Evaluation of Measurement Uncertainty ISO/IEC 17025

Have you ever followed a recipe to the letter and still ended up with a result that wasn’t quite what you expected?

In the world of testing and calibration laboratories, we face a similar dilemma every day. Instead of dealing with ingredients, however, we are faced with measurement uncertainty.

This is where the GUM (Guide to the Expression of Uncertainty in Measurement) comes in. It acts as a foolproof recipe that helps us evaluate measurement uncertainty by following a clear path that can be applied to testing and calibration methods.

Today, we will dive into the world of metrology and discover together how to evaluate measurement uncertainty using the GUM method with practical examples. This will help us to meet the requirements of the ISO/IEC 17025 standard.

Let’s get started!

What is Measurement Uncertainty?

Imagine that you are measuring the length of your desk with a ruler. Even though you think your measurement is accurate, there will always be a small margin of error, perhaps because the ruler is slightly worn or your eyes are deceived by the perspective.

That margin of error, my friend, is measurement uncertainty: an inevitable part of any measurement process that tells us how “reliable” our results are.

The definition according to the International Vocabulary of Metrology is:

“Parameter, associated with the result of a measurement, that characterizes the dispersion of the values that could be reasonably attributed to the measurand.”

In other words:

Measurement uncertainty is an estimate of the dispersion of the possible values of the measured quantity.

It indicates the reliability of the measurement result.

It is expressed as an interval of values, with an associated confidence level.

Here are some additional points about measurement uncertainty:

• Measurement uncertainty is always present, even if we use the best instruments and techniques.
• The goal of uncertainty analysis is to quantify the uncertainty so that we can make informed decisions about our results.
• There are a number of different methods for estimating measurement uncertainty.
• The most common method is the GUM (Guide to the Expression of Uncertainty in Measurement).

The Importance of Evaluating Uncertainty

In the context of laboratories accredited to ISO/IEC 17025, evaluating uncertainty is not just good practice; it is a requirement.

This is because our clients (and ourselves) need to have full confidence in the results of tests and calibrations. Think about it: if you are assessing the safety of a bridge or the purity of a drug, “almost sure” is not good enough.

You need to be “really sure”.

Here are some of the benefits of evaluating measurement uncertainty:

• It allows us to make more informed decisions about our results.
• It helps us to identify and minimize sources of error.
• It improves the quality of our data.
• It increases customer confidence in our work.
• It helps us to meet the requirements of ISO/IEC 17025.

If you are a laboratory that is not currently evaluating measurement uncertainty, I encourage you to start doing so. It is an essential part of ensuring the quality of your work.

Demystifying the GUM Method

If you’ve ever felt overwhelmed by technical terms and meticulous procedures, you’re not alone. But we’re here to demystify this process and make it as friendly as a conversation about your favorite TV show.

Definition and Basic Principles

The GUM Method, or as we like to call it, the “Great Book of Uncertainty”, is simply a guide that teaches us how to evaluate and express uncertainty in our measurements systematically.

Imagine trying to follow a new recipe without knowing how much a “pinch” of salt is. The GUM guide helps us define exactly how much that “pinch” is in the context of our measurements.

Measurement Uncertainty Components

In the world of measurement, not everything is black and white; there are many shades of gray. This is where the components of uncertainty come in:

Type A Uncertainty: This is determined using statistical methods. Think of it like when you try to estimate how many times you will need to hit a piñata before it breaks based on your previous attempts.

Type B Uncertainty: This is obtained by any means other than statistical. It’s like trying to guess how many candies are in a piñata based on its size and the sound it makes when you shake it.

Types of Uncertainty: Type A and Type B

Let’s delve a little deeper into these two types of uncertainty:

Type A Evaluation: This is based on the observed variability of the results of repeated measurements. It is our attempt to apply the scientific method to make predictions based on real data.

Type B Evaluation: This is where we put on our detective hats and gather evidence from all possible sources, except for direct experimental data. This could include manufacturer specifications, previous references, or even textbooks.

