Aeroengine Safety

Fig. "Probabilits of failure detection by NDT" (Lit. 25.2.2.1-1): This picture gives an impression about the influence of the testing procedure manual, half automatic and automatic at the identification of faults and with this at the safety of the parts. This can be of high importance for the remaining risk of parts during acute problems like material faults and LCF cracks.
The named following fault sizes are merely indicative values. Basically apply the specifications of the OEM, of adminstrations respectively in instructions and directives.
The probability of a crack growth during operation depends crucially from the fault size/crack size (volume 3, Ill. 12.2-4). Only if the non destructive testing finds with sufficient likeklihood a growth viable failure, it can guarantee the safety of the part. Naturally the individual, maximum operation load in the crack area must `fit' to the crack size. The dimensioning of parts or a risk assessment of acute problems must consider these limits. Also the fracture mechanic behaviour of the material (fracture toughness) decides about the crack size with sufficient low risk. A fracture mechanic design from experience should assume surface faults/flaws (SF) in the range of 1 mm and of 0,5 mm edge cracks/corner flaws (CF). Inner cracks/imbedded flaws (IR) are realistic supposed at 0,5 mm.

The probability of identification of a test procedure (Ill. 25.2.2.2-2) and the disposition for crack growth depends beside the fault size and loat level, also from the position of the fault in the part (sketch above).
The upper chart contains fault sizes/crack sizes which guarantee a probability of detection (= POD) of 90 %, with a reliabilityof 95 % (confidence level = CL).
It can be seen, that with a semi automatic testing, independent from the material, markedly smaller surface cracks (SF, CF) are found as with manual testing. This is especially also true with regard to the very frequent penetrant inspection (fluorescent penetrant inspection = FPI), which depends highly from the tester. However for inside located faults, the probability of detection from half automatic and manual is equal. This may be seen in connection with the here used processes ultrasonic testing and X-ray testing. It is interresting, that the fully automatic testing for surface cracks (SF, CF) finds once more markedly smaller surface faults. This is also true for not inside positioned faults. The dependence of the detectable fault sizes of the material is striking. The testability of titanium alloys, obviously is markedly unfavourable than for nickel alloys.
The chart below contains sizes of faults/cracks /of which 90 % with a confidence level of 50 % are found during an automatic testing . Here exist markedly material specific differences in the fault size.

Ill. 25.2.2.2-2 (Lit. 25.2.2.2-2 and Lit. 25.2.2.2-3): This picture shows on the example of titanium alloys, from whose crack lengths at certain test procedures, the detection of a surface fault/flaw can be expected. Indeed the eddy current testing is obviously the best. Anyway, in the shown case even for 5 mm large faults the needed reliability (confidence level) is for highly loaded parts like rotor components not sufficient sure detectable. Especially the reliability of the penetrant inspection is desillusioning. However the situation may be markedly favorable to expect, for a targeted surch for a certain failure type in a known position, like for acute problems (e.g., LCF cracks on disk grooves). The probability of detection should be evaluated in a risk analysis, for the case on hand from an experienced expert, as sufficient as possible.

Ill. 25.2.2.2-3 (Lit. 25.2.2.1-4): After a loud bang the airpland laid extremely to the side and crashed. The failed aeroengine was an elder type with low bypass ratio (sketch above).

An investigation showed, that the failure started from a crack in one of the two labyrinth tips (sketch middle right) of a distance ring (spacer, sketch below right). It resides inside the high pressure compressor between 9th and 10th stage (sketch below left). This crack formation from experience is connected with thermal fatigue cracks or hot cracks/heat tears during an intense rubbing process (volume 2, Ill. 7.2.2-9.2). The spacer consist of magnetic steel.

Component tests: A spacer was equipped with different large cracks (cuts) and was cyclic spin tested. From the crack propagation could be suggested, that an about 0,25 mm deep crack needs about 25 000 cycles to penetrate the cylindric part and further 1000 up to the ripping. For the transit from the thin cross section of the labyrinth tip into the thicker ring, the crack needs about 3000 cycles. This can be used for the determination of the interval.
From test data it can be sufficient exact suggested at the crack propagation during operation.

History: Up to the current failure, 45 cases with the fracture of a spacer during flight emerged. In 26 cases the aeroengine hat to be shut down. In 7 cases fragments esccaped (uncontained). In no case there were injured.
Unfortunately the total cyclic number of the spacer of the current case could could no more determined. The documentation was incomplete and the spacer was assembled in different aeroengines. It was magnetic crack inspected (fluorescent magnetic particle inspection = FMPI) about four years, respectively about 2 600 cycles before the accident, during an overhaul. This inspection process (Volume4, Ill. 17.3.1-4) can be classified as reliable from experience, for the configuration on hand. From the cyclic spin tests can be concluded, in spite of the lacking cycle documentation, at the crack length during the last inspection. At the time of the last crack inspection, it had already a length, which should guarantee a sure detection. However the responsible testing personnel of the repair shop showed deficiencies.

Measures:

• Eddy current testing on wing of the concerned spacer. The comparatively high sensitivity and probability of detection of this procedure (Ill. 25.2.2.2-2 and volume 4, Ill. 17.3.1-3.1) should enable a safe crack growth over about at least 8000 cycles. Not until then, a dangerous size could be anticipated. However as uncertainty remained, that the inspection on wing takes place with limited accessibility. So only cracks can be identified, which already noticeable run in the cylindric part. In such a case, there are merely about 1000 cycles up to the fracture. For this reason, the interval between eddy current tests was specified 1000 cycles.
• Exchange of the concerned spacers during the next rotor disassembly within two years or 4000 cycles. Introduction of an improved design.

