A Position Paper on Failure Assessment and Establishing Root Cause in Failure Analysis
A. K. Koul
Engineering failure analysis is an enormously important branch of engineering science. Notwithstanding the considerable ingenuity of scientists and engineers to design, build and operate sophisticated instruments, equipment, machinery and structures, sometimes their performance does not follow expectations. We see this in all types of man-made articles from transportation systems to energy delivery equipment. Gas turbine engines are no exception. There are many basic reasons:
- A part may be designed to requirements that do not fully represent the service conditions
- The operator of, say an aircraft or engine, may substantially change the way the equipment is used so that the original design assumptions are no longer valid, and damage may accumulate during service at a much faster rate than expected.
- A design may be faulty, involving inadequate safety margins of stress or operating temperatures. The choice of materials used in manufacture may be wrong.
- A defect, introduced during manufacture or assembly may remain undetected in the part during service.
- Damage introduced during service may miss detection during scheduled inspection and overhaul, and remain to cause premature failure.
The purpose of failure analysis is to determine how a component failed and what caused the failure to occur. Failure analysis is therefore one of the most important tasks carried out by the engineer because if correctly performed it should resolve any uncertainties about the validity of the original design, about the materials and manufacturing methods employed, and about the way the equipment has been © Life Prediction Technologies Inc. 2 used. If the basic design, materials, manufacturing or maintenance methods are found to be defective, then failure analysis provides a basis for sensible corrective action. This is a process often referred to as Retroactive Design.
However, the examination of the fracture surfaces or metallographic examination of the failed parts may not always be sufficient to get at the root cause of the failure of a component that may have been subjected to complex thermal-mechanical loading histories in highly corrosive or oxidizing operating environments. Examples of failures of a high strength steel bolt in an aircraft structure and a Ni-Fe base superalloy turbine disc in a land based turbine engine are presented in Figure 1 (a) and Figure 1 (b) respectively. A typical fracture surface attributed to stress corrosion cracking in stainless steel material is also included for comparison, Figure 1 (c). All fracture surfaces look very similar. Only upon examining the material manufacturing processes, part operating environments and loading conditions in detail, was the root cause of the steel bolt failure established to be hydrogen embrittlement and prolonged low temperature creep loading at 0.45 to 0.53 T/Tm of the alloy in the case of the turbine disc. In the case of the Ni-Fe base superalloy turbine disc, the client was convinced that stress corrosion cracking as opposed to creep mechanism was responsible for the turbine disc failures because the OEM including several other consultants that they had engaged, prior to involving LPTi, had reached this conclusion.
In fact, LPTi had to conduct several creep experiments in the temperature range of interest on specimens machined from service exposed discs and perform detailed fractography to convince the client that creep was indeed the underlying deformation mechanism responsible for crack nucleation and growth in their discs. A thorough understanding of the classification of fracture and deformation mechanisms is thus critical in establishing the root cause of failures in gas turbine engine components because the operative ranges of service temperatures and stresses vary significantly from one component to another throughout the engine. In addition, a quantitative understanding of the damage accumulation rates is equally important because components may be exposed to a wide range of stresses, temperatures and temperature gradients where competing deformation and fracture mechanisms can contribute to a failure.
Classification of Fracture
Several handbooks, scholarly articles and individual chapters in materials engineering text books have been written on this subject. The objective here is not to present a critical assessment of different fracture mechanisms described in the literature but simply to point out the mechanisms that are most relevant in root cause failure analysis of gas turbine components. An attempt has been made to broadly classify different fracture mechanisms on the basis of thermal-mechanical loading conditions that a component may be exposed to during service in a gas turbine engine, Figure 2. Typically, thermal-mechanical cyclic loading and time dependent steady state loading effects at elevated temperatures predominate in gas turbine components. Environment effects related to oxidation and hot corrosion damage mechanisms can also play an equally important role in initiating fracture under specific engine operating environments in gas path components. Even embrittlement effects related to hydrogen embrittlement, metal induced embrittlement and stress corrosion cracking may play a role in specific parts. A key differentiator in assessing the root cause in hot gas path components is the homologous temperature operative at the fracture critical locations and some of this information is interspersed in the classification in Figure 2.
Role of Underlying Deformation Mechanisms
A fracture critical location in a gas turbine component is typically activated if the cyclic or steady state stress and temperature combination imparts enough strain at that location to nucleate a crack. At a given temperature, service stresses may cause transgranular deformation as a result of the movement of dislocations in the form of dislocation glide and climb including cross-slip within the grain interiors and intergranular deformation due to the glide and climb of grain boundary edge dislocations in the grain boundary plane leading to grain boundary sliding (GBS). The GBS strain accumulation may be accommodated by the formation of wedge type cracks at grain boundary triple points or vacancy migration induced grain boundary voids depending on the service temperature operative at the fracture critical location. Several such deformation mechanisms may contribute towards overall deformation at a given fracture critical location but generally a specific deformation mechanism dominates more than others. The dominance of a specific deformation mechanism can be discerned from a deformation mechanism map where the boundaries of different deformation regimes are demarcated in homologous temperature versus normalized stress space for a specific grain size of the material. A typical deformation mechanism map for Cr-Mo-V steel used in land based turbine disc applications is shown in Figure 3.
