TNP AI-MSET for Commercial and Defense Aviation Landing Gear

---

Landing Gear challenges can be solved using TNP’s mature and well-proven AI-MSET to predict when mission-critical and safety-relevant mechanical, electromechanical, hydraulic and servo-motor subsystems comprising Landing Gear assets are developing incipient defect modes so that small issues can be proactively remediated between flights, and well before serious asset anomalies turn into costly and sometimes dangerous equipment failure. 

TNP’s AI-MSET provides real-time surveillance of all physical transducers: voltages, currents, RPMs for rotating elements, pressures for hydraulic components, vibration prognostics from uni-axis and tri-axis accelerometers, along with all available ambient environment parameters. In addition to real-time surveillance of all sensors associated with landing gear assets, TNP’s AI-MSET brings significant advantages to offline inspections (between flights) for high-frequency non-destructive testing (NDT) techniques used for avionics landing gear assets, including Pixelated 2D IR Thermal Imagery.

        Solution to Landing Gear prognostic challenges via TNP’s mature and well-proven AI-MSET

TNP can predict when mission-critical and safety-relevant mechanical, electromechanical, hydraulic, and servo-motor subsystems comprising Landing Gear assets are developing incipient defect modes so that small issues can be proactively remediated between flights, and well before serious asset anomalies turn into costly and sometimes dangerous equipment failure. 

TNP’s 3rd-generation Multivariate State Estimation Technique, AI-MSET, provides real-time surveillance of all physical transducers: voltages, currents, RPMs for rotating elements, pressures for hydraulic components, vibration prognostics from uni-axis and tri-axis accelerometers, along with all available ambient environment parameters.  In addition to real-time surveillance of all sensors associated with landing gear assets, TNP’s APR brings significant advantages to offline inspections (performed between flights) for high-frequency nondestructive testing (NDT) techniques used for avionics landing gear assets, including Pixelated 2D IR Thermal Imagery (Separately patented TNP techniques for enhanced defect-identification with ultra-low false- and missed-alarm probabilities).

Commercial Aviation Benefits from TNP’s solution framework:

• AI-MSET is a 3rd-generation disruptive technology that enables continuous system telemetry of complex engineering assets and advanced AI pattern recognition techniques that provide the opportunity to perform the functions of prognostic health management (PHM), disturbance prediction, system diagnostics, and incipient fault characterization in an integrated package called Electronic Prognostics (EP). 
• EP provides continuous real-time information on the operational status of components, sensors, programmable logic controllers (PLCs), interconnects, and integrated Line-Replaceable-Units (LRUs) for commercial and defense Landing Gear subsystems. 
• A separate EP function called “sensor operability validation” checks and continuously validates the health of all existing sensors, identifies and replaces any defective sensor with a highly precise “virtual” sensor based upon an analytical estimate, while APR's predictive and prescriptive maintenance capability proactively detects and identifies incipient component degradation so that tiny developing mechanical, electromechanical, or thermal-hydraulic anomalies can be remediated during scheduled preventative maintenance windows, well before turning into serious anomalies while avionic assets are in mission-critical takeoff/landing operations where overall safety margins may be compromised.

Bottom Line – These benefits translate to better Reliability, Availability, and Serviceability (RAS) of mission-critical systems, better operational risk mitigation, increased safety margins, and much lower Operations & Maintenance (O&M) costs for the carriers.

AI-MSET: DISRUPTIVE TECHNOLOGY BACKGROUND

TNP and AI-MSET -- The motivation behind TNP’s development of advanced real-time prognostic machine-learning algorithms is high Reliability, Availability, and proactive Serviceability (RAS) of mission-critical and safety-critical infrastructures. 

Definitions:

AI-MSET = AI-enabled Multivariate State Estimation Technique

SPRT = Sequential Probability Ratio Test

AI-MSET & SPRT are the industry’s leading Prognostic Algorithms– AI-MSET & SPRT artificial intelligence (AI) algorithms predict in real-time what each process metric should be on the basis of learned correlations among process variables from subsystems with dozens of sensors, large-scale assets with thousands of sensors, or fleets of assets involving millions of sensors. TNP's AI-MSET Prognostic innovations combine advanced pattern recognition methodology integrated with SPRT possessing independently configurable false-and missed-alarm probabilities (FAPs/MAPs), to simultaneously achieve high sensitivity for proactive detection of subtle anomalies in noisy process variables, but with ultra-low false-alarm and missed-alarm probabilities. (FAPs and MAPs)

TNP’s Prognostic Algorithmic Innovations combine Two Powerful Pattern-Recognition Techniques:

Sequential Probability Ratio Test (SPRT) Advanced pattern recognition technique for high sensitivity, high reliability sensor and equipment operability surveillance. Developers proved in refereed journals and international symposia that the SPRT provides the earliest mathematically possible annunciation of a subtle fault in noisy process variables. Crucial capability for Landing Gear critical-asset health monitoring: Ultra-low and separately specifiable false-alarm and missed-alarm probabilities (Type-I and Type-II errors).

Throughout both commercial and defense aviation, False Alarms are extremely costly because they result in taking revenue-generating and mission-critical assets out of service unnecessarily, while missed alarms can be catastrophic. AI-MSET with SPRT continuously disambiguates between sensor anomalies and anomalies in the monitored assets.

AI-MSET:  Online model-based fault detection and identification. AI-MSET predicts in real time what each process metric should be on the basis of learned correlations among all process variables. AI-MSET incorporates the SPRT to monitor the residuals between the actual observations and the estimates AI-MSET predicts on the basis of the correlated variables.

TNP has significantly advanced the state of the art versus original MSET1 and now offers the next generation of predictive health and prognostics which we call AI-MSET.

