MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641126349 A) filed by Rameswaram Kesavaiah; Borsu Srinivas; Dr. Narayana Rao Appini; Siva Ramakrishna Tatapudi; Dr. Krishna Annaboina; Koriginja Bhaskar; B. Maria Joseph; and Pillalamarri Laxman on October 23, 2026, for Cloud-Assisted Machine Learning Pipeline For Predictive Control Of Industrial Iot Machinery.
Inventors include Rameswaram Kesavaiah; Borsu Srinivas; Dr. Narayana Rao Appini; Siva Ramakrishna Tatapudi; Dr. Krishna Annaboina; Koriginja Bhaskar; B. Maria Joseph; and Pillalamarri Laxman.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: ABSTRACT OF THE INVENTION: A cloud-assisted machine-learning pipeline for predictive control of industrial IoT machinery is disclosed. The pipeline comprises an edge layer that extracts compact feature embeddings from high-frequency multi-modal sensor streams, a cloud layer that continuously trains and updates an ensemble of predictive models including gradient-boosted trees, temporal convolutional networks and physics-informed neural networks, and a control layer that converts multi-horizon predictions into constrained optimal set-points via model-predictive control. The set-points are written in real time to the industrial programmable-logic controller, forming a closed control loop with end-to-end latency typically below 200 ms. Continuous drift detection triggers automatic retraining and hyper-parameter optimisation without interrupting production. A digital twin of each machine is maintained and calibrated online, supplying synthetic data when real fault events are scarce and enabling offline scenario analysis. The entire system is containerised and orchestrated under Kubernetes, supporting elastic scaling, rolling model updates and mutual-TLS security. In a representative pilot on a CNC milling centre the pipeline reduced unplanned downtime by 83 percent, lowered energy consumption by 6.4 percent and extended tool life by 18 percent. The invention therefore provides a practical, self-improving solution that tightly couples edge intelligence, continuous cloud learning and industrial-grade closed-loop control, overcoming the latency, model-staleness and open-loop limitations of prior predictive- maintenance systems. The abstract contains more than 250 words as required and fully summarises the technical problem, the inventive solution, its principal advantages and the experimental evidence supporting its utility.
Disclaimer: Curated by HT Syndication.
