Session 1234



Session 2502 TS-3

ANOTHER DESIGN CLASS TOPIC

Teaching the Student How to Evaluate An In-Service Design Using Reliability Engineering

Ronald O. Stearman*, Member ASEE & Monte C. Buschow**

Abstract

In response to the Air Force’s Reliability and Maintainability 2000 higher educational program, reliability studies were incorporated into a required senior-level design and testing course in the aerospace engineering curricula over 10 years ago. This included a one-week introductory statistics lecture series and a three hour statistics workshop. This workshop was followed by a two weeks introductory reliability lecture series plus one additional three-hour reliability workshop. Case studies are typically reviewed during the workshop periods. The students in this senior design class are also divided into groups of two to four after they choose a senior design topic they would like to work on for the semester. These projects are derived from recommendations within the industry whenever possible. A certain number of these design projects involve the evaluation and redesign of existing hardware that has proven to be unreliable as determined from their service records. The student group working on such a project is first assigned the task of evaluating this inservice design employing the reliability tools he has or will learn from the classroom lectures and workshop on reliability. The group is then advised to conduct an experiment or an analysis with the help of the instructor that will reveal the physical mechanism creating the problem. In a further attempt to illustrate the design evaluation process to the student, formal critical design reviews are conducted for all projects three times a semester employing both the oral and written modes of communication. More informal weekly conferences are also maintained with each group in addition to the regular lectures and workshops. In spite of these efforts the final design reviews and evaluations are often felt to be quite subjective to both the students and the faculty involved. In an attempt to make this design evaluation process more quantitative this instructor has tried to introduce to the students one quantitative evaluation tools derived from the reliability arena that helps to separate a poor design from those that are destined to be more successful. The students are also shown how to evaluate the growth of reliability through the redesign cycles of a product in service. Using concepts from reliability as a design evaluation tool first came to the instructors attention through its application by the Boeing company in solving a puzzling unreliable aircraft fly-by-wire controls problem for the Army. The techniques will be illustrated in this paper by reviewing two case studies conducted over the past 10 years within this senior design class.

* Bettie Margaret Smith Professor, Aerospace Engineering and Engineering Mechanics, The University of Texas at Austin - Instructor: The Design and Testing of Aerospace Structures - The University of Texas at Austin: ron@aeroel.ae.utexas.edu

** Staff Engineer - International Space Station Mechanisms and Maintenance Systems Engineer - Barrios Technology - Houston, Texas; mbuschow@

I. Introduction

No one disputes the need for articles to be reliable. The growth of consumer associations is a clear indicator of public concern over reliability and the less than perfect reliability in domestic products. Organizations such as the airlines, the military, and the civilian space agencies are aware of the high costs of unreliability. Consequently in recent years reliability has been placed on an equal footing with cost in contract award evaluations by both the DOD and NASA (1).

The renewed emphasis on design in our aerospace engineering curricula in recent years was prompted by both our Accrediting Board of Engineering and Technology (ABET) as well as by our past and present Industrial Visiting Committees. The result is an increased emphasis on the level of both detail and conceptual design content in our current program from freshman through the senior years.

Since the effects of unreliability in a design tend to be very costly to alleviate difficulties after the fact, and because unreliability can be closely coupled to safety, product liability, and possibly ethics, the inclusion of the reliability issue in our design course appeared beneficial in helping tie together some of these requested critical last semester issues. Emphasizing reliability at the early stages of the design process keeps product costs under control, enhances safety, and helps to avoid product liability. It was for the above reasons coupled with the added stimulus as outlined in the following background section that the topic of reliability was integrated into our senior design and testing course starting in the fall semester of 1987 at the recommendation of the US Air Force.

After teaching an element of reliability in our required senior design course for several years an additional benefit of this subject became evident. Although critical design reviews are conducted in our senior design classes in both the written format and in oral presentations before an engineering audience, the final design evaluation is often quite subjective. In some design cases, hardware is built and its successful performance is often attributed to an outstanding design project. On the other hand, if hardware is built during the project, reliability life cycle tests can lead to added information about the quality of the design. Some design projects involve design improvement or design case studies where a prior history of the design in service is available. In these cases a reliability study can be quite beneficial in determining the level of design improvement. Many times, however, the design student is not exposed to such tools and is often left without any quantitative evaluation procedures other than his subjective intuition and past experience. The present paper addresses this issue by proposing and demonstrating the use of a reliability tool as one additional aid for the student and instructor in their design evaluation. The concept of using reliability as a significant evaluation tool for in service designs was first suggested to the author by a study conducted by the Boeing Company in assisting the Army and Sikorsky Aircraft in an EMI problem with the Black Hawk fly-by-wire stabilator control system design. (2)

Unfortunately, senior design classes tend to become a catch-all or the last course in which important topics can be brought before the student prior to graduation. Such topics include practice in the written and oral communication skills, engineering ethics, safety considerations in design and in the workplace, and product liability to name a few. The universities attempt to introduce these topics to the student in addition to the basic tasks of teaching the student a little about the design process itself and how to interact with others as part of a team for problem solving. Although it's generally not within the scope of a one or two-semester design course to evolve a seasoned designer, the graduating senior should at least be aware of the design process even thought he or she may work in other areas of a company. Hopefully, the design courses will at least accomplish these tasks with the acquired reliability tools sharpening their ability to critically evaluate engineering products, and tying together issues such as safety, ethics, and product liability.

II. Background

The introduction of reliability considerations into undergraduate engineering curriculum is essential for meeting the needs of current and future customers. The need for greater attention to reliability issues was brought to the forefront on April 25, 1980 when the rescue attempt of forty-seven American hostages held by extremists in Iran failed. In debriefings that followed the aborted attempt, reliability, or the lack thereof, was identified as a major contributing factor to the failure. Figures 1 and 2 (courtesy of World Book - Yearbook 1981, John S. Marshall, Photography Director) illustrate the complete destruction of a significant part of the mission hardware at the rendezvous point without a single enemy weapon having been fired; the US equipment simply self-destructed.

.

Reliability in terms of readiness rates for the seven-year-old Sikorsky RH-53 helicopters that participated in the rescue attempt was estimated to be between 36% to 47% (3), (4). The initial complement of helicopters numbered eight. An absolute minimum of five helicopters was needed to complete the mission. Since it was estimated that at least one helicopter would fail to start during the critical launch phase of the mission, the plan was halted once three helicopters were determined to be non-operational. Given the maximum estimated reliability, a minimum of ten helicopters should have been used. However, the addition of two more helicopters would have required an additional C-130 Hercules aircraft for refueling and would result in an unacceptable increase in the likelihood of detection by Iran and the former Soviet Union. Choosing to ignore the reliability issue in the mission design resulted in an international embarrassment for the nation.

Other factors should be mentioned to be fair with regard to choice of the RH-53 helicopter for the mission. It was chosen due to a combination of factors not necessarily related to reliability. Excellent fuel range, in-flight refuel capability, aircraft availability, adaptability for shipboard operations and the fact that the Iranian military had six RH-53's in their military inventory all affected the decision to employ this helicopter (5). One conclusion that may be drawn, however, is that although reliability issues may have been considered, they certainly did not impact the final decision to use the RH-53. As a most unfortunate effect, eight American servicemen died in this mission due to a low speed collision between one of the RH-53's and a C-130 during repositioning for refueling.

An interesting antidotal story grew out of this experience suggesting that one of the key players in this internationally witnessed event was an alleged “Colonel Hershel Walker”, one of the Commanders in Charge of the operation. Reliability concerns stemming from the event made such an impact on him that several years later, while assigned to the Chairman of the Joint Chiefs of Staff, he initiated a reliability and maintainability program designed to take America's military forces into the year 2000. Appropriately enough, the program was simply called "R&M 2000". Recognizing the impact of academic America on the end-user products of the military-industrial complex, “Colonel Walker” was determined to integrate the concepts of R&M 2000 into the nation's undergraduate engineering design curricula. To this end an initial group of thirty-seven design professors of engineering from various universities around the country attended a Reliability Engineering Design Workshop given by the Air Force Institute of Technology's Air Institute at Wright-Patterson Air Force Base in Ohio between June 15 and July 10, 1987. The senior author of this paper was one of these first thirty-seven professors to attend the workshop. The objective of the workshop was the "Integration of reliability concepts into the contents of the participating scholar's design course materials" (6).

The daily format of the R&M 2000 educational workshop included a four-hour seminar/workshop in the morning with the principal focus on reliability design concepts, and a four-hour scholar practicum in the afternoon concentrating on the integration of reliability concepts into the scholar's own design course lecture notes, homework problems, assignments, and exams. An extensive set of notes was also provided to each scholar at the workshop. These notes were a collection of learning and teaching aids, some of which could be directly incorporated into almost any design class due to the interdisciplinary nature of this subject matter. The scholar was also encouraged to develop one or more reliability case studies that could be used as semester design class projects. Through this approach, even though the formal reliability lectures might not occupy more than two to four weeks of a design class semester content, the group of students choosing the reliability design project could keep the topic before the entire class throughout the whole semester. One of the major concerns in this reliability issue is in the selection of an interesting case study or group project every semester. Although a good reliability case study has not always been forth-coming every semester, two of the more successful case studies are presented in the following description of the undergraduate reliability curricula which also demonstrate to the student a quantitative tool for evaluating a design.

