As I looked at the images of item # 9618 the MLG up-lock roller had three stresses worthy of examination for failure analysis:

• yoke bolt shear stress
• bearing stresses on threaded cylindrical section of fitting and hex bolt
• stress on door frame attachment to the hex head bolt

No stress failure seemed to have occurred to item # 9618. However, the absence of significant contamination from the slag transition line to the end of the bolt hex head suggests the possibility that tension forced the MLG up-lock roller away from the attachment connecting door frame to hex bolt. This should be an important concern for non-destructive examination (NDE) since pulling a plug out of the door frame should be the most unacceptable failure scenario related to the three stresses.

For the sake of argument distortion-energy theory is assumed: S_sy = 0.577S_y

Calculating allowable tension load ( F ) on the assembly according to yoke bolt shear stress:

S_sy = 0.577S_y = (0.577)x(185 kpsi)x( 1000 psi / 1 kpsi ) = 106745 psi

D_o = 0.624 in

D_I = 0.255 in

F = S_sy x 2 x p x ( D_o² - D_i² ) / 4 = 54385 lbs

Calculating allowable tension load ( F ) on the assembly according to bearing stress on threaded cylindrical section of fitting:

S_sy = 0.577S_y = (0.577)x(207 kpsi)x( 1000 psi / 1 kpsi ) = 119439 psi

N = 18 threads per inch

p = 1/ N = 1 /18 inches per thread

A_t = 0.203 in²

A_r = 0.189 in²

h = 0.625 in

F =  S_sy x h x ( A_t - A_r ) / p =  18812 lbs

I doubt design requires more than nine tons of tension loading so the loads appear acceptable. However, the comparison indicates that the threaded cylindrical section of the fitting should fail before the yoke bolt. I have the impression the assembly has not experienced any maintenance to adjust bolt torque and problems could be undetected if limited visual examination is the only available NDE.

Calculating allowable tension load ( F ) on the assembly according to stress on door frame attachment to the hex head bolt:

An assumption was made concerning the type of material used for the door frame so the following calculation is subject to change.

S_sy = 0.577S_y = (0.577)x(50 kpsi)x( 1000 psi / 1 kpsi ) = 28850 psi

A_hex = 1.5348 in²

A_dia = 0.2476 in²

F =  S_sy x ( A_hex – A_dia ) =  37136 lbs

The door frame attachment stress calculation indicates a greater allowable value for load than the bearing stress on threaded cylindrical section of fitting. However, the calculation can be misleading if no other stress factors are considered. Creep due to high temperatures could have influenced the door frame material more than the materials of item #9618; and thus, significantly reducing the allowable load. Fatigue could exacerbate the problem too. Again, limited visual examination of the attachment presents a problem that should be addressed by better NDE.

Note that I did not use safety factors in the preceding analysis since I was evaluating failure scenarios. Normally, I would expect design to incorporate a minimum safety factor of four. It is my understanding the bolts were initially torqued with the landing gear door closed until the door skin and the orbiter fuselage skin were flush to one another. Afterwards, the tiles were bonded onto the skin and the assemblies were never adjusted again. In my opinion, only one of the four assemblies could maintain the appearance of the flush condition with a reduction in safety factor even if the other three assemblies had defect since that one assembly would acquire the load normally distributed among all assemblies. However, I believe this presents an unacceptable risk if the only non-defective assembly could experience some type of failure during flight.

Now, permit me to discuss an example of mechanical part functionality. About two decades ago I was supporting a major piping replacement project at a nuclear power plant since microscopic stress cracks were found in the nuclear boiler piping. I volunteered to assist in examining the insertion and extraction of probe instrumentation sensors designed for monitoring radiation levels inside the boiler of a nuclear reactor. The nuclear fuel had been removed, but it was still a hazardous assignment due to residual radiation and heat. I won’t get into how much protective gear I donned, but I had to position myself directly beneath the boiler and observe the movement of the sensors. Whenever movement occurred I would indicate by hand to another person about ten feet away who would communicate my response by a radio headset to the control room operators. The importance of the procedure was to determine whether a mechanical part behaved like a mechanical part since the capability to achieve the designed monitoring locations was just as important as the capability to achieve the designed instrumentation signals. This procedure was an opportunity not missed even though not related to the piping stress crack problem. Note that the CAIB Final Report has examples of opportunities missed.

