Motor bearing failure analysis is a term often heard from manufacturers by users of motor bearings. When a bearing fails, manufacturers are asked to identify the cause, often conducting what’s known as failure analysis, and presenting a failure analysis report.
However, some reports can be unclear, leaving non-specialist engineers puzzled by the conclusions drawn. To tell you the truth, professional engineers, those who deal with bearing applications, often find many reports obfuscating, taking advantage of the non-specialist engineer’s unfamiliarity with the field of bearing failure analysis.
To help non-specialist engineers understand, I will attempt to discuss some basic concepts of bearing failure in a casual manner, at least to prevent motor engineers from being confused when reading bearing failure analysis reports in the future.
Why are many bearing failure analysis reports “inaccurate” or “useless”?
Bearing failure analysis involves analyzing the failure of the target bearing. According to ISO standards, there is a standard classification for bearing failure analysis, and the first task for a bearing engineer is to classify the failure mode of the target bearing according to a unified standard.
The purpose of this classification is to deduce the common causes of these failures through typification of failure modes. Here, bearing engineers face several challenges:
First, the accuracy of failure classification. Failure modes are often classified through recognition of the surface morphology of the bearing, but the surface of bearings varies greatly.
The failure mode images provided in standard classifications often contain descriptions that are difficult to quantify with language or certain data, so these classifications are heavily reliant on experience. Engineers with insufficient experience often struggle to classify accurately.
Second, a bearing’s failure is often a mix of several failure modes. Determining a rough sequence during identification poses a serious challenge for bearing engineers. Finding the initial failure is crucial in identifying the cause of failure.
Third, the potential causes deduced from the standard classification of motor bearing failure generally match large categories, but to identify the specific cause requires investigation of surrounding factors.
Fourth and most troublesome, even the ISO standards state that these failure types cannot cover all bearing failures 100%. Encountering a failure mode that cannot be classified is a severe test for bearing engineers. Although this problem is usually solvable and the chances of encountering this situation are indeed low.
People often say that motor bearing failure is a “ghost” issue. Frankly speaking, are there “ghost” issues? Yes, but in my personal experience, in 20 years of service, I have heard of them but never seen them.
For both general mechanical engineers and bearing application engineers, I sincerely say, “Don’t blindly believe that you have encountered a ghost, there aren’t so many ghosts for us to encounter. Saying this is often an excuse for a lack of professional ability, isn’t it?”
In actual bearing failure analysis work, failure analysis reports issued by certain authoritative bodies often take the following form:
1. Documented state of the observed bearing (via photos, etc.)
2. Failure type categorization based on the observed surface condition of the bearing.
3. Citing ISO or relevant standards to indicate potential causes, based on the failure classification of the bearing.
However, these report conclusions are often unhelpful for mechanical engineers in identifying the cause of failure (or sometimes found to be inaccurate).
This is mainly due to:
Firstly, the report only classifies based on what is observed. This “observation” might only be surface markings on the bearing, and classification analysis is conducted based on these markings. But the operating conditions surrounding these markings are not considered in the analysis. This often overlooks crucial triggers of failure.
Therefore, these seemingly consistent reports often struggle to identify the root cause of failure.
Secondly, for authoritative bodies’ failure analysis reports, the classification of failure types is generally not wrong (though some classifications by other organizations or manufacturers can be embarrassingly flawed), but the outcomes leading to failure are directly referenced from possible causes of such failures in the standards.
This practice often lacks practical significance. The possible causes in the standards cover a very broad scope.
Under the current case conditions, many surrounding factors need to be eliminated, and as much as possible, causes with smaller associations should be ruled out, providing the user with a clearer potential reason. Without this step, the analysis report often has limited value to on-site engineers.
Of course, sometimes the above situation is unavoidable. For example, authoritative testing structures, in order to ensure rigorous causality logic in the report and to ensure that all judgments are supported by standards, have no choice but to produce this type of “flawless” yet “useless” failure analysis report. Naturally, this situation is beyond the scope of technical discussion and will not be delved into here.
Through the above analysis, we can understand that bearing failure reports are influenced by capabilities, conditions, and other considerations, which can make reading difficult for non-professional engineers, and even make the reports unusable directly.
However, from another perspective, engineers know that as long as they master the art of pattern recognition, writing reports like the one above is quite straightforward. But this is not true bearing failure analysis. After all, any analysis report that can’t directly guide field operations is merely deceptive.
In fact, to avoid the aforementioned “deception,” mechanical engineers need to learn certain methods for diagnosing bearing faults and conducting failure analysis. Delving deep and refining these skills can be challenging, but basic mastery for discerning deception is achievable.
