Skip Navigation

Technical service provider and
supplier for simulation and testing

Anovis Blog by Tobias Clauß

2/5: From Requirements Specification to Solution

Tobias ClaußSales engineer

Back
Tobias Clauß

Tobias Clauß

Sales engineer

Hi, I'm Tobias,

Throughout my professional career, I’ve always worked in the field of acoustics. My specialty lies in the perception of sound in an industrial context. So, essentially, it’s always been about listening carefully. For over six years now, I’ve been in charge of our customer projects in the area of end-of-line test stands. I usually oversee test stands for years at a time, so I’ve seen quite a bit. In this blog , I share real stories from my work—stories that no AI could ever come up with. Why am I doing this? By sharing our firsthand experiences and lessons learned from countless EOL projects, I hope to help you get a robust, reliable test stand up and running more quickly.

The test scenario highlights a hastily configured system

Do you have a set of specifications for an EOL test system?

Very good. So a lot has already happened. Requirements have been gathered and coordinated internally; perhaps there have already been discussions with the test bench manufacturer, the quality assurance team, production, or purchasing.

But here's the key point:

A set of specifications is not yet a finished test task.

In my experience, there’s often a long way to go in terms of joint clarification between what’s in the specifications and what’s supposed to function reliably on the test bench later on. That’s exactly where requirements analysis begins.

Because if the specifications say, “Detect unusual noises,” then everyone in the room has a general idea of what that means. But a testing system can’t really make much of the term “unusual” at first. It needs clear criteria. It needs an operating condition, a measurement window, a sensor system, a signal characteristic, and a definition of what will later be considered a good part or a defective part.

Or, to put it another way:

The specifications often outline what is desired. The requirements analysis clarifies how this can be turned into a functional test. And that is exactly what this article is about.

A set of specifications is not the same as a test strategy

I regularly receive inquiries that include a set of specifications. That's helpful and usually a good starting point. But we almost always have follow-up questions afterward.

Not because we want to make things more complicated. And not because the requirements specification is poor, either. Rather, it’s because a requirements specification often doesn’t yet contain all the information we need to derive a robust test strategy from a requirement. Typical phrasing includes, for example:

  • "The system is designed to detect unusual noises."

  • "Defective bearings must be identified."

  • "The inspection must be performed within the cycle time."

  • "Measurement data should be stored."

  • "Good/Bad rating required."

These are important points. But we need a more focused approach for our testing strategy. For example:

  • What exactly is an unusual noise?

  • How does a faulty bearing show up in the measurement signal?

  • Is the error even measurable under the relevant operating conditions?

  • What is the difference between good parts, defective parts, and borderline cases?

  • How much natural dispersion does the product provide?

  • What decision should actually be made based on the signal later on?

If these issues remain unresolved, you can, of course, still work with assumptions. But in projects, assumptions are like little pebbles in your shoe: at first, you hardly notice them. Later, you wonder why every step feels more uncomfortable.

What I Mean by“Understanding the Client’sAssignment”

For me, a good EOL project starts with the question:

"What problem are we actually trying to solve?"

To understand the customer's task, we examine the product, the process, the quality objective, and the available data. This transforms a described requirement into a verifiable task.

1. Understanding the Product

Electric motors, gearboxes, pumps, compressors, and actuators each have their own operating and fault characteristics.

Some products deal with rotational speed and modes. Others deal with pressure pulsations, switching operations, bearing noise, resonances, or short pulses. That is why we must first understand:

  • What does the product do?

  • What options are available?

  • Under what conditions is the test conducted?

  • Which errors are critical?

  • When does the error actually occur?

A test system evaluates a specific behavior under specific conditions. Or to put it more simply: If I want to know whether an engine is running smoothly, I also need to know when and how I run it.

2. Understanding the Testing Process

The next point is the process. Where is the test stand located in the production line? What happens before it? What happens after it? How is the part clamped? How stable are the speed, load, temperature, and pressure? Especially for acoustic and vibration tests, the environment plays a bigger role than one might initially think. Another clamping point, a loose cover, an air pulse, a neighboring test stand, or a vibrating shop floor can all affect the signal. We need to be aware of these influences early on.

A test bench must provide reliable measurements under real production conditions. That is why we ask, for example:

  • How is the part mounted?

  • Is the testing procedure always the same?

  • Is there any noise or external vibration?

  • Is the measurement taken during startup, at the operating point, or during shutdown?

  • Will there be enough time to ensure stable measurement conditions?

These questions determine whether the measurement will be reliable later on.

