Intellectual Property India Publishes Patent Application for Ai-Based Acoustic Anomaly Detection And Remaining Useful Life Estimation System For Predictive Maintenance Of Industrial Machinery

MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641087844 A) filed by Immanual R; Sanjana N; Poornima Dhandapani; Sangeetha Shanmugam; Ramkumar Elangovan; Glory Priyadharshini Jeyabal; and Tharun Bhuvaneswari Shanmugavel on July 17, 2026, for Ai-Based Acoustic Anomaly Detection And Remaining Useful Life Estimation System For Predictive Maintenance Of Industrial Machinery.

Inventors include Immanual R; Sanjana N; Poornima Dhandapani; Sangeetha Shanmugam; Ramkumar Elangovan; Glory Priyadharshini Jeyabal; and Tharun Bhuvaneswari Shanmugavel.

The application for the patent was published on July 24, 2026, under issue no. 30/2026.

Abstract: An AI-based acoustic anomaly detection and remaining useful life estimation system (100) for industrial machinery comprising an acoustic sensing subsystem (102) with a microphone (104), an audio preprocessing pipeline (108) that computes a short-time Fourier transform and mel-spectrogram, normalises it, and segments it into overlapping spectrogram patches, and an autoencoder defect detection engine (120) comprising an encoder sub-network (122), a bottleneck layer (124), and a decoder sub-network (126), trained exclusively on spectrogram patches derived from healthy machine sound. A reconstruction error calculation module (128) and anomaly threshold module (130) flag anomalous patches, while a health index computation module (132) converts the reconstruction error into a normalised health index. A degradation trajectory module (134) and curve fitting and extrapolation module (136) fit and extrapolate a degradation model to a failure threshold, and a remaining useful life estimation module (138) computes a continuously updated remaining useful life, triggering a maintenance alarm module (140) when the remaining useful life falls below a configurable maintenance lead time. A data logging module (142) records full traceability metadata. The system achieves acoustic anomaly detection accuracy exceeding a receiver operating characteristic area under curve of 0.88 and is deployable using low-cost, non-contact microphone hardware, making it accessible for small and medium manufacturing enterprises.

Disclaimer: Curated by HT Syndication.