For laboratories operating under the ISO/IEC 17025 standard, mastering these components is crucial. It not only ensures that we comply with international standards, but also gives us the confidence to defend our results to clients or any interested party.

Mastering Type A Uncertainty

Now that we are familiar with the basics of the GUM Method and know that there are two types of uncertainty, it is time to put on our gloves and enter the world of Type A uncertainty.

This is the part where things get a little more “mathematical”, but don’t worry, we will take it step by step.

Statistical Methods for Type A Uncertainty Evaluation

Type A uncertainty is our statistical friend. It is calculated from the variability of repeated measurements.

Imagine you are throwing darts and recording where they hit the board. Some will be closer to the center, others further away, but if you throw enough darts, you will start to see a pattern.

This pattern is what helps us calculate Type A uncertainty.

Data Collection:

First, you need data. The more measurements you have, the better. This is like having more attempts to throw those darts.

Statistical Analysis:

Then, we use statistical tools to analyze this data. We calculate the mean, the standard deviation, and from there, the Type A uncertainty. If these words sound like Chinese to you, think of the standard deviation as a way to measure how much your darts are scattered around the center.

Experimental Data Analysis

This step is crucial. It is not enough to just collect data; we need to understand it. This is where statistical analysis comes into play:

1. Mean: This is basically the middle point of your measurements. If you threw darts at 5, 6, 7, 6, and 5 meters, your mean would be 5.8 meters.

2. Standard Deviation: This tells us how much your measurements vary from the mean. A small standard deviation means that your darts landed close to each other (and to the mean), which is good.

Practical Calculation Examples

Let’s put this into context with a simple example. Let’s say you are measuring the length of the same metal piece several times. Your measurements are 10.1, 10.2, 10.2, 10.3, and 10.2 cm.

1. Calculating the Mean:

We add all the measurements and divide by the number of measurements:

(10.1 + 10.2 + 10.2 + 10.3 + 10.2) / 5 = 10.2 cm

2. Calculating the Standard Deviation:

We use a statistical formula to calculate the standard deviation:

Standard Deviation = 0.06 cm

3. Obtaining the Type A Uncertainty:

The Type A uncertainty is calculated from the standard deviation:

Type A Uncertainty = Standard Deviation / √n

Type A Uncertainty = 0.06 cm / √5 = 0.04 cm

For laboratories under the ISO/IEC 17025 standard, this process is not only an exercise in compliance, but also a demonstration of competence and reliability. Your clients, and in fact anyone interested in your results, can be confident that what you report is backed by rigorous and transparent analysis.

Navigating Type B Uncertainty

Having taken a look at Type A uncertainty in our previous section, it is time to dive into the mysterious world of Type B uncertainty. Are you ready to become an uncertainty detective? Let’s go!

Sources of Type B Uncertainty

Type B uncertainty is like that box of chocolates where you never know what you’re going to get. It is not based on the data you collect, but on everything else. Here are some of the most common sources:

1. Manufacturer’s specifications: When the manufacturer of your equipment tells you that its accuracy is ±0.1%, that goes straight into your Type B uncertainty budget.

2. Published data: Those values you find in textbooks or scientific articles also count.

3. Previous experience: Yes, your experience and that of your team are also valid. If you know that a certain type of sample always gives you problems, that is something to consider.

4. Other factors: Temperature, humidity, electromagnetic interference… if they affect your measurements, they also come into play here.

Estimation Methods for Type B Uncertainty

This is where we put on our wizard hats and transform that qualitative information into quantitative estimates. It’s not black magic, but almost:

1. Use specifications directly: If the manufacturer says ±0.1%, that’s what you use.
2. Convert qualitative data to quantitative data: This may involve some detective work. For example, if a scientific article describes a range of values, you can use those ranges to estimate your uncertainty.
3. Worst-case scenario analysis: Imagine the worst thing that could happen (within reason) and use that for your estimate.

