JMPer Cable

siminatjmp

Imagine you're a quality engineer developing a new product. Waiting until products actually fail could take months or even years, making it difficult to evaluate reliability early in development. Fortunately, many products degrade gradually over time. For example, a phone battery slowly loses its capacity after repeated charging cycles, and a pair of running shoes gradually loses cushioning with continued use. By measuring this degradation, you can gain valuable insight into when failures are likely to occur.

After hours – or even days – of collecting and analyzing degradation data, you're ready to assess and predict system reliability. Naturally, several questions may arise:

First, how do you determine what type of data you are working with?
Next, which type of reliability analysis is appropriate for your data?
And most important, how can you efficiently unlock these analyses using JMP?

Take a breath – no need to worry. In this blog, we guide you step by step through the process. We begin by discussing different types of degradation data so that you can better understand the data you’ve collected. Next, we introduce the relevant platforms in JMP for degradation analysis. Finally, we’ll explain why – and when – to choose a specific platform to fully leverage the powerful reliability and survival analysis features available in JMP, with more detailed discussions provided in separate blog posts.

Along the way, we address the following questions:

What is degradation in general? How do we measure it?
What is degradation data?
How should covariates be incorporated into degradation analysis?
Which degradation analysis platforms are available in JMP?
How do you choose the right platform for your degradation data?

Figure 1. Overall flowchart for degradation analysis in JMP 19.

Alright, let’s begin our journey into degradation analysis.

What is degradation?

Before we dive into the technical details, let’s first build a general understanding of what we mean by “degradation.” If you’re already familiar with this term, feel free to jump ahead to the sections that focus on data and degradation analysis in JMP.

Theoretically, as noted by Condra (1993)^[1], reliability can be viewed as “quality over time.” Over the course of a product’s lifetime, its performance rarely remains constant. Instead, products often degrade gradually over time. For example, a car tire loses tread after thousands of miles, an air conditioner becomes less efficient with age, and a cutting tool gradually becomes dull with repeated use. Eventually, the product may no longer meet its intended performance requirements. This gradual decline in performance is known as degradation.

In a traditional life test, products are observed until they experience a complete failure, such as a light bulb burning out or a motor stopping. In contrast, degradation analysis monitors product performance over time and considers a product to have experienced a “soft failure” when its performance falls below a predefined threshold, even if it is still functioning. Importantly, degradation can often be observed well before a product actually experiences catastrophic failure, which can provide a valuable opportunity to understand reliability behavior earlier and more informatively than traditional failure-time data alone.

How do we measure degradation?

Now that we understand what degradation is, the next step is deciding how to measure it. First, a performance characteristic that reflects product degradation (such as battery capacity, light output, or material strength) is selected as the response variable. This characteristic is then measured repeatedly over time, with time serving as a primary predictor variable. Depending on the study, other experimental factors, such as temperature or humidity, may also be recorded or intentionally varied. Together, these measurements form degradation data.

How these measurements are collected depends on the study design. For example, some studies observe the same unit over time, whereas others measure different units under varying stress conditions. These differences in study design determine the structure of the resulting degradation data and, ultimately, the choice of analysis method.

Types of degradation data

In practice, degradation data generally fall into one of several categories. The two most common categories are:

Repeated measures degradation testing (RMDT) data.
Accelerated destructive degradation testing (ADDT) data.

Let’s take a closer look at each type

1. RMDT data

RMDT data occur when the same unit is measured repeatedly over time without being damaged. Because the measurement process is non-destructive, the unit remains functional and can be observed continuously.

For example, imagine monitoring the battery capacity of an electric vehicle. Each month, you record the remaining capacity of the same battery while it continues to operate normally. Since the battery is not destroyed during testing, you can track its gradual decline over time. This is a classic example of RMDT data.

Figure 2. Repeated measures degradation testing example.

Ask yourself: Does your data set contain multiple degradation measurements collected from the same unit over time? If the answer is yes, your data likely falls into the RMDT category. This type of data allows us to observe the degradation path of each unit in real time.

2. ADDT data

Now consider ADDT data. In this setting, the measurement process is destructive, meaning each unit can only be measured once. The test itself damages the unit, so it cannot be observed again. Instead of repeatedly measuring the same unit over time, degradation is assessed using measurements collected from different units at different time points.

RMDT is appropriate when the degradation characteristic can be measured repeatedly on the same unit without affecting its future performance. In contrast, ADDT is used when the evaluation permanently destroys the unit, making repeated measurements impossible. Since the degradation path cannot be observed directly for a single unit, it is reconstructed by combining measurements from multiple units.

For example, consider testing how the strength of an adhesive bond changes over time at higher temperatures. Several samples are exposed to the same temperature but for different lengths of time. After a specified exposure time, a sample is removed and pulled until the bond breaks. The force required to break the bond tells us how much strength it has retained. Because the bond must be broken to measure its strength, that sample cannot be tested again. A different sample is therefore needed for each exposure time. The same process can be repeated at several elevated temperatures to accelerate degradation. By combining measurements across different temperatures and exposure times, we can characterize how the bond degrades over time and predict its performance under normal-use conditions.

Figure 3. Accelerated destructive degradation testing example

Incorporating covariates in degradation models

In degradation studies, degradation may depend not only on time but also on other factors that affect the overall degradation process. When these covariates change over time, they are referred to as dynamic covariates.

Unlike static covariates, dynamic covariates capture how a unit’s characteristics evolve throughout the observation period. Their histories provide an exposure profile that can help characterize the degradation experienced by each unit.

For example, consider two vehicles operating under similar environmental conditions. Their usage characteristics, such as speed, braking frequency, and driving load, may change from day to day. One vehicle may experience frequent short trips with repeated acceleration and braking during one period, followed by longer highway trips during another, while the second vehicle may have a different usage history. Although the vehicles operate under similar environmental conditions, these different exposure histories may lead to different degradation paths. This illustrates a key feature of dynamic covariates: degradation may depend on a unit’s exposure history, rather than on a single static measurement.

Figure 4. Degradation data with dynamic covaraites example

Which platforms are available in JMP 19?

Figure 5. Degradation analysis platforms available in JMP 19.

The degradation analysis tools currently available in JMP 19 can be found under the Analyze menu, within the Reliability and Survival section. There are three main platforms available for conducting degradation analysis:

Repeated Measures Degradation: Used when the same unit is measured multiple times over time using a non-destructive process. It allows you to observe the full degradation path of each unit.
Destructive Degradation: Designed for situations where the measurement process damages or destroys the unit, meaning each unit can only be measured once.
Degradation: A more general platform that provides flexible modeling options for analyzing degradation data under different conditions.

The appropriate approach to degradation analysis depends primarily on how the degradation data are collected. Different data structures may be analyzed using different JMP platforms, with some overlap in functionality among the available platforms.

In the second blog of this series, we focus first on the Repeated Measures Degradation platform. We discuss what it is designed for, how it works, how to prepare the appropriate data for analysis, and the potential limitations of this approach.

Stay tuned! In the next blog, we dive deeper into the technical details.

References

[1] Kromholtz, G. A., & Condra, L. W. (1993). A new approach to reliability of commercial and military aerospace products: Beyond military quality/reliability standards. Quality and reliability engineering international, 9(3), 211-215.

[2] Clark, J. M., Min, J., Li, M., Warr, R. L., DeHart, S. P., King, C. B., Lu, L. & Hong, Y. (2026). What quality engineers need to know about degradation models. Quality Engineering, 1-22.