OPtimise POwder pRoducTion UsiNg Intelligent Sensing Technologies (OPPORTUNIST)

Project ID: 2228cc1456 (You will need this ID for your application)

Research Theme: Manufacturing The Future

UCL Lead department: Mechanical Engineering

Department Website

Lead Supervisor: Chu Lun Alex Leung

Industry partner: Carpenter Additive

Stipend enhancement: £1,500 pa

Project Summary:

Metallic powders play a crucial role in the global transition towards industrial 4.0 and a net-zero society as they are being used for a wide range of digital manufacturing technologies and industries, including press & sinter, metal injection mould, and additive manufacturing. Gas atomisation (GA) is one of the most common powder production processes owing to its excellent chemistry control, wide particle size distribution, and high flexible capacity. Currently, the nozzle of the GA unit often suffers from cogging and melt splashing during GA, damaging the nozzle and reduce production yield. In worst case scenario, the damage nozzle may pose safety hazards and halt production. These anomaly events are governed by complex gas flow and gas-liquid interactions within milliseconds; their underlying mechanisms are not well understood as they are difficult to monitor and predict. This project aims to develop a novel imaging and data analytics platform to understand the fundamental mechanisms of the GA process, continuous monitoring and predict anomaly events. The aim is divided into four objectives: (1) a GA simulator will be developed at UCL coupled with an ultra-fast speed imaging platform (up to 1MHz) to reveal and elucidate the liquid breakup, cogging, and splashing mechanisms. The GA simulator will be used to test fluid with various viscosities to simulate molten metal at various compositions/temperature. (2) The collected data will be used to develop data-driven machine-learning (ML) models that will evaluate a combination of time-series data, fluid viscosities, gas flow velocities, nozzle geometry, etc. (3) Such imaging platform will then be replicated in a GA unit at Carpenter Additive to collect and provide real-world training data to generalise the ML models developed in the laboratory. (4) Finally, an ensemble ML model will be deployed to test its suitability and generalisability for prediction of anomaly events.