A further motivation for the engineering design professor to introduce reliability into an undergraduate design curricula, is illustrated in figures 3 and 4. These figures show the importance of identifying engineering design changes early in the design cycle to avoid the power law cost growth that will accompany these changes if made later on after the design matures moving into the production and operational service phase of its life. These engineering change cost charts, originally developed by the Martin Marietta Company, were called to our attention by Mr. Johnny Doo, Manager of Technical Engineering, Sino Swearingen Aircraft Company in San Antonio, Texas. Mr. Doo, who works with those of us in the Aerospace Design Program at The University of Texas, says the chart correlates closely with several years of his design experience with both Sino Swearingen and Fairchild Aircraft, where he was involved in both the development of Executive Business Jet aircraft and Commuter Liner Turboprop aircraft. Product liability has become a major cost factor since the chart was initially conceived, so the author has taken the liberty of extending the charts to include this cost factor as liability sometimes dictates design changes. Figure 4 illustrates a replot of the Figure 3 data in more quantitative terms to illustrate the power law growth of engineering design change costs as the design matures through its production and service life phases. This is certainly a motivation for the engineering design professor to emphasize an early reliability assessment of the evolving design to determine changes in the product as the design evolves.

Figure 3: Engineering Change costs at various levels of the design maturity normalized to a one dollar cost at the CAD stage, source: Martin Marietta (7)

[pic] Figure 4: Illustrating power law cost growth for engineering design changes made through the maturing phases of the design, source:

Martin Marietta (7)

III. Undergraduate Reliability Curricula

In the Department of Aerospace Engineering and Engineering Mechanics at The University of Texas at Austin, reliability concepts are introduced within the senior-level class titled "Design and Testing of Aerospace Structures". After a brief review of basic statistics, attention is turned to defining reliability in the context of a life cycle test. The test data is interpreted in terms of a time dependent probability or cumulative distribution function (CDF) and failure rate is defined as a probability density function (PDF). The concept of hazard rate is developed, and then the various distributions most commonly used in reliability analysis are examined. Generally, this discussion includes the binomial, Poisson, exponential, gamma, normal, lognormal, and Weibull distributions. The use of graph inference methods is introduced as a practical system identification tool, and students are given a data sample and asked to evaluate the reliability parameters. That is, the most probable reliability function is then determined for the sample using various types of supplied graph inference paper. The final reliability topic involves analysis of "sneak circuits" using examples found within the industry and their potential solutions. At the completion of this exercise, the student's level of familiarity with reliability theory is sufficient to introduce actual case histories for review and critical design analysis. Since reliability analysis can aid in the determination of design quality by evaluating component reliability, students may be tasked to evaluate a structure based on failure data from successive component design modifications to also assess the reliability growth of a design. The reliability approach to design evaluations in the classroom will be illustrated with two examples. One of the examples is the reliability case study developed by the senior author at the Air Force Workshop in 1987, while the second example evolved from one of our more successful and current design class projects studied by the second author.(8) In the former case a general aviation baggage door was evaluated from in-service records along with a simulated life cycle test on a baggage door. A wind tunnel model was also designed and tested to evaluate the impact of the failure on the aircraft stability and control. The second case study example refers to an engine mount truss structure redesigned over six times in the past 10-year history. This later study illustrates how design improvements (i.e. reliability growth) can be tracked through the redesign cycles utilizing a simple reliability design tool. (8)

Case I Design Study - The Opening Baggage Door:

This case study was developed by the senior author while attending the 1987 Air Force workshop designed to help achieve Reliability and Maintainability by the year 2000 through an educational thrust. The case study also provided the senior author with over five design project activities over the next three-year period. These projects used reliability evaluation tools along with laboratory and wind tunnel testing techniques to help understand or determine the physics promoting the unreliability.

Following the directions from the R&M 2000 workshop the students were introduced to this case study through a “request for proposal document” where they were considered to be the contractor seeking the proposal award. (9) The work statement section from the request for proposal which outlines the program objectives and program history are included below:

Request for Proposal on the Redesign Study of an Aircraft Nose Cone Baggage

Door and Locking Mechanism

Program Objectives:

The contractor shall conduct an engineering design evaluation of the current baggage door and locking mechanism, hereafter referred to as the baggage door system. The baggage door is located on the fuselage nose cone section of a twin engine general aviation type aircraft. The evaluation study shall include a reliability analysis of the existing baggage door design. A probable failure mode and/or system components shall be identified which contribute to the system unreliability. Upon completion of the existing baggage door system design evaluation, the contractor shall perform a redesign of the baggage door system. The contractor shall demonstrate through a reliability analysis and a reliability demonstration test that his redesign should lead to an improvement in the reliability of the system performance.

