Generation of Non Invasive Measurement Signals for Digital Twin Based Knee Joint Deterioration Simulating of Drilling Employers

Drilling employers’ knee joints deteriorate while influencing and developing by high pressure underground works, with undemand health issues influence productivity and downtime. Intermediate signal for simulating deterioration processes is necessary for helping health monitoring and work planning. Using acoustic emission(AE) as testing material, this study leverages generative AI agents and innovative prompting to generate simulation data, to explore signal generation and pattern classification of healthy and osteoarthritis(OA) knee signals among various age levels. The study is divided into two parts: AI driven knee AE generation(Ex-1) and pattern classification of (Ex-2). Ex-1 verifies the generation of healthy and OA simulation signals. A benchmark dataset consists of 53 knee joint AE signals, categorised by age and health status into 5 groups, namely, healthy-young(H1), healthy-middle-aged(H2), healthy-elderly(H3), OA-middle-aged(OA1), OA-elderly(OA2), are used for simulation. Through prompting, the GENAI-agent produces 20 signal sets per group, as well as corresponding novel designed visual based profiles(HD), and the autonomous pattern classification workflow could be done by using AIGC in order to verify the capability. The proposed approach for AE detection in knee joint inspection shows that AIGC generated AE hits closely match actual data, except H3, differing by ~100 hits. Ratios for others range 0.8–1.2. Generated AE waveforms highly resemble measured ones, with correlation coefficients 0.76–0.99 and standard deviations <0.1. Euclidean distance based classification achieves ~97% accuracy, indicating potential for digital twin reconstruction of knee biomechanical deterioration via pre-trained models and prompt engineering. For pattern classification, GENAI-agent outperforms GENAI, exceeding 90% accuracy in OA1/OA2 diagnosis versus GENAI's 80%, demonstrating feasibility in occupational health management with minimal model modification. These findings enabling missing data compensation and high-precision, full-life cycle digital twin generation for managing occupational health risks for O&G employers. With integration of proposed approach into existed digital twin system, it is possible to help decision makers to make even comprehensive decisions.