Many engineers performing failure diagnostics also conduct bearing failure analyses. But for the average mechanical engineer, are bearing failure diagnostics and failure analyses the same thing? Moreover, what’s the connection between failure diagnostics, vibration monitoring, and vibration analysis?
Let’s delve into it
What is failure diagnosis?
As the name suggests, failure diagnosis involves diagnosing failures. To understand what failure diagnosis is, we must clarify two concepts.
First, what is a failure? There are various definitions for failure, and I won’t delve into academic references here. You’re free to look up basic concepts on your own. But in layman’s terms, a failure is relative to a normal state. If a piece of equipment isn’t functioning normally, we can say it has a failure. Thus, the failure diagnosis we’re talking about is diagnosing the “abnormal” condition of equipment. Whether we’re talking about the “normal” or “abnormal” condition, both are used to describe the state of the equipment.
By the way, the meaning of state monitoring is to supervise and detect the state of the equipment. This encompasses both the monitoring of the equipment’s “normal” state and the supervision of its “non-normal” state.
And what about “diagnosis”? It’s essentially inspection and judgment. The diagnosis of equipment failure is the inspection and judgment of the equipment’s failure state. What’s the purpose of inspection and judgment? It’s to find out where and what the problem is. Where the problem lies is the “localization” judgment in failure diagnosis, and what the problem is, is the “accountability” judgment in failure diagnosis.
The further use of the conclusion from the failure diagnosis is the subsequent analysis of the causes leading to the failure, as well as the formulation of corresponding measures. This process is akin to a treatment process. Sometimes people often include the identification of the cause in the failure diagnosis. This is also appropriate.
Is fault diagnosis the same as failure analysis? Or vibration analysis?
From our understanding of fault diagnosis, it’s clear that it’s a process of identifying and judging the failure state of a device. In fact, when mechanical equipment fails, it exhibits certain characteristics, such as excessive vibration, overheating, or loud noise. At this point, the device might have sustained internal damage, or it might not.
Engineers utilize fault diagnosis to adjust the malfunctioning parts. If all internal components of the device are undamaged, then the fault can be eliminated after appropriate adjustments, returning the device to normal operation.
If an internal component of the device is damaged, then engineers often need to repair or replace the component. After the repair or replacement, the device fault is eliminated and the device returns to normal operation.
In the aforementioned situations, when a component of the device is damaged, we say that the component has failed. Analyzing the failed components is what we call failure analysis.
It’s evident that failure analysis is an evaluation of the failures. If a component is entirely without issues, then even though it may be in the fault diagnosis phase during operation, failure analysis is not conducted. Failure analysis is the analysis work conducted after a failure has occurred, which could be at the initial, mid, or late stage.
From the above, it’s clear that device failure analysis is a small part of the larger scope of fault diagnosis.
Returning to the topic of fault diagnosis, changes in the operational status of equipment can be detected by monitoring systems. As we know, vibration is a critical parameter (or state variable) reflecting equipment performance.
During operation, we supervise the machine’s conditions through its vibration signals. When a fault occurs, the vibration signals respond accordingly, allowing us to analyze potential details of the malfunction. Vibration analysis encompasses all these tasks.
However, the state of operating equipment is not solely characterized by vibration. Other parameters, such as noise and temperature, also play significant roles.
We employ various parameters to analyze the operating state of the equipment, with vibration analysis forming a part of this process. Vibration analysis can occur both during equipment state monitoring and fault diagnosis.
It’s evident that vibration analysis is one of the methods used in the process of equipment fault diagnosis. Aside from vibration, we employ many other approaches when conducting fault diagnostics, such as studying operational parameters, analyzing temperature and noise, and even examining the surface morphology of bearings (failure analysis of bearings), among others.
By now, you should understand the scope and differences between fault diagnosis, vibration analysis, and failure analysis.
What can vibration analysis do that failure analysis cannot? What can failure analysis do that vibration analysis cannot?
This question may seem lengthy, but understanding it can prove particularly useful when dealing with analysts who claim to be “omniscient”, allowing you to instantly expose their limitations.
The primary approach to bearing failure analysis is the examination of the bearing’s surface morphology. In other words, it’s essential to inspect the bearing raceways, rolling elements, cage, end faces, inner surfaces, outer surfaces, and so forth.
Hence, a comprehensive bearing failure analysis requires disassembling the equipment, even the bearing itself. It’s crucial to observe characteristic patterns to make judgments and discernments. From this restriction, it’s clear that an engineer cannot perform failure analysis before disassembling the machine.
Although you can conduct oil analysis if you extract lubrication, this does not constitute failure analysis for the bearing housing. Therefore, it’s impossible to perform failure cause analysis.
