Coronavirus (Covid-19): Latest updates and information
Skip to main content Skip to navigation


Select tags to filter on

Python Library Launched: Bayesian Deep Learning for Manufacturing 2.0!

Sumit Sinha, Dr Pasquale Franciosa and Prof. Darek Ceglarek launched a python library implementing artificial intelligence algorithms for root cause analysis of assembly systems. The work includes a Python (Bayesian Deep Learning for Manufacturing) and Matlab (Variation Response Method (VRM)) integrated library. Key highlights of the library include:

1. Multi-task attention based Bayesian 3D U-Nets enabled by Bayes-by-Backprop and Flipout, for Root Cause Analysis of manufacturing systems with uncertainty quantification

2. Deep Reinforcement Learning models for control and correction using customizable CAE based (multi-physics) manufacturing systems

3. Low latency integration of Matlab environments and TensorFlow models

4. Closed-loop training, continual and transfer learning integrations for rapid training and scalability

5. Integration of 3D Grad-CAMs for architecture interpretability

6. Keras Tuner for rapid architecture prototyping and selection

7. Multi-station manufacturing case studies for training and benchmarking

Tue 26 Jan 2021, 11:41

Journal Paper published in IEEE Transactions on Industrial Informatics (Impact Factor: 9.1): Object Shape Error Response using Bayesian 3D Convolutional Neural Networks for Assembly Systems with Compliant Parts Date

Sumti Sinha, Dr Pasquale Franciosa and Prof. Darek Ceglarek published a paper proving a transformative framework for achieving Zero Defect Manufacturing. The paper proposed a novel Object Shape Error Response (OSER) approach to estimate the dimensional and geometric variation of assembled products and then, relate, these to process parameters, which can be interpreted as root causes (RC) of the object shape defects. The OSER approach leverages Bayesian 3D-Convolutional Neural Networks integrated with Computer-Aided Engineering (CAE) simulations for RC isolation. Compared with the existing methods, the proposed approach (i) addresses a novel problem of applying deep learning for object shape error identification instead of object detection; (ii) overcomes fundamental performance limitations of current linear approaches for Root Cause Analysis (RCA) of assembly systems that cannot be used on point cloud data; and, (iii) provides capabilities for unsolved challenges such as ill-conditioning, fault-multiplicity, RC prediction with uncertainty quantification and learning at design phase when no measurement data is available. Comprehensive benchmarking with existing machine learning models demonstrates superior performance with R2 =0.98 and MAE = 0.05mm, thus improving RCA capabilities by 29%

Tue 26 Jan 2021, 11:36

Paper presented at 18th IEEE Conference on Industrial Informatics (INDIN) by Sumit Sinha: Object Shape Error Response using Bayesian 3D Convolutional Neural Networks for Assembly Systems with Compliant Parts

Sumit Sinha presented his work co-authored by Dr Pasquale Franciosa and Prof. Darek Ceglarek. The presentation proposed the Object Shape Error Response methodology extending traditional Object Detection in Deep Learning. It is the first uncertainty enabled model overcoming shortcomings of state-of-the-art models for Root Cause Analysis of Assembly System.

Tue 26 Jan 2021, 11:28

Catapult WMG Workshop - Get Connected: When CAE simulation meets Artificial Intelligence, 4th February 2021

The workshop demonstrated the use of Digital Twins, Multi-Fidelity CAE Simulations and Deep Learning in manufacturing that enabled right-first-time capability to reduce engineering changes during installation, commissioning and production. The developed software was demonstrated and made available with open-source code. Key external speakers included managers from Jaguar Land Rover (JLR) and technical experts from Mathworks (MatLab). Key Internal Speakers included researchers within the Digital Lifecycle Management (DLM) group. The event included 50 attendees from both academia and industry including major industries such as BMW, Rolls-Royce, JLR and Constellium. The attendees were from diverse backgrounds such as automotive manufacturing, CAE simulations, data science and software development. Key presentations included:

1. Deep learning enhanced digital twin for Closed-loop In-Process Quality Improvement by Prof. Darek Ceglarek

2. Variation response method CAE simulation suite by Dr Pasquale Franciosa

3. Deep learning in manufacturing: predicting and preventing manufacturing defects by Sumit Sinha



Tue 26 Jan 2021, 11:26

Older news