Performance History of Current Design:

A fifteen year history of maintenance and defect reports, two airworthiness directories issued during this period with little apparent success, and unexplained clear weather accidents of the twin engine aircraft have prompted the National Transportation Safety Board (NTSB) to issue a safety memorandum to the Chief of the Federal Aviation Agency (FAA) in the spring of 1987. This memorandum basically recommends a redesign of the Nose Cone baggage Door System. Several recommendations are also made in this memorandum as to the types of design modifications that should be accomplished. The contractor shall review the maintenance and defect reports, the technical NTSB accident summaries provided, and the NTSB memorandum to the FAA to obtain a more detailed history of the baggage door system design problem. The contractor should also review the oral statements of an FFA flight safety officer and practicing instructor pilot on the supplied cassette tape to understand a pilot's perspective who experienced this problem and survived. The contractor shall also use these referenced documents and those that are supplied by the contract monitor to assist his engineering design evaluation and redesign of the baggage door system.

[pic]

(Double-click to hear audio clip of an instructor

pilot’s in flight experience with the opening baggage door problem)

A review of the basic baggage door system configuration will be beneficial prior to a review of its design history. Figure 5 illustrates the basic nose cone baggage door configuration of the aircraft. The door is hinged at the top and opens upward.

[pic]

The nose cone wheel well extends into the baggage compartment but is not sealed to prevent air leakage from the nose gear wheel well into the baggage compartment. Details of the baggage door system subassembly linkage associated with the locking mechanism are illustrated in Figures 6a and 6b along with the link brace supporting the door in its open position.

A barrel type lock activated by a key adjacent to the door handle provides access to the baggage compartment. Additional details of the locking mechanism subassembly installation in the baggage door structure are illustrated in Figures 5, 7a and 7b.

One past design modification (increased) the length of the locking pin illustrated in the lower part of the sides of the door in figures 6 and 7. The locking mechanism can be easily removed from the baggage door by the removal of the one cover plate shown removed in the lower center section of Figures 6b and 7b. Finally, Figure 7b shows a sketch of the baggage door locking pin linkage mechanism, demonstrating an over-center kinematic locking effect.

Additional details providing the historical summary of the baggage door problem can be found in references 9 through 11.

Baggage Door System Design Evaluation:

This portion of the study had the benefit of a fifteen-year service history of the baggage door reliability problem. This service history was public domain information catalogued by the FAA and the Department of Transportation in Australia as maintenance and defect reports (MDRs) which are currently referred to as service difficulty reports (SDRs). A reliability graphical inference method was the tool to be employed for the design evaluation. The actual design evaluation is based upon estimating what the statistician refers to as the hazard rate function. Originally, this function was employed in the insurance industry for modeling the distribution of human mortality. The parameters of the modeling are those associated with the versatile Weibull distribution function.

The Weibull distribution or reliability function is usually expressed by the following equation for a two parameter model, where R is the reliability function, ‘t’ is the time in service, ( a shape function and ( a scale parameter related to the characteristic life of the system being modeled.

[pic] (1)

The corresponding failure rate function for this distribution is minus the time rate of change of the reliability function. This is given for the two-parameter Weibull distribution function of equation (1) by:

[pic] (2)

The hazard rate function is obtained by normalizing this failure rate, function f(t) by the instantaneous reliability R(t). The parameters to be identified from service records are ( and (.

In general reliability studies, failures are sometimes found to occur after a finite failure-free time (. This fact is used to modify the two parameter Weibull reliability function by adding this additional failure free time parameter ( as follows:

[pic] (3)

The factor ( is often referred to as the location parameter or the minimum life. (12) This Weibull modeling is well suited to a design evaluation study due to its flexibility to model any segment of the hazard rate function. This hazard rate function is often referred to as the “bathtub curve” as is seen from its general shape as depicted in figure 8.

[pic]

Figure 8: Qualitative behavior of the hazard rate function illustrating the fundamental types of failure modes(6).

The interpretation of the shape parameter ( is critical and helps to identify the presence of a particular failure mode and the quality of the design. The following section discusses the typical values and corresponding implications of various values for (: (12)

( < 1 implies infant mortality. This can be caused by a variety of factors, including design flaws, inadequate burn-in, inadequate stress screening, production flaws, misassembly, poor quality control, overhaul problems, and solid-state electronic failures.

( = 1 implies random failures. In this case, one can suspect maintenance errors, human errors, foreign object damage (FOD), “acts of God,” mixtures of data from two or more failure modes of different (, or system failures due to shock overloads. This also describes failures of systems composed of a large number of components where random failures across the components are quite likely.

1 < ( < 4 implies early wear-out. The most common causes of failures in this range are low cycle fatigue, corrosion, and erosion. This is also the area that characterizes most bearing failures.