On the other hand, equipment vibration analysis discriminates between normal and faulty operation through collected vibration signals. Hence, it’s possible to collect vibration signals without necessarily needing to disassemble the equipment.
However, at its core, vibration analysis can only perform location analysis, not responsibility analysis. For example, it’s almost impossible to distinguish between a dent in the raceway caused by contamination and one caused by improper installation using vibration analysis. However, this distinction can quickly be made through failure analysis.
In summary, failure analysis requires visual access to the failed surface, and nothing can be done before that. Vibration analysis can at most identify which surface has an issue, but it cannot carry out in-depth categorization.
Having said all this, I haven’t even started on what failure analysis is. However, understanding these fundamental concepts is essential, and they offer many benefits.
Although our previous discussions on bearing failure analysis have not directly addressed the immediate challenges many engineers face, the knowledge remains highly significant. Many individuals, in their haste to understand bearing failure analysis, are eager to match types and images in hopes of quick categorization and so-called “diagnosis.”
This method may seem fast, but in reality, it often proves to be the slowest route. This is because the categorization of motor bearing failure types is always predicated on a basic understanding of the concepts. Without this, such categorization is either inaccurate or still incapable of guiding the resolution of actual onsite problems.
However, at this juncture in our discussion, we will delve into the categorization issues within motor bearing failure analysis. We won’t adopt the conventional methods of exposition found in articles and written introductions.
Instead, beyond the definition of categories, we will explore deeper issues, such as the mechanisms of these failures and the relationships between these failures and their causes.
Let’s start with fatigue. According to the classification of bearing failure analysis, fatigue always comes first. What exactly is fatigue?
From a force perspective, when the contact surface withstands positive pressure, its interior experiences shear stress (if you’re not familiar with terms like shear stress and tensile stress, you’re encouraged to consult related content in theoretical mechanics).
As for bearings, when the rolling element passes over the raceway, the internal shear stress distribution in the raceway is as shown in the following illustration:
The image reveals the distribution of shear stress beneath the contact surface (an indicative illustration only; for the actual diagram, please refer to the relevant literature). It is evident that the shear stress peaks on the surface and just beneath it.
As the bearing rotates, the rolling elements cross a fixed point on the surface, with the shear stress increasing before dissipating. This cycle repeats as each subsequent ball rolls over the same point.
In other words, the interior of the bearing ring experiences cyclic shear stress due to the rolling compression of the balls. Consequently, the number of shear stress cycles beneath the bearing surface per rotation equates to the number of balls.
When metals are subjected to cyclical shear stress, the internal structure of the metallic material undergoes changes, leading macroscopically to the emergence of cracks. To illustrate this without resorting to theory, consider a commonplace example.
If we bend a thin wire once, it doesn’t break; however, after repeatedly bending it back and forth, it snaps. This is a familiar childhood experiment. Upon close examination of the break, one can see radial tearing at the break point. The radial tearing direction corresponds to the axial shearing direction, clearly demonstrating the impact of shear stress.
For those interested, this experiment can be repeated as an early fatigue test. For instance, bend the wire several times, but stop before it breaks and inspect the bend. At this point, you’ll notice numerous fine cracks. These are the telltale signs of fatigue fracture.
Applying this example to bearings, the internal shear stress status of the bearing ring and rolling elements during operation is strikingly similar to that within the wire. As can be seen from the image above, the maximum shear stress in the bearing occurs on the surface and just beneath it at a certain depth.
Given an equal number of rotations, the interior of the bearing experiences the same number of shear stress cycles, leading to the formation of initial cracks at the point of maximum shear stress.
Cracks have appeared within the bearing ring or the rolling elements, then these cracks expand, leading to spalling. This is what we refer to as fatigue spalling. The process has been discussed in many books, and thus won’t be elaborated on here.
Once we understand the causes of fatigue, there are several questions that still need clarification:
1. Since maximum shear stress occurs both on the surface and internally, is it possible for spalling to begin at two different locations?
2. What is the relationship between the principle of fatigue and identifying the causes of fatigue?
Though these two questions may seem elusive, they provide a basis for answering many practical issues encountered in the field.
In the above content, we discussed how bearing fatigue is related to the cyclic shear stress caused by the cyclic pressure on the bearing material. There are two aspects of this concept that warrant a more profound understanding.
Firstly, bearing fatigue is related to the magnitude of the shear stress. Given the same cycle count, larger shear stresses will lead to failure sooner. In understanding this, we can intuitively see that bearing fatigue is related to the bearing load.
The bearing load determines the normal force between a rolling element and the bearing ring, which, in turn, determines the size of the internal shear stress in the metal. (While the material itself can also influence this, we will discuss it later. For now, let’s isolate the concept of the load to facilitate understanding.)
