"Early Warnings: Diagnostics and Prognostics in Aero Engines," ME Online Web Exclusive, Jan. 24, 2005

for 1/24/05

Early Warnings — Diagnostics and Prognostics in Aero Engines

Raising the Level of Aviation Safety Through Non-Intrusive Detection of Cracks in Blades and Disks

by Mike Drumm

The current state of aero engine design in the United States is unparalleled. Engines are more powerful, lighter, more durable, efficient and typically more cost-effective than ever before.

This state is attributable to the quality of engineering talent at the engine manufacturers and their suppliers, and to the U.S. government's relentless, strategic drive to have the best engine technology powering American warplanes. Air propulsion superiority, both military and commercial, is not just a buzzword. It's a sustainable goal that over the long term becomes a way of life to many in the aero propulsion business.

But making anything better, faster, and cheaper, pushing technologies to their practical extremes, and developing new methods and materials to leapfrog to new capabilities always increase the uncertainty of an outcome. New materials (e.g., anisotropic substances like composites) may be stronger and lighter, but many times they cannot be characterized or modeled with existing codes, and require new methods to qualify them for safe use in flight.

Making fans, compressors, and turbines lighter to shed weight, and using materials with high strength-to-weight ratios sometimes induce high vibrations with complicated resonances. These vibrations can cause blades to have large deflections at a broad range of frequencies, resulting in blade cracking and sometimes in-flight failures. Stressing new designs to their limits, especially at high temperatures, can cause cracks in disks that are difficult to detect while the engine is at rest, eventually resulting in uncontained failures (i.e., the engine goes ka-boom and pieces of metal come flying out of the casing).

To prevent these potentially life-threatening situations from occurring, and to help lower the operating and maintenance costs of fleets, designers and planners have begun to look into key enabling technologies that provide not only early warning of potential disasters, but are also substantive indicators of the real-time health of engine components and the overall system.

Diagnostics and prognostics are the terms used to describe these technologies.

Diagnostics uses hardware and software to monitor rotating systems for abnormal behavior, while looking for faults and changes in normal patterns. A diagnostic system succeeds when it declares a fault that is not a false alarm.

The fact that a fault has been declared, however, does not necessarily mean that immediate action need be taken. That is where prognostics kick in. Prognostic software combines inputs from multiple sensors, operating data, and failure histories in an attempt to predict the life of a component, and to determine what maintenance is required and when.

A blade slightly damaged by some foreign object entering the engine may be deformed, but if it is not affecting performance to any great degree and if a crack is not growing, replacement might be safely postponed to a regularly scheduled maintenance shutdown.

Many existing health monitoring systems use fault detection and diagnostic techniques, but are not focused on automating prediction of future machine conditions. This prognostic approach is based on the current state of the component or machine, and on available operating and failure history data.

An example of operational prognostics is a system developed by Impact Technologies in Rochester, N.Y., to determine the optimal water wash interval to mitigate gas turbine compressor performance degradation due to salt deposit ingestion. The system uses advanced probabilistic modeling and analysis technologies to forecast the future performance characteristics of the compressor to yield the optimal Time To Wash (TTW), incorporating both cost and performance into the equation.

The technology is described in a 2003 paper, "Prognostic Enhancements to Gas Turbine Diagnostic Systems," written by Michael J. Roemer of Impact Technologies and several others involved in the project, and published by the Institute of Electrical and Electronics Engineers.

Detecting cracks in disks and blades, or monitoring other mechanical anomalies in rotating systems requires specialized equipment, previously used only in test cells and during development programs. Recent advances in measuring blade stress non-intrusively and developing high-temperature sensors that can measure blade vibration from outside an engine casing have brought to the forefront the possibility of using these systems in an operating engine. Combining these technologies with software for prognostic engine health maintenance has created an opportunity to increase safety, enhance performance, and lower maintenance costs of engine systems.

Non-Intrusive Stress Measurement

To understand the actual stress on a blade or disk used to require placing a strain gauge on a component, running thin wires onto the part, up the rotating shaft, and out through a device called a slip ring. This technique was fraught with problems: Gauges came flying off at high speeds; the gauges had significant temperature limits; each gauge could measure strain only in one direction on one blade; the slip ring had a limited life and needed to be cooled with refrigerant (which raised environmental issues), and, most important, slip rings and strain gauges were not reliable enough to be used in operating engines to continuously monitor component health.

To solve these problems non-intrusive stress measurement systems were developed that could measure blade deflection by detecting the time-of-arrival of every blade on every revolution. Depending on the number of sensors used and their location, older non-intrusive systems could measure blade deflection (vibration) circumferentially, and newer systems like those from Hood Technology of Hood River, Ore., were extended to additionally measure radial and axial deflection.