( > 4 is typical of old age or rapid wear-out. Many manufactures use ( values in this range to specify and control the quality of vendor parts. Typical old-age failure modes include some forms of erosion and corrosion, brittle material failures such as found for ceramics, other material property failures, and some bearing failure modes. (12)

In its most basic form, the method used to investigate the various Weibull distribution constants involves the graphical inference method employing, for example, the use of Chartwell Weibull probability paper. This paper is scaled on the ordinate with a double natural logarithm scale that represents the cumulative percentage failure. Originally, the percent of cumulative failures was given by the mean rank method and the value for the ordinate was estimated using the following equation:

[pic] (4)

where i = index of the latest failed sample and

n = sample size.

The mean rank method was employed by Weibull in his original paper outlining his method of statistical reliability analysis (13). Appendix G of the new Weilbull Handbook (12) also reviews several other ranking methods. Although the student can carry out the system identification by hand using the various forms of Weibull graph paper, illustrated in appendix H of the new Weilbull Handbook (12) a relative recent development of WeibullSMITH™ Software has been developed for the PC, which is well suited for the design classroom(. The case studies conducted in this paper employed a current version of this software.

In the baggage door design evaluation study a decision was required as to whether a two-parameter or three-parameter model was needed. According to Abernethy (12) certain criteria should be met before implementing the three-parameter Weibull model:

1. The uncorrected Weibull plot should show concave downward curvature.

1. There should be a physical explanation of why failures cannot occur before time [pic]

1. A sample size [pic] of at least 15 samples (preferably 20) should be available.

2. The correlation coefficient [pic] should significantly increase and be above the critical correlation coefficient when using the three-parameter Weibull distribution.

Analysis of the Baggage Door Assembly reveals the following characteristics with respect to the stated criteria:

1. The uncorrected Weibull plot indicates a concave downward curvature.

1. The failures cannot occur before [pic].

1. The sample size [pic] is 25.

1. The correlation coefficient [pic] increases from 0.9502 to 0.9788. Although this is slightly below the critical coefficient 0.9792 calculated by the program WeibullSMITH™, calculation at seven significant figures reveals a difference of less than four-hundredths of one percent.

Additionally, the coefficient of determination [pic], another measure of goodness of fit, increases from 0.9029 in the two-parameter model to 0.9581 in the three-parameter model.

A summary of the results is included in the following table as derived from figures 9 and 10, representing a graphical inference method of identification of the Weibull parameters from the service data.

Table 1 - Summary of Weibull Analysis Results

|Parameter |Two-Parameter |Three-Parameter |

| |Weibull Model |Weibull Model |

|Sample Size [pic] |25 |25 |

|Minimum Life [pic] (hours) |0 |101.12 |

|Characteristic Life [pic] (hours) |798.69 |761.74 |

|Shape Parameter [pic] |1.4076 |0.9447 |

|Correlation Coefficient [pic] |0.9502 |0.9788 |

|Coefficient of Determination [pic] |0.9029 |0.9581 |

Accepting the better fit of the three-parameter Weibull model, we can now quantify our confidence in the value of the estimate of the shape parameter [pic]. From Figure 3.7 in O'Connor (14) , we can determine the factor [pic] necessary to evaluate these limits. For a sample size of 25 and a desired confidence of 95%, [pic].

The confidence limits are calculated using the relationships (14)

[pic]

[pic]

Therefore, we can say that we have 95% confidence that

[pic]

Returning to the value of 0.9447 for [pic], we can conclude that a high probability exists that the design demonstrates failure distribution characterized by infant mortality (early failure). According to Abernethy (12) , shape parameters less that one "lead us to suspect:

5. Inadequate burn-in or stress screening

6. Production problems, misassembly, quality control

7. Overhaul problems

8. Solid state electronic failures."

Similarly, Lewis (15) comments that "... defective pieces of equipment, prone to high failure because they were not manufactured or constructed properly, cause the high initial failure rates of engineering devices. Missing parts, substandard material batches, components that are out of tolerance, and damage in shipping are a few of the quality weaknesses that may cause excessive failure rates near the beginning of design life." Lewis continues "... the preferred method for eliminating such failures is thorough design and quality control measures ..."

The included two Weibull plots illustrated in figures 9 and 10 were generated with the service history data obtained from the FAA and the Department of Transportation in Australia. The first plot, included for comparison only, is based on the two-parameter Weibull model, while the second plot is based on the three-parameter Weibull model.

Error bounds are included that utilize Fisher's Matrix method. Nelson (16) provides a thorough discussion of this method. Although Johnson (17) preferred the beta-binomial approach, significant advantages are attained when using the Fisher's Matrix bounds, especially for moderate size samples. (12) Finally, the median rank method is utilized during the analysis.

The Weibull baggage door reliability study, just outlined, indicates a failure (door opening in flight) during the infant morality phase of its design life. Failures due to wear out, which are certainly more desirable, are not the general design failure mode. This represents a quantitative engineering validation of the appropriate stand taken by the NTSB in requesting in the Spring of 1987 that the FAA mandate a redesign of the aircraft baggage door.

