Predictive maintenance and edge analytics

IBM and Siemens AG are prominent companies in the organization data, with 894 and 695 documents in the window, respectively. Cisco Systems, Schneider Electric, Honeywell, ABB Group and Nvidia are also active. Their capabilities span edge computing, industrial automation, condition monitoring, fault diagnosis and digital-twin platforms.

Applus Inc. is the sole startup named in the organization data, with 161 documents in the window. The available information does not describe a specific edge-maintenance product or establish its technical performance. Its presence indicates activity, but the evidence is not enough to compare startup prospects or identify a clear market leader.

Lovely Professional University has 20 documents in the window, while the data lists 16 for the University of Business and Technology and 14 for Kongu Engineering College. Their work covers machine-learning maintenance, real-time industrial sensor data and lightweight embedded models. These figures indicate research activity, not a measure of research quality.

The funding data lists $9,233,707 across 9 grants since 2020 and 3 funding rounds totaling $327,329,945. These are distinct funding categories and may include broader industrial or infrastructure projects rather than funding dedicated solely to predictive maintenance and edge analytics.

The latest listed round is Siemens AG's $210,780,000 project and infrastructure finance entry from March 2026. Earlier entries include $111,280,445 for Siemens AG in April 2025 and a $5,269,500 grant for University College Dublin in June 2025. The Siemens entries are not identified as dedicated predictive-maintenance funding.

KinNECTed AI Inc and 2541980 ALBERTA LTD. lead the listed patent holders, with 2 patents each. Troxler Electronic Laboratories, Goco, Katz Water Technology LLC and Schneider Electric Systems, USA Inc. each have 1 listed patent. These counts describe the holders shown in the data, not their overall patent portfolios.

Patent activity is declining in the measured window: 8 patents are listed, with growth of -40.0%. That contrasts with 341 publications in the last year and publication growth of 221.7%. The difference suggests that recent technical activity is more visible in research output than in new patent filings.

Recent developments include FusionNet from Swinburne University of Technology in August 2026 and Telit Cinterion's edge AI SDK announcement in September 2026. FusionNet targets rare faults, sensor degradation and interpretable results, while the SDK supports machine-learning models running on cellular modules. These developments connect model design with practical on-device deployment.

The main approaches are signal processing, physics-based prognostics, supervised fault diagnosis, unsupervised anomaly detection, edge inference and digital-twin methods. Signal processing works with equipment data; learning methods find patterns in labeled or unlabeled signals; physics-based methods encode asset behavior. Edge inference concerns where models run, while digital twins add a virtual asset context.

The four application areas identified are discrete manufacturing (1,123 documents), process industries (207), energy and utilities (300), and transportation and fleets (395). The examples range from monitoring factory equipment and industrial processes to maintaining infrastructure and vehicles. Each setting brings different sensor, reliability and connectivity requirements.