Closed Loop Digitalised Data Analytics and Analysis Platform (DAAP) for Intelligent Design and Manufacturing of Power Electronic Modules

Power electronic modules (PEMs) and higher-level systems play an increasingly important role in adjustable-speed drives, unified power quality correction, utility interfaces with renewable energy resources, energy storage systems, electric or hybrid electric vehicles and more electric ship/aircraft. The power electronic technologies provide compact and high-efficient solutions to power conversion but deployment of power electronic modules in such applications comes with challenges for their reliable and safe operation.

This project aims to address four key challenges which the power electronics manufactures, and PEM end-users continue to face:

Challenge 1: No in-line and non-destructive inspection methods for PEM package quality and internal integrity assessment (wire bonds, die attachment and encapsulant) embedded within the production line.

Challenge 2: No comprehensive PEM data on design-quality-reliability characteristics, no processes for chartreisation and test data integration and management, and for data modelling and analysis.

Challenge 3: No advanced capabilities for accurate assessment of PEM deployment risks and for lifetime management.

Challenge 4: No or limited data is fed back from end-users to PEM designers/manufacturers, no application-informed design and manufacturing quality.

The project seeks to develop a digitalised Data Analytics and Analysis Platform (DAAP) for PEMs. The following novel and beyond current state-of-art developments in the project address the above stated challenges:

1) Non-Destructive Testing (NDT) with real-time data acquisition capability. A novel technique for NDT using LF-OCT imaging will be enhanced and optimised to provide quality data for individual PEMs. The proposed NDT method can quantitatively measure the mechanical deformation of gel-encapsulated bonding wires down to nanometer level. It can capture an entire cross-sectional image without any mechanical scanning, providing novel capability of running in-line with the packaging process.

2) Quality Predictions using AI and Machine Learning (ML): Research on integration and use of multiple data formats and sources, including standard datasets of electrical parameter test measurements, image data from in-line LF-OCT, and off-line X-ray and other imaging techniques, will be undertaken. The integrated data will underpin the accurate and automated quality evaluation of each individual PEM by enabling the development of ML and Deep Learning models. The modelling capability will enable packaging quality evaluations based on comprehensive sets of design and packaging process attributes.

3) Reliability Predictions. Current state-of-art in design-reliability and in-service degradation modelling for PEMs will be advanced through the proposed inclusion of manufacturing quality characteristics and design attributes in the reliability predictions. This will result in enhanced knowledge and more accurate, quality-informed reliability modelling and insights into the relations between design, quality and reliability by analytics of manufacturing and end-user data.

4) Data-Modelling-Optimisation Capabilities' Integration. The proposed integration (DAAP) of data, information exchange, and different modelling capabilities with multi-objective optimisation methods will be a novel development. The proposed optimisation routines will provide new capabilities for power semiconductor packaging design (e.g. module architecture, materials, interconnect solutions, application-specific reliability performance, etc.) and optimal process control on the manufacturing line.