3. Understanding the Quality Goal

Then we need to clarify why the testing is being done. Is it for safety? For comfort? For functionality? To prevent complaints? For process monitoring? Or to meet a specific customer requirement? That makes a difference. A noise that the end customer hears later and complains about is evaluated differently than a characteristic that’s used only internally for process monitoring. A safety-related defect is different from a comfort issue. That’s why we need to know:

  • What happens to a defective part?

  • Will it be released, rechecked, or reworked?

  • Are there any borderline cases?

  • Who makes the decision when the results are inconclusive?

  • What documentation is required?

This is often where it becomes clear whether all parties involved are truly working toward the same goal. Quality management, test bench supervisors, production, and engineering should work together to clarify these issues early on.

4. Understanding the Data Landscape

Finally, there’s one point that’s often underestimated: the data. If measurement data is already available, that’s very helpful. If there are also known good parts, bad parts, and borderline cases, that makes it even better. After all, a lot can be deduced from data:

  • What is the difference between good and defective parts?

  • Which characteristics are stable?

  • What is the typical range?

  • What types of malfunctions occur regularly?

  • What kind of evaluation method could work reliably?

If measurement data is missing, we can plan a preliminary investigation, measure sample parts, or propose a phased approach. It only becomes difficult when everyone believes the data is clear—and it later turns out there are three Excel rows, two gut feelings, and an old cell phone video with background noise from the factory floor. Such clues help us find our bearings. For a robust testing strategy, we need meaningful measurement data.

"Conspicuous" is not an evaluation criterion

“That part sounds off.” I can certainly relate to that statement. It’s completely understandable from a human perspective. When someone has experience with their product, they often hear or sense very quickly that something isn’t right. But that’s not enough for an automated testing system. After all, “off” is, first and foremost, a perception. The testing system needs to turn that into a measurable characteristic. That is the actual translation step in requirements analysis.

A few examples:

Customer Testimonial

What might be behind it

Verifiable characteristic

"The engine is whistling"

Tonal content at a specific rotational speed

Level in a frequency band

"The camp sounds rough"

Broadband or pulsed vibration components

Parameter derived from the vibration signal

"The transmission is rattling"

Periodic anomaly dependent on rotational speed

Order analysis

"The pump sounds different."

Abnormal pattern during operation

Comparison of Signatures

"That thing really stands out."

Still unclear

Analyze first, then evaluate

We need to derive a measurable test requirement from the description “noticeable.” To do this, we will clarify the following together:

  • What causes this abnormality?

  • When does it occur?

  • Is it reproducible?

  • Is it visible in the signal?

  • Can it be distinguished from normal dispersion?

  • Can it be measured under standard conditions?

This hard work will save you trouble later on.

How a defect pattern Becomes a Test Characteristic

The process of turning a defect pattern into a test characteristic usually involves several steps.

  1. Record the error description: What is observed? Who noticed it? When does it occur?

  2. Review sample parts or data: Are there good parts, defective parts, or borderline cases?

  3. Define operating conditions: Is the measurement taken at a constant speed, during acceleration, under load, or at idle?

  4. Check the sensors and measurement position: Can the relevant signal even be detected clearly?

  5. Identifying Stable Signal Characteristics: How can the error pattern be reliably identified across different parts?

  6. Deriving the evaluation logic: threshold value, attribute combination, pattern matching, or another approach?

  7. Validate using actual parts or data.

This approach leads from an initial assessment to transparent, validated criteria for automated end-of-life decisions.

Common Misconceptions About Evaluation Criteria

Terms that are interpreted differently often lead to difficulties. There are three misunderstandings that I see particularly often.

Misconception 1: “We only need one threshold value”

Threshold values are convenient. They are clear, easy to explain, and easy to document. But not every type of error can be neatly resolved with a single threshold value.

Why?

Because real-world products vary. Because different variants behave differently. Because temperature, speed, load, or installation conditions affect the signal. And because a good part can sometimes be louder than a bad part—just for a different reason.

A threshold value can work very well. But it must be appropriate for the failure pattern. Sometimes multiple characteristics are needed. Sometimes variant-specific thresholds are required. Sometimes the analysis also shows that the characteristic originally assumed is not robust enough. Recognizing this early on helps us find a more suitable characteristic in a timely manner.

