CQM is helping Philips predict the lifetime of LED lighting. That’s the hardware bit, but what about the 'reliability' or lifelong behavior of complete systems, which of course includes software and the interaction between hardware and software? To what extent does one affect the other? Few tools are yet available when it comes to revealing this overall picture. In fact, there’s even a serious shortage of theoretical frameworks. Which is why Philips Lighting and CQM are together doing pioneering work in mathematically modelling the reliability of entire electronic systems.
Almost everything in the world is directly or indirectly controlled by electronic systems. And because our dependence on these systems is growing, it’s increasingly important to be able to quantify and predict their reliability under various conditions.
CQM's senior consultant, Marc Schuld, is an expert in this area. "Any product that consists of just two interacting components is in fact a system. The more components and types of components that are added to the product, the more interactions there are, and the more complex the system becomes. And the more complex a system is, the more difficult it is also to predict the reliability.”
Looking at software-reliablility differently
Generally speaking, quite a bit is already known about the lifetime behavior of hardware. But it’s different with software. “There are no obvious causes for software defects, such as variations in the creative process, wear and tear, or physical stress from working conditions such as temperature, humidity or vibration. Software defects arise precisely from those things that are not directly visible or foreseeable. For example, a misinterpretation of the design specifications; mistakes when writing the code; inadequate or improper testing; memory leaks; or incorrect or unexpected use of the software. All defects that, in contrast to hardware, are not functions of usage. And that means you have to approach software reliability differently."
CQM quantifies error detection and debugging process and ties in with software development
"Even the best engineers, with the best knowledge and equipment, can’t prevent software defects. What you can do is conduct tests on the software designed to answer certain key questions. For example, how likely is it that there are serious errors in the software? And, going a little deeper, what is the likelihood of errors occurring while the software is being used within a given period by users with pre-specified user profiles? What matters most is that you can quantify these answers. And you have to include the uncertainty of the answers in the confidence intervals. From which the project manager can decide, on the basis of the predetermined reliability objective, whether the software is ready for release or requires more testing. Together with the Philips Lighting Reliability Manager, Willem van Driel, we’ve now developed a methodology for retrieving the data with which such questions can be answered. For this, we estimate software reliability growth models (SRGs) that describe the error detection and debugging process. This methodology is tied in with the software development process."
Quantitatively predict interactions between hardware and software
When predicting the reliability of complete systems, it’s not just about the individual hardware and software. The many interactions between the two also play an important part. For example, obsolete hardware can lead to software defects, and conversely software can accelerate the deterioration of hardware. "Let’s return to the example of Philips' LED systems. The software keeps the light output of a LED light bulb at the right level by increasing the current throughout the product’s life. But the higher thermal load may, for example, cause the solder joints to degrade, causing the lamp to fail earlier. The interactions between hardware and software must therefore also be quantitatively predicted in detail and included in the overall reliability. In other words, complex and still relatively unchartered territory.”
Would you like to know more about reliability?
Would you like to know more about lifetime predictions of hardware ór software? Please feel free to contact Bert Schriever of Marc Schuld.