[pic]

Figure 9: Two-Parameter Weibull Distribution - Baggage Door Problem

[pic] Figure 10: Three-Parameter Weibull Distribution - Baggage Door Design Problem

Life Cycle Test

The next phase of the group study invloved designing an appropriate life cycle laboratory test and/or further design analyses to understand the physical mechanisms promoting the unreliability. Consequently, this group designed a life cycle laboratory test on a nose cone and baggage door substructure as shown in Figure 11.

[pic]

Figure 11: Life cycle test set-up

Two shakers were employed to simulate propeller blade passage frequencies and the resulting pressure pulses on the baggage door and nose cone shell as shown in Figure 11. Approximately 25 hours of testing was involved in this study with no observed failures. During this time, however, the interesting observation was made that when the door wasn’t locked the simulated propeller aerodynamic pressure pulses and resulting door vibrations rotated the door latching handle in an oscillating manner to the open position.

This demonstrated to the students what a vibration environment could do to structural assemblies. The students were also exposed to this case of zero failures in a life cycle test and shown how the chi-squared function could be used to estimate a lower bound characteristic life for the baggage door’s time based upon its test duration. This is expressed by equation 6 where T would be the life cycle test time in hours, 100(1-() the upper confidence limit for ( and [pic]; 100(1-() is the upper 100(1-() of the chi-squared distribution with two degrees of freedom. The testing time of 25 hours, however, was too short to obtain a sharp lower bound estimate. Several door opening closing and locking cycles were also conducted manually which confirmed the NTSB findings that the baggage door handle could be closed and locked without the side locking pins ever engaging in their mating holes.

In essence, the locking pins and links had an inadequate stiffness allowing column buckling to the extent that the over center locking mechanism would function and the door could be locked after the pin linkages had buckled. Figure 12 illustrates just such an event that occurred on an aircraft whose baggage door was found locked but without the locking pins engaged with their mating sockets. Some lateral plate buckling of the circular rotating disk driving the pins was also observed.

[pic]

Figure 12: Damaged Locking Pin Mechanism

Wind Tunnel Test

A second group in the design class, also investigating the physics of this problem, focused on how the open baggage door influenced the aircraft stability and control. They obtained the plans for a 1/12 scale powered radio control model of the aircraft and built it to fly on a cable mount system. A description of the radio controlled model, its wind tunnel cable mounting system, its eleven channels of instrumentation, and a discovered vortex interference mechanism are all illustrated in Figure 13.

Figure 13 : Details of 1/12 scale radio control model flying on cable mount system in a subsonic wind tunnel

The start-up of the wind tunnel tests on the 1/12 scale radio controlled powered model with an opening nose cone baggage door are shown in the following Quicktime movie:

[pic]

(Double-click to see movie illustrating 1/12 scale radio controlled powered model in wind tunnel supported by means of a cable system simulating free flight conditions.)

The next Quicktime movie demonstrates an aircraft destabilizing mechanism caused by the baggage door tip vortex that develops when this door comes open. This tip vortex interacts with the horizontal stabilator as a lifting surface and may jump back and forth across the stabilator depending upon the trim condition of the stabilator. Our example from this second movie sequence shows the model undergoing a porposing oscillation thought to be caused by the switching of the vortex across the stabilator tail plane. The asymmetric canard configuration resulting from this opening baggage door undoubtedly also plays a role in this upsetting mechanism. A final sequence in this second Quicktime movie also suggests this vortex interaction phenomenon. A bead tied to a long string taped to the tip of the baggage door is shown spinning in this vortex core near the stabilator leading edge. A slight change in stabilator trim pulls the vortex core close to the bottom surface of the stabilator causing a sudden and uncontrolled pitch up of the model which hits the ceiling of the tunnel.

[pic]

(Double-click to see movie of two unstable wind tunnel model upsets induced by an opening baggage door.)

It is felt that this wind tunnel study of a powered radio controlled model flying on a cable support system with an opening baggage door provided some physical insights into the aircraft destabilizing mechanism.

Case II Design Study - Engine Truss Reliability Growth

As part of this group project, two students were tasked with evaluating a welded tublar truss engine mount currently in use on a popular turboprop commuter aircraft. The mount had been experiencing cracking, particularly in the vicinity of welded joints near the engine attachment points. The actual truss life was noted to be far below the intended and predicted design life of 20,000 hours. Data for the various damaged engine trusses were obtained from FAA Service Difficulty Reports. Consistent with extreme value studies, only first-time cracking failures were considered. That is, a repaired truss reentered into service that cracked again was not considered. Six truss types were identified in the preliminary investigation. Only four of the six were found to have statistical relevance to the study, as the remaining two truss types were only in limited use and sample data was insufficient. The various truss types were indicative of redesigns that evolved over a ten year period and were developed with the goal of extending the characteristic life of the truss. Modifications included increases in truss tube diameters and tube wall thicknesses along with the addition of gusset plates to the elements and joints noted to crack most often.