Ill. 25.2.2.2-4 (Lit 25.2.2.2-4 ): Already since 1966 gamma rays are used for the X-raying of civil aeroengines. At this procedure the attention applies for

• Wear/abrasion.
• Deformations: Creep (sketches below).
• Offset/loosening.
• Distances, gaps (Fig. "Assembly fault causing exit of fragments").
• Deteriorations, large cracks.
• Missing of parts or part sectors (sketches above).
• Failures of parts.

The displayed cases apply for elder aeroengine types. At several newer the application of this testing procedur was considered by design.

Ill. 25.2.2.2-5 (Lit. 25.2.2.2-5 and Lit 25.2.2.2-6): With gamma rays of isotopens also some interior problems of aeroengines can be monitored by „X-raying“. This enables the smaller size of the radiation source, compared with a X-ray tube. This procedure is, depending from the accessibility, applicable at several locations at the whole length of the aeroengine. In the shown case, turbine entrance guide vanes (nozzles) are inspected. Thereby deformations (Ill. 25.2.2.2-5) and crack development respectively break outs may be concerned.

Ill. 25.2.2.2-6 (Lit. 25.2.2.2-7, Lit. 25.2.2.2-11 and Lit. 25.2.2.2-12): With a sufficient intense radiation source (e.g., linear accelerator, frame below, volume 2, Ill. 7.1.4-17) it was succeeded from 1970, to X-ray whole aeroengines (sketch above). This took place at the aeroengine on the test rig. Also pulse frequencies between 50 and 500 Hz have been possible in this installation. Thus dynamic processes like the movement of components can be filmed/recorded. The investigation, takes place in before determined zones of the aeroengine.

So gaps like at blade tips and labyrinths (middle sketches) could be monitored. These can be related different efficiencies of individual components, bearing loads (volume 2, Ill. 7.2.1-2 and Ill. 7.2.1-3) and rubbing processes (volume 2, Ill. 7.0-4). For this, firstly the cold condition and the for a comparison interresting stationary condition was documented. These enable an evaluation of the results of the operation. Also changes of the gaps during cooling at stand still, are of high interest (volume 2, Ill. 7.1.2-9.1). It is important for example, to know the time period at which different thermal expansions of the components lead to the jamming of the rotors. This concernes especially the shafts of the gas producer. These are not easy controllable from the outside for free motion . In this connection also the bowing of rotors during the cooling phase (rotorbow, voöume 2, Ill. 7.1.2-9.2) is to mention.

The analysis of the pictures needs specific expertise and experience. Even then, when the region of interest is identified. For the layperson often the changes in the image documentation are not identifiable. Here expert knowledge is demanded like for x-ray photos of the medicine.

Ill. 25.2.2.2-7 (Lit. 25.2.2.2-8): Also the oil analysis (SOAP, Fig. "Specifications for oil contaminations" and Ill. 22.3.4-2) can be considered in the broadest sense asnon destructive test. The pictures show examples about the successful application of the procedure.

Ill. 25.2.2.2-8 (Lit. 25.2.2.1-9): Oxidation protection coatings (diffusion coatings, overlay coatings) of the turbine blading enable long operation lifetimes. Therefore its condition is of high interrest.
Two systems are used (volume 3, page 11.2.3.3-4):

Diffusion coatings mainly with aluminium and chromium (volume 3, Ill. 12.4-8 uad volume 4, Ill. 16.2.1.8.1-2).

Overlay coatings of the type MCrAlY (details above in a metallographic section; volume 4, Ill. 17.3.2-5).

With the operation time, the coating of both types will be consumed by oxidation from the surface. Additonally diffusion processes take place. These change the coating structure and the composition. Therefore the deterioration is especially interesting for the evaluation of the remaining lifetime and so for the logistics.
For this purpose, thermography can be used (volume 4, Ill. 16.2.2.3-19). Here differences of the thermal conductivity between base material and from service changed coating, are sufficient. The cooling progression after the process determined heat impulse, allows conclusions at the coating thickness (diagramm below). This thermography technique enables at the most coating systems a fast and repeatable exact measurement of the coating thickness. However, for this representative parts must be removed, as long as no invasive technique (through borescope openings?) is available (chapter19.2.4).

Ill. 25.2.2.2-9 (Lit. 25.2.2.1-3): At a multitude of aeroengines of this type the feet of the fan blades had to be ultrasonic tested for crack formation. Reason was the fracture of a blade after about 14 000 start-stop cycles. Based on an intense investigation of the crack development, in a service bulletin a check with the binocular (30x) was demanded. It replaces the penetrant inspection(!), demanded in the manual. This obviously was not sufficient to guarantee the necessary safety.

The fan blades had been removed from the aeroengines on wing and underwent an ultrasonic testing. If a crack was found, besides the whole blading, also the disk had to be exchanged.

Comment: Perhaps LCF cracks are concerned (volume 3, Ill. 12.6.1-6). These develop in the contact surfaces of the feet, mostly in combination with deteriorating fretting. This can be traced back at blade vibrations and/or movements of the blade foot (volume 2, Ill. 6.1-15.2). The blades slip during the widening of the disk grooves under centrifugal force. With fretting, also the bad penetrant inspectability could be explained, because the concerned surface can be rough and small cracks can be smeared.