SAM – Self Adapting Model-based system for Process Autonomy

This project presents an innovative and new digitization system for self-adapting models, with potential to improve the competitiveness of Norwegian process industries. Value is created for the end-users by improving process monitoring, process understanding and process control and adding autonomous models. Significant benefits are expected from the innovations in the project. Bilfinger's BCAP service analyses historical data from production processes to identify key improvement parameters at a given plant. The addition of new modules to Bilfinger's BCAP product will make them a preferred digital platform provider and opens possibilities for significant growth in revenue and profit. Making SAM widely available can give massive competitive benefits to Norwegian and European process industry. SAM is coordinated by Bilfinger Industrial Services Norway AS with Eramet Norway AS, Equinor ASA, Yara International ASA, Elkem ASA, Norsk Hydro ASA, Boliden Odda AS, REC Solar Norway AS as industry partners. The R&D partners are SINTEF AS and University of South-Eastern Norway. Boliden's project goal is to optimize the zinc and sulfate contents in purified process fluids before electrolysis. During the project a new instrumentation solution for fluid density has been evaluated and put to use. The new solution gives more accurate and stable calculation of zinc content in the solution. Other density measurements and measurement principles have been considered to gain insight into other parts of the process. Continued work has been done to model parts of the process. An expanded measurement campaign was performed recently to facilitate tests of spectroscopy analysis of process liquids. Elkem and Eramet have installed multiple sensors: a thermographic camera, weight sensors and a level measurement system. These will reduce the need for human presence in the challenging environment of the tapping area (hot, open melts, dust, noise…) and ensure better data collection from the process. The SAM project has identified and installed sensors that can survive in this challenging environment and replace visual observations and manual measurements. Equinor has focused on testing and analysis of vibration sensors and measurements of multiphase flow in pipes. Field data from acoustic sand detectors has also been pre-treated and analysed. For many applications it is possible to determine the flow rate bases on such signals, but further research is needed to separate and individually measure flow rates for each phase (oil, gas, water). Different modelling approaches have been tested to predict flow rate and composition; however, more robust methods must be developed before the models can be applied in the field. Hydro seeks to understand and control of the heat balance when designing an electrolysis cell for production of aluminum. It is important to be able to measure heat flux with sufficient speed and accuracy. In the project Hydro and SINTEF have evaluated several new ideas to improve our knowledge of and ability to measure the heat balance of our electrolysis cells. With a combination of theoretical analysis and practical testing, the partners have gained a basis for choosing the methods to be developed farther. A novel sensor concept will be patented and possibly commercialized. REC Solar has improved the solidification process by implementing results from the SAM project, and has now achieved satisfactory levels of nitrogen and carbon impurities in their silicon, which at high concentrations can disrupt the customer process. In 2021, the project changed its focus upstream to the new, in-house-developed process for kerf melting. Process stability and yield are sensitive to small variations in temperature and chemical composition. We aim to establish a data model where temperature, weight, and power measurements are connected to image analysis of video from melt pouring. We will build a database with key parameters based on time-series data from the process. Yara: has performed a comprehensive study of in-line dust measurement techniques in one of Yara's NPK production plants in Porsgrunn, with the aim to optimize operation and better understand the causes of dust emissions. Several production campaigns for relevant focus products has been carried out. Two campaigns have been carried out with production of a specific NPK product associated with high NH3 emissions. In collaboration with Rheality Ltd, a piezo electric element has been installed in a process line. One sampling campaign has been carried out, and a model using AI to predict process fluid viscosity is under development.

Bilfinger: Gjennom BCAP og SINTEFs "Bedrock"-plattform, er det utviklet nye algoritmer for dataanalyse og optimering. Nøkkelparametere for forbedring i produksjonsprosesser er identifisert. Utvidelse av BCAP med nye moduler kan gjøre dem til foretrukket digital plattformleverandør med betydelig vekst i omsetning og fortjeneste. Yara: Gjennom omfattende studier er online støvmålinger optimert for å gi best mulig informasjon til NPK-produksjon, med mål om å redusere støvutslipp. Det har vært oppfølging av produksjonskampanjer for å modellere kritiske prosessvariabler og forbedre produktegenskaper. En piezoelektrisk enhet er installert i samarbeid med Rheality Ltd for å måle viskositet og forbedre prosesskontroll og -egenskaper. Bruk av støvmålinger og AI-modeller for viskositetsprediksjon vil optimere produksjon og driftsstabilitet. REC Solar: Implementering av SAM-resultater har forbedret stivningsprosessen og oppnådd tilfredsstillende nivåer av nitrogen og karbon. Datamodell basert på tidsseriedata og eksterne signaler som temperatur, vekt og kraft er under etablering. Implementering av SAM-resultater forbedrer størkningsprosessen og øker produksjonsutbytte Hydro: Det er utviklet en ny sensor for måling av varmebalanse i elektrolyseceller som kan monteres utvendig på cellen. Den nye sensoren vil patentbeskyttes og deretter forsøkes kommersialisert gjennom en 3.part. Slike sensorer for nøyaktig måling av varmebalanse kan forbedre design og drift av aluminiumproduksjonsprosessen. Equinor: Vibrasjonssensorer og akustiske sanddetektorer gir mulighet for nøyaktig prediksjon av strømningshastighet og sammensetning i rørledninger. Videre forskning kreves for nøyaktige beregninger av hver produsert fase. Prediksjon av strømningshastighet og sammensetning i rørledninger kan forbedre prosesskontroll og optimalisere produksjon. Boliden Odda: Nytt instrument for tetthetsmåling gir bedre nøyaktighet og stabilitet i beregning av sink- og sulfatinnhold. Utforsking av andre tetthetsmålingsteknikker kan gi ytterligere forbedret prosessforståelse. Effekter inkluderer forbedret prosesskontroll og produksjonsytelse, og høyere kvalitet. Eramet og Elkem: Installasjon av termokamera, vektsystem og nivåmålingssystem i utfordrende tappingssoner har redusert behovet for mennesker i dette krevende arbeidsmiljøet (varme, flammer, damp, støy og støv) og har forbedret datainnsamlingen. Det er fremtidige kostnadsbesparelser ved reduksjon av slaggmengde, og HMS-gevinster ved erstatning av manuelt arbeid i krevende miljø. Disse virkningene og effektene viser hvordan SAM-prosjektet har bidratt til å forbedre konkurransedyktigheten og innovasjonen i de deltagende norske prosessbedriftene. Utviklingen av teknologi og kompetanse vil bidra til den norske prosessindustriens mål om klimanøytralitet innen 2050.

The primary objective is to demonstrate that, by adapting digitalization methodologies to process industries, significant optimization of the processes and reduction of their environmental impacts can be achieved. The innovations include optimization and control of industrial production processes using big data analytics, new online sensors and data-based models. In close cooperation with the project partners the innovation will lead to the development of an algorithm for self-adapting model, which the partners can potentially integrate into their existing data systems at the end of the projects. The research will involve combining existing process data with new data gathered from a grid of sensors assembled in an innovative manner. These new process data and the data-driven models will be used to define the optimum set points for the processes. The R&D challenges are many-fold, where the most critically are: i) harsh environment makes on-line measurement non-operational in practice, ii) identifying the most causal process variables from correlated data, iii) the majority of available data have poor quality – resulting in poor predictive models, iv) unexpected changes in critical process variables not integrated in the data-based models (e.g. future change in raw material quality) which reduce the autonomy potential of the SAM module. By involving several process industries, this project will give researchers the opportunity to develop generic algorithms suitable for a broad range of processes.