
Industry
Painting, Production and Maintenance
Area
Distribute maintenance insights worldwide
Effect
Real-time AI-models integrated in an App
(03)
PROJECTS