In the parameter identification the data were first plotted on exponential and lognormal graph inference paper as a first attempt to characterize the behavior of the failing trusses. When this attempt proved inconclusive, the students plotted the data on the more versatile Weibull graph inference paper. The resulting four plots (one for each of the four trusses included in the study) indicated an alarming trend - rather than increasing the characteristic life of the truss, the augmented strength in each redesign of the truss was actually decreasing the characteristic life. A summary of those results are depicted in Figure 14.

[pic]

Figure 14. Characteristic truss life [pic] as a function of truss type. Truss types are listed in chronological order from left to right.

Additionally, analysis of the shape function, a characterization of the type of failure rate, indicated that this important value was also decreasing and, as of the last design, indicated the possibility of failures in the infant mortality region of the "bathtub curve" (region I) or early failures as shown in Figure 8. Figure 15 illustrates the critical transition of the truss design cycles toward an infant mortality mode of failure. Thus the product redesign improvements did not improve anything but only led to an increased unreliability.

[pic]

Figure 15: Shape parameter [pic] as a function of truss type. Truss types are listed in chronological order from left to right. Upper and Lower bound indicate confidence limits of 95%.

Subsequent analysis of the data using the Duane model (12) (also used for determining reliability growth) also verified that the truss cracking problem was getting worse as the six design iterations progressed over the past ten years. Additional research indicated that the dislodgment and delimitation of engine vibration isolator mounts was another ongoing failure problem. Studies are currently underway in this area and seek to not only evaluate the reliability of the vibration isolation mounts, but to determine if a causal relationship exists between these three apparent failure modes. This research design class topic was extended to a master's thesis for one of the original undergraduate students, the second author of this paper.

The physical mechanism leading to the unreliability problem is thought to be due to the manner in which the engines are supported on this aircraft. That is, the engine is supported through compliant vibration isolator mounts attached to the engine truss structure. All of these four mounting points are located in a single steam wise plane approximately a foot behind the engine and propeller combined center of gravity. This mounting system allows a whirl degree of freedom of the engine and propeller system during a significant gust encounter or landing impact maneuver. This whirl motion is thought to be promoting the high degree of unreliability that appears to be evolving as the redesign cycles progress. That is, the design “improvements” are forcing the design toward to an infant mortality mode of design life. In essence, the designers are trying to solve the problem through strength considerations when stiffness (i.e. eliminating the whirl degree of freedom) must also be considered. A lack of good stiffness integrity can in turn lead to a whirl flutter instability as illustrated in these Quicktime movies.

[pic] [pic]

Side View Front View

(Double-click to play movie of stable whirl motion)

[pic] [pic]

Side View Front View

(Double-click to play movie of unstable whirl motion)

The current industry standard practice for mounting engines with high power ratings is through attachment points located in two streamwise stations and not all in one streamwise plane. This practice basically suppresses the whirl degree of freedom thought to be the physical mechanism leading to the decreasing reliability growth experienced by this design.

One such industry standard engine mounting example is illustrated in figure 16 which is the redesigned Lockheed Electra turbo prop structural mounting system. Arrows show the two streamwise mounting points and a yaw damper to suppress the lateral motions of the engine propeller system.

[pic]

IV. Graduate Level Reliability Curricula

Currently, a graduate level course in reliability is offered in the Department of Aerospace Engineering and Engineering Mechanics at The University of Texas at Austin. The course, titled Reliability Engineering, is offered as a special topic by the Structural Dynamics group approximately every two years. It begins by defining reliability and other key terms and examines the effects of reliability issues on economics and history. The course then provides a solid reintroduction to statistical mathematics before exploring methods of modeling and quantifying reliability. At this more advanced level, the course introduces predictive methods and system modeling. Reliability issues in design are examined to include probabilistic stress analysis, failure mode, effects and criticality analysis, fault tree analysis, and Monte Carlo simulations. Finally, case studies are examined. Since the classes are typically small in size, the case studies portion may be abbreviated and research of interest to the student may be included with the student reporting on his or her reliability research interests.

V. Conclusions

Topics in reliability have been taught as an integral part of the undergraduate structural design and testing course in aerospace engineering at The University of Texas at Austin over the past ten years. A graduate course in reliability has also been taught on a less frequent basis. During this time frame it was found that its implementation in the undergraduate design course did not require any significant dollar investment and also enhanced the instruction of topics related to safety, product liability, and ethics which are also essential to the understanding of the design process. The reliability tools provided to the course are versatile enough to aid in the determination and evaluation of a given design quality and reliability growth through successive design cycles when the design or designs have been in the field for a sufficient length of time to have established a service record. Reliability assessments of preliminary paper designs can also be carried out within certain estimated confidence levels using industry and government generated databases such as the Government and Industry Data Exchange Program (GIDEP). Finally, the reliability teaching experience at the senior design class level indicates that working real world reliability studies significantly stimulates student interest and performance in learning more about the critical review involved in the design process.

