Prognostic modelling options for remaining useful life estimation by industry

Over recent years a significant amount of research has been undertaken to develop prognostic models that can be used to predict the remaining useful life of engineering assets. Implementations by industry have only had limited success.

By design, models are subject to specific assumptions and approximations, some of which are mathematical, while others relate to practical implementation issues such as the amount of data required to validate and verify a proposed model. Therefore, appropriate model selection for successful practical implementation requires not only a mathematical understanding of each model type, but also an appreciation of how a particular business intends to utilise a model and its outputs.

This paper discusses business issues that need to be considered when selecting an appropriate modelling approach for trial. It also presents classification tables and process flow diagrams to assist industry and research personnel select appropriate prognostic models for predicting the remaining useful life of engineering assets within their specific business environment. The paper then explores the strengths and weaknesses of the main prognostics model classes to establish what makes them better suited to certain applications than to others and summarises how each have been applied to engineering prognostics. Consequently, this paper should provide a starting point for young researchers first considering options for remaining useful life prediction. The models described in this paper are Knowledge-based (expert and fuzzy), Life expectancy (stochastic and statistical), Artificial Neural Networks, and Physical models.

*Integrated Systems Health Management includes fault detection, fault diagnosis (or fault isolation), and fault prognosis. We define prognosis to be detecting the precursors of a failure, and predicting how much time remains before a likely failure. Algorithms that use the data -driven approach to prognosis learn models directly from the data, rather than u sing a hand -built model based o n human experti se. This paper surveys past work in the data driven approach to prognosis. It also includes related work in data -driven fault detection and diagnosis, and in model -based diagnosis and prognosis, particularly as applied to space systems.

https://ti.arc.nasa.gov/m/profile/schwabac/AIAA-39300-874.pdf

A Survey of Data -Driven Prognostics

Integrated Systems Health Management includes as key elements fault detection, fault diagnostics, and failure prognostics. Whereas fault detection and diagnostics have been the subject of considerable emphasis in the Artificial Intelligence (AI) community in the past, prognostics has not enjoyed the same attention. The reason for this lack of attention is in part because prognostics as a discipline has only recently been recognized as a game-changing technology that can push the boundary of systems health management. This paper provides a survey of AI techniques applied to prognostics. The paper is an update to our previously published survey of data-driven prognostics.

A Survey of Artificial Intelligence for Prognostics.

http://www.123seminarsonly.com/Seminar-Reports/049/75568950-Gas-Turbine.pdf

Gas turbine performance prognostic for condition-based maintenance

A key to producing reliable engine diagnostics and prognostics resides in fusion of multisensor data. It is believed that faults will manifest effects in a variety of sensors. By 'integration' (fusion) of information across sensors detections can be made of faults that are undetectable on just a single sensor. Data to support development of prognostic techniques is very rare. The development requires continuous collection of significant amounts of data to capture not only "normal" data but also capture potential fault event data well before the fault is detected by existing techniques, as well as capture data related to rare events. The collected data can be analyzed to develop processing tailored to new events and to continuously update algorithms so as to improve detection and classification performance and reduce false alarms. IAC in collaboration with the Air Force and the Army is developing a testbed to perform data collection and to develop fusion techniques for gas turbine engine health monitoring. The testbed and examples of its operation are presented here.

A testbed for data fusion for engine diagnostics and prognostics

The US Army and South Carolina Army National Guard have developed an automated vibration monitoring and fault diagnostic tool as part of the Vibration Management Enhancement Program (VMEP). This paper outlines the VMEP system rationale, design and the vibration diagnostic algorithms used to find faults in the helicopter rotors, engines, and drive train. The concepts that differentiate VMEP from traditional Health and Usage Monitoring Systems (HUMS) implementations are described. Specifics of the system implemented in the Apache and Black Hawk helicopters found in the National Guard are shown. The software architecture used in the on-board component, the ground-based component, and the development station is presented. The data collection effort during the initial field trials is described and examples of vibration test data and algorithm processing results are presented. We believe that the system presented here represents a way of meeting many of the integrated vibration diagnostics goals for Army Aviation.

Automated Helicopter Vibration Diagnostics for the US Army and National Guard

: Automated systems to perform aircraft diagnostics and prognostics are of current interest. Development of those systems requires large amounts of data (collection, monitoring, manipulation) to capture and characterize normal, known fault events, and to ensure data is captured early on in a fault progression to support prognostic system development. Continuous data collection is also required to capture relatively rare events. Data collected can then be analyzed to assist in the development of automated systems and for continuous updating of algorithms to improve detection, classification, and prognostic performance. IAC, in collaboration with the Air Force and Army, is developing a testbed on which to perform data collection, and develop diagnostic and prognostic processing techniques using Army helicopter vibration and engine performance data as part of the Army's Vibration Management Enhancement Program (VMEP). VMEP and the testbed being developed for collection and processing of VMEP data are described here.

A testbed for data fusion for helicopter diagnostics and prognostics

A method for health monitoring for an aircraft includes the steps of obtaining vibration data for the aircraft, obtaining navigation data for the aircraft, and determining a measure of health of the aircraft using the vibration data and the navigation data.

Methods and systems for health monitoring for aircraft

Helicopter rotor smoothing (track and balance) is a periodic maintenance task required to minimize rotor induced aircraft vibrations at the fundamental (once per revolution) rotor frequency. We have designed and implemented a general, neural network based software system for rotor smoothing. Neural networks provide non-parametric mappings between the spaces of adjustments and vibration measurements. In the network training process, these mappings are extracted from experimental data without any assumptions about their functional form. On the other hand, when the experimental data available for training is not complete, simulated data based on model dependencies (linear or other) may be also incorporated into the neural network model. The neural networks are easily updated (retrained) if new data becomes available thus allowing the system to evolve and mature in the course of its use. The customization of the system for helicopters of different types is facilitated by general-purpose software for application development, which includes preparation of flight data and neural network training. It is worth noting that the prototype applications developed up to date required relatively modest amount of flight data (20-30 flights). The neural network system has been applied to Apache (AH-64), Blackhawk (UH-60), and Kiowa Warrior (OH-58D) helicopters as part of the Vibration Management Enhancement Program (VMEP). Preliminary results are very encouraging. In the verification tests, we were able to shorten the smoothing time for AH-64, with all of the rotor smoothing procedures completed in 2 to 4 flights. In all cases, the neural network approach produced solutions with experimentally verified low vibration levels and small track split. The system has also demonstrated the ability to detect errors in implementation of the smoothing adjustments. Application to other types of helicopters is considered.

P. Grabill

Papers

A testbed for data fusion for engine diagnostics and prognostics

Automated Helicopter Vibration Diagnostics for the US Army and National Guard

A testbed for data fusion for helicopter diagnostics and prognostics

Methods and systems for health monitoring for aircraft

Neural network system for helicopter rotor smoothing