These systems can report the actual stresses seen on the blades; detect blade flutter and rotating stall; measure the effects of foreign object damage; detect synchronous resonances that are the source of high-cycle fatigue in blades; measure disk deformation and unusual blade stretch that are the precursors of low-cycle fatigue cracks in disks; measure individual blade mistuning and even help separate complicated, coupled resonances in bladed disks.

Many experienced testers are emotionally tied to strain gauges and use them to "calibrate" or verify data from non-intrusive systems. Equivalent analytical work is required for both strain gauge and non-intrusive measurements to ultimately produce details of blade stress; but an advantage of non-intrusive systems is that one probe can take data from every blade on every revolution, and some sensors can even be outside the engine casing. These non-intrusive measurements require no interruption at all in the gas path, and remove the sensor itself from the major environmental effects of the engine: temperature, foreign objects, blade rub, moisture, corrosion, etc.

Non-Intrusive Sensors

Optical sensors were the early probe of choice for non-intrusive systems because of their ability to withstand high temperatures, and because of the resolution associated with time-of-arrival measurements using light. Eddy current probes have been developed that have the same resolution as light probes, can measure both blade time of arrival and tip clearance in the presence of contaminants like oil, and can also be used at the same temperatures as light probes. In addition, some types of eddy current sensors can be used to measure through the case of a jet engine, instead of requiring an uninterrupted line of sight to the target, like capacitive and optical probes.

The simplest eddy current sensors generate a static magnetic B field from a permanent or electro-magnet that penetrates most casings, and induces eddy currents in passing blade tips. These eddy currents generate their own magnetic fields, and because of the flux exclusion principle, temporarily exclude the magnetic flux from penetrating the blades. This causes a time rate of change in the primary magnetic field, which is detected by a coil in the sensor. This allows the probe to make high-resolution measurements of blade passings from outside the case. Both optical and eddy current sensors can be made in the same form factor and have been used by Hood Tech's non-intrusive measurement system to make concurrent, real-time measurements of blade time of arrival and tip-clearance.

Figures 2 and 3 show a comparison between measurements taken with eddy current sensors and strain gauge data acquired on the same component. The eddy current sensors were measuring through 2.5 mm of case material.

Disk cracks can also be detected using blade-tip sensors and a non-intrusive system. If an FEA model of the cracked disk is available, the confidence of crack identification can be greatly increased with knowledge of expected disk deformation patterns. Figure 4 is an example of a disk deformation pattern predicted by a linear-elastic finite element model that was used to predict crack growth in a seeded fault engine test.

To simplify this analysis technique, an assumption can be made that the deflection pattern changes in amplitude but not in shape as the crack grows larger. In this way, a single calculated deflection shape can be fit to the data, yielding the best estimate of maximum blade tip deflection. With knowledge of the relationship between maximum blade tip deflection and crack size, the crack size is directly estimated.

In looking for disk cracks in a rotating engine or in a spin pit, many times radial, axial, and circumferential vibration are all measured. The system searches for asymmetric disk deformations caused by growing cracks. Comparison of the measured deformations with precalculated predictions enables the system to discriminate among several hypothesized crack types.

Sensitivity to cracks in the disk face, in the inner diameter bore, and in a blade retention slot have all been demonstrated. A paper, "Low Cycle Fatigue Rotor Burst Prognostics Using Blade Tip Sensors," published on the ExSell Web site, www.exsellinc.com, describes the technology.

When investigating some blade cracks, the system monitors the resonant frequencies and mode shapes of blade resonances to detect possible changes. Cracks in different areas of a blade produce different responses, so it is sometimes possible to characterize a crack by detecting a change in the signature of the blade vibration.

Integrating Elements

Through-the-case sensors, driven by compact high-speed electronics, can now make measurements of blade deflection and clearance in real-time, and convert these readings into information that describes the health of a component. Blade stress, blade cracks, disk cracks, flutter, rotating stall, foreign object damage, high-cycle fatigue, and low-cycle fatigue can all be identified and tracked using current hardware and sensor systems.

Coupled with finite element models, operational data and failure histories, algorithms and software can be developed that both diagnose a problem, and predict the failure of a component, allowing a maintenance system to be developed that optimizes safety, cost, and performance. New generations of engines that use these capabilities will have increasingly robust diagnostic and prognostic systems that work together to raise the level of aviation safety even higher.

Mike Drumm is president of ExSell inc., a Concord, Calif., consulting company in the area of diagnostics and prognostics for aircraft engines and power generation systems.