Misconception 2: “One defective part is enough as a reference”

A single defective part is better than no defective parts at all. But it’s rarely enough to establish a robust inspection process. After all, even defective parts can vary. Bearing damage can vary in severity. An assembly error can manifest differently depending on operating conditions. A noise can sometimes be very clear and other times barely noticeable. If we have only a single example, there is a risk that we will recognize that exact part well—but not the full range of possible failure patterns.

Better options are:

  • unambiguous good parts

  • obviously defective parts

  • Borderline cases

  • various versions

  • Parts from actual production

The procurement of these parts should be addressed early in the planning process.

Misconception 3: “We’ll figure that out later”

Fundamental questions belong in the requirements analysis. In real-world testing, fine-tuning and clarification of new observations follow.

For example:

  • What is a defective part?

  • Which error scenario should actually be covered?

  • Under what conditions does the error occur?

  • What data or sources support this?

  • What constitutes a passing grade?

If these questions aren't answered in time, a technical implementation can quickly turn into a workshop under time pressure. And workshops under time pressure rarely have the charm of a good Rhineland evening. They're more like a construction site just before it rains.

What Anovis Can Do at This Stage

Simcenter Anovis—or Anovis for short—helps turn an idea into a testable task as early as the requirements analysis phase. For example, by:

  • Analysis of Existing Measurement Data

  • Comparison of Good and Defective Parts

  • Search for relevant signal characteristics

  • Evaluation of Possible Sensor Locations

  • Developing a Sound Testing Strategy

  • Derivation of a Robust System Configuration

What's important to me is this: Anovis does not replace a shared understanding of the test task. Anovis helps make that understanding technically tangible.

If a customer says, “This part sounds rough,” we can use measurement data to check whether this roughness is evident in the signal. Perhaps in a specific frequency range. Perhaps as impulse-like behavior. Perhaps depending on the rotational speed. Or perhaps it isn’t stable enough for a reliable EOL assessment. That, too, is a valuable finding.

A good proposal outlines the technical capabilities, the necessary requirements, and the risks.

Self-Assessment: How Clear Is Your Audit Task?

If you are currently preparing an EOL project, the following questions will help you quickly assess how well your testing requirements have already been defined. Answer each question individually with: Yes, Partially, or No.

  1. Can you describe the exam question in one sentence?

    For example:

    "The system is designed to detect error Z in product variant X at the defined operating point Y."

    If the description gets longer than the beginning of a novel, that's a sign that there's still a need for clarification here.

  2. Is the failure pattern described in technical terms?

    What happens physically, and how could it be measured?

  3. Are there any reference parts or reliable measurement data?

    Good parts, defective parts, and borderline cases are extremely helpful—especially when they come from real-world production conditions.

  4. Is the normal product distribution known?

    Only those who are familiar with the normal range of variation can determine what is truly unusual.

  5. Is the relevant operating condition clear?

    Measurements must be taken under the operating conditions in which the fault occurs.

  6. Have borderline cases been discussed from a technical perspective?

    Borderline cases show whether the testing strategy is truly effective.

  7. Is it clear who approves the technical evaluation?

    In the end, we need someone to say, “Yes, this rating aligns with our quality goal.”

Evaluation

If you can answer most of the questions with “Yes,” your test case is already well prepared. If many of your answers are “Partially,” that’s completely normal. That’s exactly what the requirements analysis is for. If several answers are “No,” you should refine the test case before configuring the system. This will save you time, money, and frustration later on. Or, in short: It’s better to answer a few uncomfortable questions at the beginning than to look for expensive answers on the test bench later.

Conclusion: Clarity First, Then Configuration

A good EOL test system requires that the test task be understood, described, and translated into measurable characteristics. During the requirements analysis, a request is transformed into a robust project.

We need to clarify:

  • What is to be identified?

  • Why is this relevant?

  • How does this show up in the signal?

  • Under what conditions are the measurements taken?

  • What data is available?

  • What decision should the system make?

  • Do all the people involved really understand the same thing?

Once these questions have been thoroughly answered, the project is well prepared. The goal: a test bench that operates reliably.

Did you enjoy this post? Stay up to date!

Our marketing team says that a blog thrives on its serial nature. That's why you'll find a new post here every second Monday of the month!

Don't want to miss a thing? Follow me on LinkedIn to learn more about the latest news and trends in quality assurance, EOL, and acoustic test benches.

tc.png

Tobias Clauß

Sales engineer

Close cookie consent banner

Share website analytics data

We use analytics cookies to evaluate the use of our website and improve our content. In doing so, personal data is processed by Google Ireland Limited (e.g., IP address, usage behavior) and may be transferred to the United States.

We use cookies for

Language selection