Skip to Content
Exit Menu
Mr Bojie Sheng

Mr Bojie Sheng
Research Fellow

Brunel Innovation Centre - Cambridge

Summary

Bojie Sheng received his BSc degree from the School of Electrical Engineering, Wuhan University, China in 2008. After he obtained his PhD degree in 2014 in Glasgow Caledonian University, UK, he joined the University of Strathclyde as a research associate before joined BIC as a Research Fellow in 2018.

Bojie is a highly motivated, diligent and responsible researcher specialised in electrical/mechanical engineering and machine learning/computer vision, with strong skills in both computer software and hardware development. His research is mainly focused on condition monitoring, anomaly detection, defect diagnosis and prognosis.

Qualifications

BSc, PhD

Responsibility

My responsibilities for six innovative ML/computer vision projects about robot, drone, medical device, green farm, additive manufacturing and battery monitoring applications are developing computer vision/deep learning algorithms (MLP, CNN, GRU, LSTM models) for image/time series data classification, regression and object detection.

In the robot project, I developed different CNN models with image augmentation methods including GAN for image classification for a marine robotic able to perform recognition and cleaning of biofouling on monopiles of offshore wind turbines. Prune and quantization techniques were also applied to deploy the DL models in the robot embedded system for real time application. In the drone project for bridge inspection, I developed different DL algorithms to process different types of data: DL models were developed for image classification (CNN) and object detection (YOLO) based on image data from cameras, different time series data processing methods (STFT, Wavelet) were applied for structure thickness estimation based on ultrasonic testing data while ultrasonic (conventional ultrasonic and phased array ultrasonic) data were converted into images through STFT, then 4 CNN models were developed separately for good/substandard data classification, surface type classification, defects classification and thickness estimation (regression). In the medical device project, 1D CNN model was developed to process IR and RF data for blood glucose concentration prediction to monitor the health of a person with diabetes. In the green farm project, MLP/GRU/LSTM models were developed to process time series data from different sensors (temperature, soil moisture, leaf wetness, NPK, etc.) for greenhouse condition monitoring. In the additive manufacturing project, data pre-processing methods (median filtering, DWT compression, PCA, etc.) and ML methods (KNN/MLP) were applied to process eddy current data for defects classification.