Applications and Evaluation Examples

Let’s put this into context with a real-world example. Imagine you are calibrating a scale that, according to the manufacturer, has an accuracy of ±0.1 g. Additionally, you know from experience that the temperature variability in your laboratory can affect the measurement by ±0.05 g.

Here is how you could proceed:

1. Manufacturer’s specification:

• The accuracy stated by the manufacturer of ±0.1 g is directly added to your Type B uncertainty budget.

2. Temperature influence:

• Based on your temperature records and how they affect measurements, you estimate an additional ±0.05 g of Type B uncertainty.

3. Combining uncertainties:

• In the end, you would combine these Type B uncertainty sources with your Type A uncertainties (obtained from repeated measurements) to obtain a total uncertainty estimate.

This total uncertainty would accurately reflect the confidence you can have in your measurement results.

Combining Uncertainties: The Art of Accuracy

Having explored the intricacies of Type A and Type B uncertainties, it is time to get our hands dirty and learn how to combine these puzzle pieces to get a complete picture. This, my friend, is where the magic really happens.

Methods for Combining Type A and Type B Uncertainties

Think of combining uncertainties like making a smoothie. You have your fruits (Type A uncertainty), your milk or water (Type B uncertainty), and you need to mix them together to get that delicious final result.

This is where the root sum square (RSS), also known as the quadratic combination rule, comes into play. Don’t worry, it’s not as complicated as it sounds:

1. Square the values:

• Take each of your uncertainties, both Type A and Type B, and square them.

2. Add the squares:

• Add all those squared values together.

3. Square root of the sum:

• Take the square root of that total sum, and there you have it, your combined uncertainty.

Calculation of Combined Uncertainty

Let’s put this into practice with a simple example. Suppose you have:

• Type A uncertainty: 0.05 units
• Type B uncertainty 1: 0.02 units
• Type B uncertainty 2: 0.03 units

Here’s how you would calculate the combined uncertainty:

1. Squares:

• Square each of the uncertainties:

• 0.05^2 = 0.0025
• 0.02^2 = 0.0004
• 0.03^2 = 0.0009

2. Sum:

• Add all the squared values:

• 0.0025 + 0.0004 + 0.0009 = 0.0038

3. Square root:

• Take the square root of the sum:

• √0.0038 = 0.0616 units

Result:

• Combined uncertainty: 0.062 units

Your combined uncertainty is approximately 0.062 units. This means you have a more accurate and reliable estimate of the true uncertainty associated with your measurement.

Expanded Uncertainty: Expanding Our Confidence

After combining Type A and Type B uncertainties to obtain the combined uncertainty, the next step is to calculate the expanded uncertainty.

This step is essential to provide a wider confidence interval around the measurement, which gives us a better estimate of how “confident” we can be about our results.

What is Expanded Uncertainty?

Expanded uncertainty is defined as the interval within which the true value of a measurement is expected to lie, with a given level of confidence.

It is calculated by multiplying the combined uncertainty (which represents the variability of the measurement) by a coverage factor (k), which reflects the desired confidence level.

In general, the confidence level is set at 95%, which means that there is a 95% probability that the true value lies within the expanded uncertainty interval. However, this level can vary depending on the specific requirements of the analysis or the applicable regulations.

Coverage Factor (k):

• The value of k is selected based on the desired confidence level. For a confidence level of 95%, a value of k = 2 is used.
• Higher values of k (such as k = 3) are used to obtain wider uncertainty intervals with a higher confidence level (99.7%).
• Lower values of k (such as k = 1) are used to obtain narrower uncertainty intervals, but with a lower confidence level (68%).

Example:

Suppose the combined uncertainty of a measurement is 0.05 units and a confidence level of 95% is desired.

• Expanded uncertainty = 0.05 units * 2 = 0.1 units

In this case, we can be 95% confident that the true value of the measurement lies within the interval of 0.1 units above and below the measured value.

In summary:

• Expanded uncertainty is a measure of the accuracy of a measurement.
• The larger the expanded uncertainty, the lower the accuracy of the measurement.
• It is important to choose the appropriate confidence level for the task at hand.