I’ll try to relate this example to the up-lock roller. It is all well and good that inspections are effected, but I have concerns. Visual examination of deformity, damage, cracks or breaks in the hex bolt seems extremely limited without its removal or testing the torque or tightness. How do you visually determine that the hex bolt has not sheared across the diameter somewhere in the threaded engagement section? Do the 0.030" tile step & gap measurements really determine the state of the hex bolt if a shear across the diameter in the threaded engagement section only creates a microscopic break gap of 0.020"? Do the detailed landing gear functional tests impose loading typical of flight conditions experiencing vibrations, temperature extremes and the cumulative dynamic forces imposed on connecting hardware? Does the 0.376" hex x 0.781" length at end of hex bolt have test purpose potential means for applying torque to the hex bolt?

Let’s revisit that thread stress calculation. Previously I had calculated with assumption that the failure would occur at the fitting threads under full threaded engagement. I tried to show that failure, if it occurs, would probably occur there first. Note that the pitch and length of threaded engagement are important. Assume a shear occurs across the diameter of the hex bolt just below the first thread in the threaded engagement section near one of the ends of the engagement length leaving a microscopic break gap of 0.020". This time the stress is applied to the material of the hex bolt and there is no effective full threaded engagement since there is a 0.020" microscopic discontinuity. What is the allowable load?

Calculating allowable tension load ( F ) on the assembly according to bearing stress on threaded section of hex bolt:

S_sy = 0.577S_y = (0.577)x(185 kpsi)x( 1000 psi / 1 kpsi ) = 106745 psi

N = 18 p = 1/ N = 1 /18 A_t = 0.203 in² A_r = 0.189 in²

Since stress is transmitted through only one of 18 threads in the one inch length,

h = 1 in / 18

F = S_sy x h x ( A_t - A_r ) / p =  1494 lbs

There is a significant difference between 18812 lbs and 1494 lbs of allowable tension. Note that the "as-found" bolt alignment is offset from the yoke/roller configuration and some bending moment should be anticipated as the hook apparatus pulls on the roller in order to achieve tension. Therefore, this allowable tension should actually be reduced via significant safety factors since this stress theory is based on average stress exacerbated by the possible bending of the thread. I could do other analyses assuming the break occurs leaving less than one complete thread. However, NASA may discover by doing the math that the allowable tension could theoretically approach zero with this type of failure scenario. I'm not too concerned about shearing at a point in the bolt length absent of thread engagement since such scenario failure would probably become immediately apparent by applying tension with the hook. However, a break near the most vulnerable end of the threaded engagement could be less detectable if the allowable stress is marginally sufficient to close the door without thread failure. By the way, extremely cold temperatures make metal brittle so consideration of maximum-shear-stress theory should also be considered where S_sy = 0.50S_y. Again, I’m not using any safety factors in this calculation. I also have insufficient data to predict anticipated tension loads and all possible stresses for each roller assembly. That said, it is possible to incur a thread failure at other locations along the bolt without decreasing the allowable tension requirement if enough good threading still exists near the most vulnerable end of threaded engagement. Therefore, my analysis should be used as a guide for a more comprehensive study, including the appropriate determination of whether critical or non-critical classification should be assigned to up-lock roller assemblies.

• NASA review stress calculations for the MLG up-lock rollers
• NASA review application of available NDE techniques for MLG up-lock rollers
• NASA perform additional testing of item # 9618 for failure assessment

Relevancy

• Confidence in the closeout photos of the link assembly on Discovery's main landing-gear door was replaced by a borescopic examination for similar problems.

The author of this report makes no assertion of professional qualifications and assumes no liability for the content. The author’s intent is to raise awareness of potential problems for appropriate NASA personnel to resolve.

Additional references can be obtained indirectly from my LH MLG Up-Lock Roller Trajectory Analysis Report of October 29, 2003.

The following images of specifications, along with my interpretations, were derived from Grumman Aerospace Corporation drawings provided by Les J. Boatright: