4-year studentship available for UK, EU & International* students, who possess a first class or 2.1 (Honours), or equivalent EU/International qualification, in the relevant discipline of Computer Science, Engineering, Physics or Mathematics. The candidate should have the following
The NMIS (National Manufacturing Institute Scotland) Doctorate Centre in Advanced Manufacturing, along with Howden Compressors are looking to jointly fund PhD studentship in the area of industrially applied data analysis for providing real-time insights into machinery condition.
Howden Compressors, in partnership with PTC & Microsoft have developed a platform to monitor the air & gas handling assets they design, test and manufacture for their clients all over the world. Platform has been commercialised & assets connected all over the world.
Assets ranging from small industrial fans to large multi-mega Watt bespoke compressor packages designed for hydrocarbon gas applications in potentially explosive environments. The assets have been connected & process & machine health data is being continually streamed into the platform from these assets we now have a historical dataset for a range of the fan & compressor technologies. A Howden engineer can study data & give guidance to solve a problem or advise on root cause.
Working with experts at NMIS & Howden Compressors to develop algorithms that will provide improved prognostics and insights into asset quality, including expected time to component failure and evaluate the energy efficiency of certain systems. The algorithms will be developed using the labelled historical datasets & may use statistical methods, machine learning or artificial intelligence. Successfully developed models on historical data will then be transitioned to the Howden platform for further testing an evaluation as part of the wider decision support platform.
The project aim is to develop a set of tools and processes within the platform that can automate the detection of the anomalous behaviour across the entire spectrum of air and gas handling assets with respect to the health or the performance of the asset. Majority of data is time-series but may include frequency domain. The intention is to use the wide array of published models such as 1D CNN’s or alternatives and optimise them based on the data format and desired outcomes.
Specifically regarding the health of the assets the project should first conduct an audit of the serviceable components on the connected assets and then be able to measure, score the health / remaining life of the component, and track the model accuracy, by utilising the instrumentation on the package. This instantly provides benefit to the end users as they will then be able to extend maintenance intervals or schedule them forward to prevent unplanned downtime. When the asset is provisioned with high frequency vibration monitoring instrumentation an automated reporting tool to score harmonic peaks and advise on potential failure modes is to be developed.
With regards to the performance monitoring on the assets that use huge amounts of electrical energy the objective here is to not only advise on the efficiency of the asset by looking at the energy consumed and the delivered flow versus the original design calculations but also asses control strategy of the assets on site.
On completion the current Uptime platform will become more scalable, by removing a large portion of the manual reporting and automating it, than it currently allows the business itself to pivot more towards a repeatable aftermarket servitisation model. The models developed to support the Uptime platform will be demonstrated on a subset of systems within the Howden portfolio but will be designed to allow adoption within other systems after an appropriate model hyperparameter tuning and training period.
Based on your current searches we recommend the following search filters.
Based on your current search criteria we thought you might be interested in these.
A decentralized, data driven health monitoring and diagnostics platform based on Artificial Intelligence (AI) and wearable/portable Internet of Medical Things (IoMT) sensors
Anglia Ruskin University ARU