Expanded Uncertainty Calculation and Interpretation

Calculating the expanded uncertainty is straightforward once the combined uncertainty is known:

Expanded uncertainty = k × Combined uncertainty

Where:

• Expanded uncertainty is the uncertainty expressed with a level of confidence.
• k is the coverage factor.
• Combined uncertainty is the uncertainty obtained by combining all sources of uncertainty.

Practical example:

If the combined uncertainty of a measurement is ±0.062 units and a 95% confidence interval is desired, using a coverage factor of 2, the expanded uncertainty would be:

2 × ±0.062 = ±0.124 units.

Interpretation:

• ±0.124 units indicates that the true value of the measurement lies within the interval of 0.124 units above and below the measured value, with a confidence of 95%.
• In other words, we can be 95% confident that the true value of the measurement lies between (measured value – 0.124) and (measured value + 0.124).

Importance of Expanded Uncertainty in ISO/IEC 17025

For laboratories following the ISO/IEC 17025 standard, evaluating and reporting expanded uncertainty is fundamental.

This value provides clients and stakeholders with a clear understanding of the level of confidence they can have in the laboratory’s results, which is crucial for making informed decisions based on those results.

Here are some key points about the importance of expanded uncertainty in ISO/IEC 17025:

1. Transparency: Expanded uncertainty allows clients and stakeholders to know the accuracy of the laboratory’s results. This makes it easier to compare results between different laboratories and assess the suitability of the results for a specific purpose.

2. Confidence: Expanded uncertainty provides a solid basis for decision-making. By knowing the accuracy of the results, clients and stakeholders can make decisions with greater confidence, knowing that the laboratory’s results are reliable.

3. Compliance: The ISO/IEC 17025 standard requires laboratories to evaluate and report expanded uncertainty. This means that laboratories that do not comply with this requirement are not complying with the standard.

4. Continuous improvement: Evaluating expanded uncertainty can help laboratories identify areas where they can improve their processes and increase the accuracy of their results.

How to Express Uncertainty in Measurement Results

Think about the last time you explained something complicated. Maybe you used an analogy or a story to make it more accessible. With measurement uncertainty, we do something similar, but instead of stories, we use standardized language and format that anyone in the scientific community can understand.

Be clear and concise: When expressing uncertainty, your goal is to be as clear and direct as possible. For example, if your measurement result is 10 units with an expanded uncertainty of ±0.5 units, you can express it as: 10 ±0.5 units.

Include a confidence level: We normally express uncertainty with a 95% confidence level. This means that, based on your measurement and the calculated uncertainty, you are 95% confident that the true value lies within your expressed range.

Include the coverage factor K: It is not enough to report the confidence level; you must also include the value of the coverage factor K.

Use the correct units: Make sure that the uncertainty is expressed in the same units as your measurement, or in relative terms if that is more relevant.

Here are some additional tips for expressing uncertainty:

• Use the symbol “±” to indicate the expanded uncertainty.
• Do not use the terms “error” or “accuracy” when referring to uncertainty.
• Be consistent in how you express uncertainty throughout your work.
• If you are unsure about how to express uncertainty, consult with a qualified expert.

By following these tips, you can ensure that your uncertainty is expressed in a clear, concise, and accurate manner. This will help others to understand your results and make informed decisions based on them.

Interpreting Results and Making Comparisons

Once you have expressed your uncertainty, the next step is to interpret what it means in the context of your work.

For example, if you are comparing your results with those of another study or with a reference value, the uncertainty allows you to see if there is really a significant difference or if any variation could simply be attributed to the margin of error.

Significant comparisons: Use uncertainty to determine whether your results are consistent with expected values or with the results of others. If the uncertainty ranges overlap, there may not be a statistically significant difference.

Uncertainty-based decisions: In some cases, the magnitude of your uncertainty may influence the decisions you make next. For example, if the uncertainty is too high, you may need to improve your methodology or perform additional measurements.

