Big data and machine learning for prediction of corrosion
Corrosion is a disease to materials and poses a significant risk to many industrial facilities and structures such as pipelines, storage tanks, boilers, heat exchangers, and other equipment and systems. Overall, corrosion is one of the critical materials failure mechanism in industry and annually responsible for costs amounting to trillions of euros. In the EIT RawMaterials up-scaling project CorTools, corrosion on-line monitoring and prediction software tools are developed to respond to the needs of raw materials industry.
We need to understand the behaviour of construction materials of tanks, containers, pipes, mixers, i.e. stainless steels in the operation conditions. In this project, we focus on raw materials processing conditions and are creating a materials selection tool, a virtual expert and corrosion on-line monitoring solution.
Elina Huttunen-Saarivirta, Research Professors, VTT Technical Research Centre of Finland Ltd
The unique Cortools consortium: Boliden Harjavalta Oy, DMS, Ferritico, Outokumpu Stainless AB, Metso:Outotec, Tecnalia, ZAG and VTT Technical Research Centre of Finland Ltd, enables the co-creation and industrial validation of the tools. For the industrial project partners, the developed tools enable more than 1 M€ annual savings through the prevention of corrosion failures. For the European raw materials industry in whole, the project may allow billions of euros in savings.
The software tool will make it easier to support our technical marketing and sales. The tool will decrease the need for time-consuming corrosion testing and it will be easier for the industry to choose the most optimum stainless steel grade as construction material.
Lena Wegrelius, Head of Corrosion, Outokumpu Stainless AB
Collaboration across the whole industrial value chain is carried out in the project to cover multiple dimensions (laboratory, pilot, operational environment) in the tool prediction capabilities and thus facilitate market penetration of the software tool. The use of digital on-site measuring technologies and computer modelling will connect various operational scales, therefore contributing to an improvement in material- and energy efficiency of the raw materials processing industries.
Machine learning creates ”virtual expert”
The corrosion on-line monitoring and prediction software tools provide the raw materials sector yet another footstep towards digitalisation. Another giant progress is the link to artificial intelligence (AI). In this project, AI is used in computational modelling. Furthermore, a remote collection of on-line monitoring data allows later for the hidden causalities in the project conditions (e.g. mineral quality in hydrometallurgy) and the detected corrosion rate to be disclosed with the aid of AI. The software tool is developed by the project partner Ferritico.
We make use of the available data when creating this virtual expert. Industries have huge sets of experimentally measured corrosion data and this is a big source of information and motivation to develop a machine learning tool. We are predicting failure events based on all of the available features instead of developing complex physical or chemical models of corrosion.
Satish Kolli, Materials Informatics Engineer, Ferritico
The user needs are twofold. First, a materials selection tool, a “virtual expert”, that allows for evaluation of a corrosion risk of materials under various operation conditions emerged. Based on machine learning, this “virtual expert” finds the safe operating conditions for a given material and estimates the lifetime of materials in specific conditions. Second, on-line corrosion monitoring systems in industrial facilities reveal the actual corrosion rate of the materials and makes it possible to evaluate the remaining lifetime of the components.
