A digital twin approach to noise exposure modelling and risk analysis

Excessive noise is a major hazard in oil and gas, where heavy machinery and steel structures often exceed regulatory limits. We introduce a Noise Exposure Digital Twin (NEDT) that fuses historical surveys, live sensors, and equipment data into a real-time sound-propagation and dose model. At its core is a physics-informed neural network (PINN) surrogate of the Helmholtz equation, coupled with a personal-exposure module and a Dynamic Noise Exposure Risk Index (DNERI) that combines instantaneous sound pressure level (SPL) with cumulative dose. Applied to Asset X (a North Sea offshore platform) in a simulated operational pilot via three-day replay of Noise Exposure Management System (NEMS) field data, the PINN inferred spatial noise maps with a ∼1.5 dB mean error and predicted daily personal dose within ±10 %. Early-warning logic provided a median ∼32-min lead time before exceeding 85 dB(A) or 100 % dose. Compared with annual surveys and limited dosimetry campaigns, the NEDT delivers continuous virtual dosimetry for all crew and robust ALARP guidance. Integrating PINNs with Internet-of-Things (IoT) data in a digital twin thus enhances noise-risk management and worker safety in high-noise operations. • Developed a noise exposure digital twin using Physics-Informed Neural Networks (PINNs) to map dynamic acoustic risks. • Integrated real-time personnel tracking and acoustic data to assess exposure and simulate safety scenarios. • Demonstrated scalability and cross-platform adaptability with minimal retraining required on new assets. • Validated predictions against real-world measurements and traditional monitoring methods.