Tenders are invited for Ai-Based Failure Prediction to Improve Operational Payload Availability - Expro plus To investigate and develop methodologies using AI, for prediction of core units and payloads failuresbefore their occurrence and assess how this information can contribute to system unavailability.DescriptionFailure detection plays an important role in monitoring the integrity of electrical components but alsounits, payloads and overall system. Failure detection is performed as part theFault Detection Isolation and Recovery (FDIR) system, however, at the moment, little is done in terms of failure prediction and assessment of changes in units behaviour. Most of the work performed in this scope targeting on ground data analysis and not on-board real-time assessment.Artificial Intelligence (AI) has the potential to contribute to a better failure prediction process by looking and analysing information coming from components/units/subsystems and assessing if theirperformance is decreasing with time. Failure and anomaly prediction capabilities based on datatrend prognostics could lead towards an increase in the overall availability of the system through either applying recovery actions before occurrence of failure, or allowing to perform a smoother and faster switch to the redundant branch which would result in less or no interruption of payload or system operation. Finally, such approach couldhave the potential to better understand the degradation effects of units and system and may lead to a better overall mission operationalavailability. The goal of the activity is to research mission payloads and understand how, throughthe use of AI, the outage can be minimised more than through the use of common and currentapproaches.The activity encompasses the following tasks:-Identification and gathering of potential data of payloads that could be used for AI development (e.g. Sentinel, MSG, Telecom, Constellations, etc.) and for overall Neural Network training- Identification of the best types of machine learning algorithms that allow to evaluate behaviours based on training data. Analysis of the statistical techniques, data distribution and probability theory. Provide methods to train and cross-validate and select as approachone of the two broad categories: supervised or unsupervised learning- Development of AI methodology to predict failures in payloads or core units with the scope of minimising outage- Apply the developed AImethodology to a small case study or proof-of-concept (e.g. at payload unit level) for which there is sufficient telemetry data available- Perform comparison between conventional availability results and the resulted availability when the new developed AI approach is being used- Provide inputs and lessons learned to the existing (i.e. SAVOIR) body of knowledge. Read less Tender Link : https://esastar-publication-ext.sso.esa.int/ESATenderActions/filter/open
We takes all possible care for accurate & authentic tender information, however Users are requested to refer Original source of Tender Notice / Tender Document published by Tender Issuing Agency before taking any call regarding this tender.