Here are some additional tips for interpreting uncertainty:

• Consider the purpose of your measurement and the level of accuracy that is required.
• Compare your results with other relevant data and studies.
• Be aware of the limitations of your measurement methods.
• Use statistical methods to assess the significance of any differences.
• Document your uncertainty and how you interpreted it.

By following these tips, you can ensure that you interpret your uncertainty in a way that is meaningful and accurate. This will help you to make informed decisions based on your results.

Practical Applications of Uncertainty in the Laboratory

Now that we have covered the journey from understanding measurement uncertainty to combining and expressing it correctly, it’s time to see how all this applies to the daily operations of a laboratory following the ISO/IEC 17025 standard. This is where theory meets practice.

Case Studies in Diverse Disciplines

Each discipline has its own challenges and quirks when it comes to measurement and calibration. Here we will explore some concrete examples that showcase how uncertainty evaluation is applied in different contexts:

Engineering: Imagine you are designing a component that needs to fit perfectly into a larger system. Accuracy is key, and the uncertainty of your measurements will determine how reliably the component can be manufactured and used.

Chemistry: Consider the task of measuring the concentration of a solution. Variations in temperature, the purity of reagents, and the accuracy of your measurement instruments can all influence the final result.

Physics: In experiments where physical phenomena are measured, such as the speed of light or the strength of a magnetic field, understanding and expressing uncertainty is crucial for validating your findings.

Other Practical Applications:

• Calibration of instruments: Uncertainty plays a vital role in ensuring that measurement instruments are calibrated correctly and traceable to national or international standards.

• Quality control: Laboratories use uncertainty to monitor the quality of their measurements and ensure that they are meeting the requirements of the ISO/IEC 17025 standard.

• Data analysis: Uncertainty is an important factor to consider when analyzing data and drawing conclusions from experiments.

• Risk assessment: Uncertainty can be used to assess the risks associated with a measurement or decision-making process.

By understanding and applying uncertainty in the laboratory, scientists and engineers can ensure that their measurements are accurate, reliable, and traceable. This helps to improve the quality of data, reduce risks, and make better decisions.

Common Challenges and Solutions When Using Measurement Uncertainty

Let’s face it, measurement is never as straightforward as we would like it to be. Here are some common challenges and how uncertainty can help address them:

Outdated Equipment: Even older equipment can still be useful if we understand its uncertainty and take it into account in our measurements.

Environmental Variables: Learning to measure and compensate for the influence of factors such as temperature and humidity can significantly improve the accuracy of your results.

Data Interpretation: Sometimes, the challenge lies in the numbers. Applying uncertainty correctly allows you to make sense of your data, even when it initially seems confusing.

Other Common Challenges:

• Human error: Uncertainty can help to identify and mitigate the effects of human error in measurement.
• Limited resources: Uncertainty can help you to prioritize your resources and focus on the measurements that are most important.
• Complex measurements: Uncertainty can help you to break down complex measurements into smaller, more manageable steps.

By understanding and addressing these challenges, you can use measurement uncertainty to improve the accuracy, reliability, and traceability of your measurements. This will help you to make better decisions, improve the quality of your work, and reduce risks.

In Conclusion:

We have come a long way together, from introducing measurement uncertainty to the practical application of the GUM Method in the laboratory. Throughout this journey, we have unravelled the mysteries of Type A and Type B uncertainties, learned how to combine them, and discovered how to express them in a way that is clear and useful for everyone.

But more importantly, we have seen how these concepts are not just academic exercises, but essential tools in the arsenal of any laboratory that seeks to comply with the ISO/IEC 17025 standard.

By embracing uncertainty and using it effectively, laboratories can improve the accuracy, reliability, and traceability of their measurements. This, in turn, leads to better decision-making, improved quality of work, and reduced risks.

So, as you continue your journey in the world of measurement, remember the importance of uncertainty. It may seem like a complex concept at first, but with a little effort and understanding, it can become a powerful tool that helps you achieve your goals.