AI-MSET combines multiple Intelligent Data Preprocessing (IDP) pattern-recognition and autonomous training/tuning/optimization algorithms, so that end-customer aviation carriers don't have to have PhD data scientists to set up and operate the Landing Gear prognostic analytics. 

TNP has developed a dense portfolio of patented innovations that leverage AI-MSET (as a core algorithm) integrated with various pre-processing, postprocessing, and optimal training/tuning algorithms so that TNP's prognostic solutions are more robust to low resolution sensors, data acquisition bandwidth limitations, missing values in time series signatures, intermittent spurious anomalies, and signal-asynchrony issues in large-scale IoT applications. AI-MSET attains higher sensitivity and better false alarm avoidance than any alternative AI Machine Learning (AI/ML) approaches, including neural networks (NNs), support vector machines (SVMs), and kernel regression. 

Traditional Threshold-Based Surveillance is Inadequate for IoT Prognostic Applications-

Traditional threshold-based prognostic approaches may use Machine Learning to distill down and coalesce important metrics for distinguishing between “normal” and “anomalous” behavior, but ultimately, the measured or derived metrics are being compared against a threshold:

• The endemic problem with threshold limit tests is the “sea saw” effect between false alarms and missed alarms.
• If the user wants to get earlier warnings for developing problems and “squeezes” the thresholds closer to the means, spurious trips result and elevated false alarm rates.
• If the user wants to avoid costly false alarms and moves the thresholds further away from the distribution means, then the assets can be severely degraded (or failed/crashed) before any alerts are generated.

TNP Disruptive Technology As Applied to Landing Gear Assets

US FAA “Grand Challenge”: 

It is difficult to predict when mission-critical landing gear technology will fail. Significant changes in the performance and/or availability of operating electronic components, interconnects, and mechanical, electromechanical, and thermal-hydraulic components may occur as a result of periods of intensive use, extremes, or cycles in environmental factors (e.g., temperature, humidity, shock, vibration), or system aging. Conventional surveillance methods that evolved from mechanical-asset prognostic health management (PHM) cannot fully allow for the effects of these factors in the determination of electronic equipment performance tolerances or test limits. This can result in apparent and actual decreases in landing gear equipment readiness and remaining useful life (RUL). Traditional PHM approaches that rely on conventional threshold-limit tests do not provide sufficiently early detection of symptoms of developing problems to permit the carriers to realize the benefits of truly predictive surveillance or condition-based and preventive maintenance. 

Discussion: 

Prognostics and Health Management (PHM) solutions for mission-critical and safety-critical assets require a comprehensive methodology for proactively detecting and isolating failure mechanisms, avoiding Type I and Type II errors (false- and missed-alarms), and estimating in real time the remaining useful life (RUL) of critical electronic components and associated subsystems. TNP's mature and well-proven AI-MSET "Electronic Prognostics" (EP) leverages AI-MSET’s capabilities to extend the envelope of PHM solutions to include the electronic systems where “thermal events” most often originate, including embedded Command-and-Control (C&C) computing elements, ASICs, programmable-logic-controllers (PLCs), actuators, and interconnecting networks systems.  No competitive ML algorithms can proactively detect the incipience or onset of thermal-dissipation mechanisms in the embedded C&C elements for mission-critical and safety-critical aviation assets.  AI-MSET has over four dozen patented innovations in the EP portfolio for determining electronic component health, performance tolerances, and quantitative Remaining Useful Life (RUL) estimation. To assure availability, flight readiness, and continued safe and reliable operation of complex electronic, power, lighting, entertainment systems, and propulsion systems, it is essential that accurate real-time information on the current state of the entire system and all its sensors and embedded C&C components and interconnecting networks be available to PHM autonomous control actuators. Such information is needed to determine the operability of control systems, the condition of active electronic components, and the status of sensory systems and other safety and operationally important subsystems and components.

A second vital feature of the AI-MSET technique that distinguishes it from conventional ML methods is that it has built‑in quantitative false‑alarm and missed‑alarm probabilities. This is important in the context of the present Landing Gear integration project, because it makes it possible to apply formal reliability-analysis methods to the customized surveillance system monitoring a variety of landing gear subsystem internal electronic components. 

Summary:

TNP algorithms and modular (pluggable) software provide:

• Real‑time signal validation for landing gear sensor electronic prognostics
• Fault isolation to identify the faulted component with persistent log information that isolates faults to the component level, the latter functionality will be essential to reduce or eliminate false alarms through enhanced root cause analysis (RCA) (this capability translates to better capability to identify and address hard-to-find faults and, in turn, reduces maintenance labor costs). 
• Incipient fault annunciation (helps the carrier identify and deal with intermittent faults, which are unpredictable and can skew data). 
• Remaining useful life (RUL) estimation with quantitative confidence factors (helps the Aviation Carrier make informed maintenance decisions – e.g., “Critical Issue: this must be fixed before deployment,” versus “Early Warning of Incipient Fault: AI-MSET RUL estimates that this issue should be addressed in the next 425 aircraft operational hours”). 
• Analytical signal replacement to enable continuing operations with failed or degraded sensors; (allows for AI to smooth data and, in turn, improve decisions influenced by data). 
• Automated sensor calibration monitoring for on‑condition maintenance. (allows for correcting the sensor output until such time as the sensor can be either recalibrated or replaced, which in turn buys the Aviation Carrier cycle time until the sensor can be replaced). TNP’s highly accurate real-time signal validation and sensor operability validation algorithm has already been approved for safety-critical applications, such as US Nuclear Regulatory Commission (NRC) approval for use in commercial nuclear reactors.