True North Prognostics

True North Prognostics (TNP)is a San Diego-based predictive analytics and prognostics company founded in 2025. TNP develops next-generation, AI-powered prognostic solutions built on more than 30 years of proven MSET innovation. These solutions detect the earliest signs of degradation in complex engineering assets, before developing faults lead to costly failures or unplanned downtime. Serving defense, transportation, energy, data centers, and advanced manufacturing, TNP helps organizations improve reliability, availability, and serviceability (RAS) while optimizing energy efficiency.

True North Prognostics is a member of the NVIDIA Inception program.

TNP’s mission is to harness the digital and AI revolution to build technologies that maximize ROI from business-critical industrial and defense assets, enable predictive and prescriptive maintenance via early warnings of tiny developing faults, and for safety-critical assets, reduce penalizing challenges to system availability goals while increasing human safety margins. Through innovation and collaboration, TNP develops tools that anticipate problems before they occur, helping industries operate more safely, reliably, and efficiently.

At the heart of TNP's work is AI-MSET™(Artificial Intelligence Multivariate State Estimation Technique), an advanced machine-learning technology designed to detect the earliest indications of mechanisms that can lead to system failures. AI-MSET learns how systems behave under normal conditions and continuously analyzes sensor data to identify subtle changes that may signal developing anomalies in engineering assets. It can also detect the onset of malicious activity, strengthening prognostic cybersecurity in SCADA systems and other business-critical IT assets and networks.

Unlike traditional monitoring systems that apply high and low thresholds to individual signals (univariate alerting), AI-MSET learns the relationships among dynamically interacting sensors in real time. This multivariate approach achieves very high prognostic sensitivity while maintaining extremely low false- and missed-alarm probabilities, even in noisy environment.

AI-MSET™ is the third generation of MSET-based prognostics, building on more than 30 years of development and real-world application in business-critical and safety-critical environments. TNP's technical team includes multiple experienced MSET innovators and practitioners, including CTO Kenny C. Gross, who led the original development of MSET at Argonne National Laboratory in the 1990s and has continued advancing MSET-based prognostics ever since.

AI-MSET extends that proven foundation with Gen3 innovations that deliver greater anomaly sensitivity, improved signal fidelity, exceptional scalability, and substantially lower computational requirements, thereby enabling earlier and more reliable detection of developing problems across today's increasingly complex engineering assets.

The original MSET was approved by the U.S. Nuclear Regulatory Commission (NRC) for monitoring applications across U.S. commercial nuclear power plants. It was also approved by NASA for Space Shuttle applications and by the Federal Aviation Administration (FAA) in 2002 for commercial aviation sensor prognostics.

MSET innovations have been published in peer-reviewed IEEE literature. Published comparisons have demonstrated advantages over evaluated alternatives, including LSTM and other neural networks, in accuracy, early-warning lead time, false- and missed-alarm performance, and computational efficiency, with computational requirements two to three orders of magnitude lower in comparable signal-monitoring applications.

For high-volume applications, AI-MSET's deterministic mathematics supports fine-grained parallelism on modern multicore CPUs and GPUs. AI-MSET can monitor dense sensor streams on comparatively modest hardware. Smaller deployments involving dozens or hundreds of signals can run on compact laptops, 1RU enterprise servers, or Raspberry Pi platforms.

TNP's technologies are applied across many fields, including:

• Defense and Aerospace: Early detection of incipient faults in defense assets across land, air (fixed-wing and rotary-wing), sea (surface and underwater), and space, as well as in commercial aircraft; applications include landing-gear prognostics, reduction of fire/smoke/fume (FSF) false alarms, and reduction of no-fault-found (NFF) events.

• Transportation: Reducing train derailment risk; detecting malicious-intrusion software hacks into Water Treatment SCADA systems, and into safety-critical railroad switch-and-crossing (S&C) systems for railroads; and characterizing the structural integrity of railroad bridges, ties, and track substrates using low-cost, mag-mounted triaxial accelerometers aboard locomotives.

• Situational  Awareness: Supporting human-in-the-loop (HITL) supervisory control of complex engineering assets.

• High-Frequency Waveforms: Processing high-frequency waveforms on low-cost computing platforms while substantially reducing sensor I/O bandwidth, including vibration (kHz), acoustics (MHz), and infrared thermal imagery (GHz).

• IT Systems: Electronic prognostics, optimized power management, software-aging detection and rejuvenation, and prognostic cybersecurity.

• SCADA Systems: Monitoring industrial, utility, water-treatment, railroad, and defense supervisory control and data acquisition (SCADA) environments to detect the early onset of malicious intrusion activity with ultra-low false- and missed-alarm probabilities.

• Manufacturing: Predictive and prescriptive maintenance that reduces downtime, scrap, and operations and maintenance (O&M) costs while enhancing product quality and reducing warranty costs and spare-parts logistics for manufacturers and their customers.

• Energy: Monitoring instrumentation and sensors in commercial nuclear power plants; improving wind-turbine efficiency and RAS; and increasing energy efficiency in HVAC and other feedback-controlled engineering systems by reducing noise in raw sensor inputs to multi-input, multi-output (MIMO) controllers.

• Lithium-Ion Asset Safety: Proactive detection and avoidance of the onset of thermal runaway in all classes of lithium-ion batteries without redesigning the batteries or their chemistry.

True North Prognostics is advancing predictive technology by combining AI, data science, and engineering to create smarter, safer, and more sustainable systems for the industries that power our world.

© 2025 NVIDIA Corporation. NVIDIA and the NVIDIA logo are trademarks and/or registered trademarks of NVIDIA Corporation in the United States and other countries.