Speaking of shear stress, the observant reader would not have trouble noticing that in the diagram of the previous article, there are two places where shear stress peaks: below the metal surface (the diagram in the above only drew underneath the raceway, but the situation within the rolling body is similar) and on the metal surface.
Readers familiar with bearing failure analysis should be able to associate this with the standard classification of bearing failure, where fatigue is divided into surface fatigue and subsurface fatigue (also known as below-surface fatigue).
Indeed, you’ve got it right. These are precisely the two locations where the maximum shear stress occurs between the rolling body and the raceway: at a certain thickness below the contact surface, and on the contact surface itself. Fatigue occurring at a thickness below the surface is called subsurface fatigue, while fatigue occurring on the contact surface is referred to as surface fatigue.
Does this description differ from what you typically find in textbooks? However, this understanding is derived from the mechanism of fatigue generation and indeed, is the origin of the classification in the standard. (As an aside, it’s worth noting that many engineers have blind faith in standards without understanding the reasons behind them, which limits the depth of their knowledge.
One should strive to understand the origin of standards, thus enhancing understanding of their appropriate application. Even flaws in some standards could be discovered in this way. Do not blindly follow standards; they are set by engineers and are intended to serve practical engineering. Anyway, let’s get back on track.)
Following this train of thought, if you look closely at the diagram in the previous article, it seems that the maximum shear stress on the surface would be greater than below the surface. Would this change in reality?
In fact, it is different. Given good lubrication, the maximum shear stress on the metal surface would significantly decrease, and might even be negligible. Therefore, if you flip through some books, you will find that there is no peak value on the contact surface in some diagrams. This is not because I drew it wrong, or others did, but because the premises are different. Others depict a well-lubricated scenario, while I illustrate a situation without sufficient lubrication.
Don’t think I’m just idly chatting—the above paragraph is quite useful. Through it, you will understand that bearing surface fatigue is related to poor lubrication. This is why it is stated in the standard that surface fatigue is related to lubrication (once again, helping everyone annotate the standard).
From the previous discussion, we can see that fatigue is related to the frequency and magnitude of shear stress. In bearings, the cycle number of shear stress is related to the number of rotations, and the magnitude of shear stress is related to the bearing load.
Thus, it’s not difficult to see that bearing fatigue is related to bearing load and speed. We also discussed surface fatigue and subsurface fatigue, noting that good lubrication can significantly reduce surface shear stress. This means that surface fatigue is mostly related to surface lubrication.
The logic isn’t complex, and I won’t expound on it here; readers can fill in the gaps themselves. The conclusions drawn from the above logic align with the potential causes suggested in the international standards for bearing failure analysis. (See, combining theory with practice isn’t difficult, is it?)
Now let’s return to the main topic, where we will discuss another critical type of bearing failure – wear.
The term “wear” is commonly encountered and fairly easy to understand. It refers to the phenomenon of surface failure caused by the displacement or loss of surface material due to friction between two moving surfaces. Despite my best efforts to avoid reciting definitions verbatim, expressing this concept remains somewhat tricky. Luckily, I have ample space here to explain this clearly.
While I strive to avoid rote memorization of concepts, I must emphasize the somewhat complex notion of “displacement and loss of surface material.” I highlight these aspects to distinguish between wear and surface fatigue.
During surface fatigue, cracks appear within the metal (indeed, under microscopic observation, the localized area appears whitish before a crack forms. Thus, when explaining “white cracks,” my personal interpretation leans toward fatigue-related explanations). When these cracks propagate, leading to local flaking of the metal (commonly referred to as fatigue spalling), only then does material loss occur on the metal surface.
Unlike fatigue, wear may quickly involve displacement and loss of the metal surface material. How do we explain this displacement and loss? In the standards for bearing failure analysis, bearing wear is divided into two categories: adhesive wear and abrasive wear. These two concepts can help clarify our understanding of wear.
Adhesive wear refers to the transfer of material from one friction surface to another. Due to this material transfer, the surface topography of the two surfaces changes, leading to the wear failure mode known as adhesive wear.
This concept may still sound abstract. Let’s use a simple analogy: it’s like a girl using foundation makeup. When she swipes the sponge across the powder, fine particles stick to it. This is adhesive wear.
If my 90% male fan base finds the foundation analogy difficult to comprehend, consider the use of a stamp. The principle is the same. Two surfaces compress, and material sticks from one surface to the other.
Of course, in the case of bearings, on a microscopic level, two metals compress together, creating a high local pressure and causing cold welding. When the contact points separate, localized material adhesion occurs. This is the process of adhesive wear.