Acknowledgements

Appreciation is expressed to the many Aerospace Engineering senior design students, who, over the past ten years, have effectively pursued several engineering reliability projects, including those reported upon in this paper. The authors would also like to thank the Air Force reliability educators for introducing them to the important field of reliability engineering through the 1987 CERM workshop and the R&M 2000 educational thrust.

1. Goodell, Frank S. Brigadier General, U.S.A.F. Special Assistant for Reliability and Maintainability Headquarters, U.S. Air Force, “R&M By Design: A Blueprint For Success,” Journal Papers on Reliability and Maintainability, Journal of Aircraft, August 1987, pp. 481-483

2. Von Achen, William, “The Apache Helicopter: An EMI Case History,” Compliance Engineering, Fall 1991, pp. 11-17 & 111-112. (See section on The Black Hawk)

3. Griffiths, D. R., "Readiness Rate of RH-53 Key Issue," Aviation Week and Space Technology, vol. 112, no. 18, May 5, 1980, pp. 22-23.

4. Moses, R. L., Freeing the Hostages, University of Pittsburgh Press, pp. 183-190, 1996.

5. Fink, D. E., "Rescue Helicopters Drawn from Fleet," Aviation Week and Space Technology, vol. 112, no. 18, May 5, 1980, pp. 24-25.

6. CERM Reliability Engineering Design Workshop, Compendium of Workshop Notes and Materials. Reliability Training Institute, 1986.

7. Personal Communication with Johnny Doo, Manager, Technical Engineering, Sino Swearingen, and Adjunct Lecturer to the Aerospace Engineering Design Program, University of Texas at Austin - Martin Marietta data base found in older archival design journal.

8. Buschow, Monte, “A Study of the Reliability of In-Service Engine Mounts,” M.S. Thesis, The University of Texas at Austin, Spring 1997.

9. Stearman, R. "Request For Proposal On the Redesign Study of the Aircraft Nose Cone Baggage Door and Locking Mechanism," University of Texas at Austin, Aerospace Engineering Department report, Fall 1987.

10. Burnett, Jim (Chairman NTSB) National Transportation Safety Board- Official Safety recommendation To the FAA Chief Administrator Donald D. Engen, Spring 1987.

11. "Safeguard" The Aviation Consumer, June 15, 1987 pg. 6.

12. Abernethy, R. B., The New Weibull Handbook. Houston: Gulf Publishing Company, 1993.

13. Weibull, W. “A Statistical Distribution of Wide Applicability,” Journal of Applied Mechanics, 1951.

14. O'Connor, P. D. T., Practical Reliability Engineering. New York, Wiley & Sons, 1985.

15. Lewis, E. E., Introduction to Reliability Engineering. New York, Wiley & Sons, 1994.

16. Nelson, W., Applied Life Data Analysis, John Wiley, 1982.

17. Johnson, L. G., The Statistical Treatment of Fatigue Experiments. Elsevier Publishing Company, 1964.

-----------------------

[pic]

Figure 1: Iranian military personnel inspect the remains of a C-130 Hercules transport.

[pic]

Figure 2: Iranian officials and army officers inspect

the wreckage of US military equipment left in the

desert after the abortive hostage-rescue mission.

$1,000,000 +

At CAD Terminaal

At Design Checking

During Design Review

During Pre-production

During Production

In the Field

For Product Liability

$1

$10

$100

$1000

$10,000

$100,000

100

101

102

103

104

105

106

Figure 5: Nose Cone and Baggage Door Under Study

[pic]

[pic]

Figure 6a:

Baggage Door Locking Mechanism and link brace supporting door in open position

Figure 6b:

Baggage Door Locking Mechanism with cover plate removed

Figure 7b: Inside view of baggage door with cover plate removed, showing over center locking mechanism

Figure 7a:

Additional Details on Baggage Door Locking Mechanism

[pic]

[pic]

( This software is distributed through the SAE or ASME Societies or the Gulf Publishing Company, Houston, TX.

(5)

(6)

[pic]

[pic]

load cell

exterior to tunnel

precision weights

[pic]

yaw

roll

pitch

(a) Cable mount location

(b) Rate Gyros location axes

rudder servo

[pic]

flaps servo

[pic]

baggage

door

servo

stabilator servo

aileron servo

(d) Cable mount location

(c) Servo Locations

Ford Engine Mount support

AFT lateral engine support and damper

AFT Vertical Engine Support

Figure 16.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download