Siddhi Bhandari

Topic: Development and validation of a neural network for the identification and evaluation of systematic geometric surface defects on car body outer skin components. 1. Creation of a defect dataset for neural network training. 2. Definition and comparison of generated images in virtual audit room with real captured images. 3. Comparison of different existing neural networks. 4. Investigations on CNN and anomaly detection networks. Hyperparameter study to modify network to fit best to the problem.

Generation of Input Data for Neural Network Training from 3D Scanned Parts (stl files) for systematic classification of Surface Defects. 1. Creation of Process flow for generation of Input data (Images) from a 3D scanned part 2. Use of Application Programming Interface (API) for automation of each step with various applications as GOM inspect, Beta CAE ANSA and META, Blender. 3. Use of Blender & META for creation of rendered images of systematic surface defects. 4. Training of CNN for detection of defect.

Virtual FEA Analyst for Structural Safety of Passenger Vehicles (Crashworthiness) - Passive Safety 1. Work for various NCAP legislations worldwide in the virtual world. 2. Load cases: High-speed frontal, Side, Rear, Low-speed crash, Pedestrian Protection, and Static strength test for door and roof. 3. Highlights: Multidisciplinary optimization for weight and cost reduction using DOE, prework on frontal impact rule with the mobile progressive deformable barrier (MPDB). Design and optimization proposals.