Manufactured products go through a number of process steps and transformations over their lifecycle. Process diagnostics and prognostics capabilities can help ensure high uptime, productivity, safety, and environmental responsibility. CEA-List is developing solutions that leverage know-how from a broad range of disciplines. Instrumentation, simulation, and data science play major roles.
CEA-List is drawing on advances in a wide range of fields to develop innovative diagnostic and predictive tools for industrial processes. Data—and how to process and extract information from it—is at the center of the institute’s research.
Specifically, we want to develop innovative methodologies, be able to measure how reliable they are, and make sure they are compatible with real-time operation.
Our solutions are designed to be integrated into industrial processes, to assist and guide human operators in their decision making and to make the implementation of fully automated processes safer and more reliable.
CEA-List leverages analysis software and extensive knowledge of algorithmics and data science, using inverse methods, heuristics, and AI, for example, to partially or fully automate diagnostics. Simulation also plays a major role in machine learning and prediction.
Depending on the industry and use case, we develop automation tools that are robust under variable operating conditions, as well as decision-assistance tools. All of our tools are designed to improve the reliability and productivity of industrial processes.
CEA-List helped develop an automated structural health monitoring tool for offshore wind turbines and, specifically, welds.
Tools in the CIVA Analysis software suite and a purpose-built module that separates information generated by the structure itself from information generated by defects were used to implement an algorithm using segmentation, characterization, and size measurement data to automatically detect defects.
This fully automated analysis brought turnaround times down from several months to just days.
Prognostics requires a series of measurements and observations of data flows and trends over time.
The predictive technologies developed at CEA-List depend heavily on the interoperability of data and models deployed across a digital measurement chain. These technologies can predict the future condition of components based on aging and degradation models of the system being monitored. Prognostics is particularly useful for developing predictive maintenance strategies and determining the lifespan of a structure.
Using shared software platforms, our researchers leverage both AI and simulation to create solutions to inspection and process control challenges. They are developing analytical and numerical models for real-time embedded applications, for example.
Today, CEA-List possesses a portfolio of powerful data analysis and diagnostic software tools that includes simulation, inverse methods based on direct models, algorithms and AI-specific tools, heuristics, and knowledge formalization. The institute’s expertise in a wide range of technologies and approaches enables a system-level vision—the most effective way to develop an appropriate solution and, ultimately, produce reliable diagnoses in real-world operating conditions.
In the field of diagnostics, the stakes are often high, especially in terms of safety. So, our solutions have to be reliable, even when we don’t have enough data. The reason we use a variety of techniques in combination with AI is to guarantee the ability of our solutions to process signals that are often complex in variable operating conditions.