By Larry Tyson
Our team is directly involved in the reliability, availability, and maintainability (RAM) of a turbo fan jet engine. The engine fuel controller (FC) is one of the main components on the engine. Its purpose is to control the flow of fuel based on pilot demand from the throttle levers in the aircraft cockpit. If the FC does not operate properly, the engine does not receive fuel in accordance with pilot demand and may shut down. As can be imagined, shutting down an engine is not a good situation in flight.Improving the Reliability of a Turbofan Jet Engine by Using the Weibull Distribution for Failure Mode Analysis (WFMA)
NOTE: FM = FAILURE MODE, MAL CODE = MALFUNCTION CODE, they are the same
In recent years, our FC has had in-flight shutdowns due to flameouts and other problems. A flameout refers to the failure of a jet engine caused by the extinction of the flame in the combustion chamber. It can be caused by a number of factors, including fuel exhaustion, compressor stall, insufficient oxygen supply, foreign object damage (such as birds or hail), severe inclement weather, mechanical failure, and many other failures.
One root cause of these shutdowns and flameouts was the FC. In-flight-shutdowns result in in-flight aborts (IFAs). An IFA is a dangerous condition that puts the pilot, aircraft, and property at risk of damage. This flameout situation requires the pilot to defer quickly to emergency procedures and can lead to loss of aircraft; it is therefore a potentially costly failure that must be corrected. Between calendar years (CYs) 1999 and 2006, our team developed and published five technical directives (TDs) to correct the flameout problem. In late 2007, with most of the TDs completed, we used reliability growth analysis (RGA), part of the IEC-61164, to monitor the effect of the TDs, and we noticed a real change in the failure mode (FM). Instead of IFAs, we were starting to see a change to another FM. This additional FM was not previously known. This is because the flameouts took precedence due to their safety impact. This condition is called a “hidden failure.” A hidden failure is where one failure (such as In-Flight Shutdowns (IFS)), hides another FM, which in this case was leaks. This is a continual story in fielded systems: when one problem is corrected, another problem will arise. Analysis of the maintenance data illustrated that we had many FMs (in aviation these are called malfunction (MAL) codes). Previously, we did know about the leaks problem, but safety due to IFSs had a higher priority. These leaks were occurring more frequently, however, and no IFAs were involved. The lead engineer requested assistance in determining if the FMs were related. The Weibull process was used to determine interdependence of the various leaks. While accomplishing this engineering investigation, we developed a series of five pie charts that show the actual analysis progression. Information on Dr. Weibull and Weibull analysis may be obtained by searching for the term “Weibull” at google.com; there is a vast amount of information on this process.
In November 2007, Dr. Robert B. Abernethy (“Dr. Bob”) taught a 4-day Weibull Life Data Analysis class in Jacksonville, Florida. The instruction involved Weibull analysis and reliability growth management and analysis. Both of these are part of the International Electrotechnical Commission (IEC) publications. Outside of his reliability of life limited data analysis, Dr. Bob may not be known; however, Dr. Bob’s 11 patents are very well known and were obtained when he was employed by Pratt & Whitney (P&W) as an engine designer. In the mid 1950’s, the US Air Force and CIA approached Mr. Kelly Johnson (Lockheed Skunk Works), desiring a new aircraft that could fly very high and fast. This aircraft became known as the SR-71 Blackbird, and its power plant is the P&W J-58 that Dr. Bob and his team of engine designers and technicians developed. The SR-71 still holds all altitude and speed records for jet engines, and it has been 54+ years since the very first SR-71 flew.
So, how did Dr. Bob help us with our jet engine problems? He helped us to take 165 FMs we had observed and reduce them to a very manageable and defined set of failures. By utilizing Weibull distributions, we were able to make sense out of this large number of failures. This was accomplished by using the likelihood contour plot (LCP). If the contours intersected, then the FMs were from the same Weibull Family and their data was merged. This process continues until there are no more LCP intersections. To visualize the process and to see the reduction of FMs, we will use pie charts. Before we get to these charts, in the first iteration of this paper, we used the FM descriptions, and this proved to be complicated. Here we will use a 3-digit number, or an alpha character followed by 2 numbers that is associated with the description, to reduce the clutter.
We will look at a series of pie charts that show the results of using the Weibull distribution for failure mode analysis. The first chart is an eyesore, because it has all 165 FMs.
In Pie Chart 2, the 95 FMs that had fewer than 3 data points were combined to form mal code (95), which makes up 1% of the data. Please keep in mind that 3 data points is the minimum required for distribution analysis. This chart also combines those FMs that fall in the cannibalizations, administrative, and no trouble found categories; these will carry the number 800 and make up 18% of total data points.
Pie Chart 3 strips the mal code (95) and 800 series of FMs that are not real failures. We also have removed all cannibalization, administrative, and no trouble found FMs. This leaves us with 56 true failures. Our failure mode analysis starts here.
From the above 56 FMs, the detailed statistical and mathematical process of the Weibull distribution analysis begins. Some of these FMs will fail the outlier test, some will fail the distribution analysis and not be the required Weibull rank regression (Wrr) or Log normal rank regression (Lrr), and others will be the Weibull 3 parameter (W3P), which cannot be used.
Pie Chart 4 is the “bottom line.” It is the very reason to perform the Weibull Failure Mode Analysis (WFMA). The final 6 FMs started out of the original 165 FMs. The numbers associated with the FMs on the pie chart originated from the original FMs. That is, this FM is arrived at by adding the FMs that make up the below FM (e.g., FM 588 is derived from the following FMs : 037 - Fluctuates/Oscillates, 177 - Fuel Flow Incorrect, 374 - Internal Failure).
At this point, our lead FC engineer had a clear understanding of the problems. This enabled him to write a local engineering specification (LES) to require the replacement of a group of internal seals during depot-level maintenance. Previously, these seals were subjectively changed based on condition. Due to the results of this LES, we now have slightly over 735 (as of 12/2010) FCs with new seals (40%+ of total FC’s). A formal technical directive (TD) would take another 2 years to complete, but by this time the FCs will be 85%+ incorporated using the LES. This saves a great deal of time and money for the military. Another benefit we saw was the 1% reduction in cannibalizations that occurred as a direct result of this analysis. This makes the FC “stay on wing” longer. In other words, the FC is not failing as often, which results in it staying on the aircraft longer. This is called time-on-wing (TOW).
In closing, we think the Weibull analysis really helps a great deal on doing engineering analyses when numerous failure points are involved. We were able to take a complex situation, perform the analysis, filter or reduce those failures down to the most critical, and do a detailed analysis of those root cause failures. The Weibull analysis provides a powerful and useful tool for the engineering community that deserves more attention and use for these types of complex analysis.
This document may be used to follow the PowerPoint presentation by the same subject title. The presentation (35 pages) is a detailed step-by-step instruction and includes technical details of WFMA.
Larry Tyson, retired, has spent 24 years in the US Naval Service. Currently working for government service, his tasks include: support equipment specialist, involved in LIFE-CYCLE-COST ( LCC ), and reliability and maintainability (R&M) for avionics, support equipment, airframes, hydraulics, and power plants, both propeller and jets. Larry is involved in the TURBOFAN Community with a concentration on R&M analysis, RCM, and RCA.
Acknowledgment: Wes Fulton and Dr. Bob assisted with this article and the power point presentation; I thank them very much. Mr. James Young, Naval Surface Warfare Division, Crane, IN: his help was greatly appreciated in the formulation of this document.