It is noteworthy that this deformation mechanism map is markedly different from the original Ashby maps and is based on experimental work and theoretical studies conducted by the author on deformation and fracture in complex engineering alloys for over three decades. Strain rate contours are also superimposed in this map to indicate the dominance of different mechanisms over different stress-strain rate and temperature regimes. It is understood that, for a given alloy, the boundaries of various deformation regimes will shift as a function of the heat treatment of the alloy if the heat treatment process changes the grain size or the relative strength of the grain interiors and the grain boundaries of the material. Apart from quantifying the actual thermal-mechanical loads that a component may be subjected to during service, this understanding of the material deformation behavior is fundamental to finding the root cause of a failure.
Damage Accumulation Rates and the Role of Microstructure
An estimate of time when a crack may have nucleated during service or which service loads are more damaging than others is often required by the clients? To arrive at this estimate, a thorough knowledge of the thermal-mechanical loading history at the fracture critical location along with quantitative relationships for damage accumulation rates for the different engine operating points within anoperational envelope are required. Empirical damage modeling approaches such as the use of Larson-Miller (LM) parameter for creep or Manson-Coffin relationship for LCF are routinely used to conduct damage modeling analysis. A fundamental flaw with empirical modeling approaches is that these models treat material as a continuum. It is now well established that, for a given thermal-mechanical loading condition, the rate of damage accumulation at a fracture critical location in a component is governed by the microstructural features present at that location in complex engineering alloys.
Microstructural discontinuities such as the presence of hard second phase particles within the grain interiors or the grain boundaries or semi-coherent precipitates used to strengthen the grain boundaries magnify the local stresses. The presence of high angle grain boundaries or even inter-phase boundaries act as sources of material weakness above a certain homologous temperature because dislocation climb is facilitated and the well known Hall-Petch strengthening effect of grain boundaries is diminished, Figure 4. This is why empirical damage models are not always reliable unless these models are calibrated using extensive material data bases generated over a wide range of stresses, strains and temperatures.
The author led the work on the development of constitutive models for creep, LCF, creep crack growth, fatigue crack growth etc. during his tenure at the National Research Council of Canada. The modeling approach that was followed used a novel idea of combining the classical deformation kinetics theories with a specific deformation mechanism related physics that was considered dominant for a given loading condition. With the result, the influence of microstructural features in controlling the rates of deformation under cyclic as well as steady state loading conditions could now easily be quantified. In any high temperature application, microstructure further evolves as a function of service exposure and this effect must also be considered for accurate assessment of any failure. The rate equations in their basic form are,
where rate of strain accumulation (𝜀̇) at any fracture critical location in a component is a function of the local stress (σ), temperature (T) and microstructure(S) and the microstructure in itself is an evolutionary function of stress, temperature and the original microstructural state prior to exposing the part to service conditions. An added advantage of using these models is that extensive material properties databases are not required for correct prediction. Instead accurate microstructural data are required. Generating quantitative microstructural data is significantly faster and cheaper than generating very large mechanical properties databases over a temperature range of interest for a given part.
Damage Evolution and Fracture
Several potential fracture critical locations may be present in a given part. Several of these fracture critical locations may be activated during service depending on the local stresses, temperatures or temperature gradients and local microstructural state at these locations. At a potential fracture critical location, localized strain accumulation may take place as a result of persistent slip band (PSB) or wavy slip band (WSB) formation under cyclic loads or dislocation glide and climb at specific microstructural sites under creep loading conditions. At high temperatures, both cyclic as well as steady state loads may contribute to strain accumulation at the same microstructural site thus leading to the classical dilemma of dealing with combined creep-fatigue interactions such as in turbine vanes in aircraft engines. Even stress redistribution may take place in highly cooled turbine parts as a result of localized creep strain accumulation. Over temperature or over stress incidents or coating spallation may also have occurred during service. All of these factors must be taken into account while dealing with failure assessment. Typically, when strain accumulation reaches a critical value at a fracture critical location, a crack may nucleate. Assessment of this crack nucleation phase is absolutely vital in establishing the root cause of failure. The formation of cracks in physics based modeling is described as,
where 𝑡𝑓 is the time to failure and a crack undergoes nucleation (𝑡𝑁) and short as well as long crack growth phases (𝑡𝑆𝐶𝐺 𝑎𝑛𝑑 𝑡𝐿𝐶𝐺) prior to final fracture. Typically, in forged components, both crack nucleation and short crack growth phases dominate the part life whereas crack nucleation phase alone plays a dominant role in the cast parts. Even high cycle fatigue (HCF) may dominate crack propagation beyond a certain point in the life of a turbomachinery part. All of these factors must be critically examined to establish the root cause in failure assessment.
Conducting fractography and metallographic studies on service exposed or failed parts may not always be sufficient to assess a failure or to determine the root cause of a failure. While this approach may work for simple engineering applications, in the case of parts operating in complex engineering systems that are exposed to a variety of thermal-mechanical and environmental loads at potential fracture critical locations it is often necessary to predict the local temperature and stress distributions and understand the impact of the sequence of loading on the root cause of a failure. A thorough knowledge of the underlying deformation and fracture mechanisms operative in advanced engineering alloys is also necessary for failure assessment.
A major problem with existing system designs including gas turbines is that parts or system designers treat the part material as a continuum and invariably use material properties data intensive empirical damage modeling techniques in conjunction with very large safety factors to establish the part life or preventive maintenance schedules. Such procedures ignore the impact of microstructural discontinuities such as the presence of hard second phase particles, grain boundaries, inter-phase boundaries, precipitate clusters etc. on the stress-strain state at a potential fracture critical location. This is why failures still happen in a majority of systems. Physics based damage modeling techniques eliminate this ambiguity.