Medical Device Link.

A related issue arises when an electronic device is especially successful in finding willing users. If a product becomes popular, then it may very well be adapted to other uses. Here, a classic example is the beeper, or portable electronic pager. Originally used by service technicians, emergency response personnel, and the like, pagers worked well and reliably for those professionals. But a new use environment emerged when teenagers became users of pagers, one where the stresses were greater. The question raised by such a scenario is this: How can a product be validated for use in its environment when the stresses experienced by the product will change if it becomes popular and is put to new uses?

Finally, consider the fact that most medical electronic equipment is actually more than just electronic. Equipment can include solid-state electronics; mechanical devices such as printers, screens, and lids; and power circuitry, including batteries and chargers. Testing a solid-state circuit board for its ability to tolerate expected conditions requires one set of procedures, while a battery or printer may have to be subjected to a different set of tests. Could a system be developed that adequately tests diverse equipment elements simultaneously?

How can a medical electronic device be validated for use in potential stress environments that have not or cannot be anticipated? The answer is to focus the testing on conditions that can break the product rather than on what might happen under the conditions it is expected to encounter. This article describes and discusses a technique and test system that uses this approach and produces quantitative design validations in just a few days.

These stresses are applied so they randomize their effects. For example, vibration would be delivered in a random six-axis profile, mechanical loads would be randomly relative to each other, and electrical loads would be randomized. The object is to apply all potential stresses simultaneously and unsystematically in order to distribute stress randomly throughout the EUT. The rate at which different failure modes accelerate in response to different stress sources is well documented.¹

The FMVT process begins with the EUT exposed to stresses that simulate the maximum expectable service conditions. Over time, the stress levels are raised. And as the test progresses, the EUT experiences failures. These failures occur at locations in the device design that accumulate stress damage faster than other areas of the product.

The failure occurrences pinpoint the weak links in the design of the tested equipment. Because all of the known types of potential stress are applied to the device randomly and with intensity increasing over time, the order of occurrence of the failure modes is generally the order of their relative significance. The stress sources applied are raised in proportion to two reference points specific to the product—the service conditions (what the product would experience) and the destruct level (how much of a particular stress will destroy the product in a short period of time)—ensuring that the damage done by each type of stress is maintained approximately in proportion. The time to failure, therefore, can be taken to be in proportion, and the resulting failure-mode progression is representative of the order of failures that would be seen in the field.²

Figure 1. Plot of the failure-mode progression for an imaginary electronic product with a relatively immature design. The design maturity of the device is 0.42.

The distinct failure modes resulting from FMVT can be plotted as shown in Figure 1.

If the first failure mode occurs early in the test period and notably before the rest of the failure modes, then the product is immature and has plenty of room for improvement. If all of the failure modes appear only after a significant period of FMVT time and in close succession, then the product is mature. Figure 2 indicates a mature design as revealed by results of FMVT. If the first failure modes of this product were addressed in redesign, the equipment's projected operational life would not change perceptibly because the next failure mode occurred at nearly the same time as the first ones. Addressing the first failure mode in the device represented in Figure 1, however, would result in a significant improvement in that product. A product's potential for improvement can be quantified as its design maturity.

Figure 2. Plot of the failure-mode progression of a device with a mature design. The design maturity of this device is 0.02.

Design maturity (DM) is represented quantitatively as the average time between failures after the first failure divided by the time to the first failure. This calculation yields the average potential improvement in product life under the accelerated test that would be gained by fixing one failure mode. The device in Figure 1, which has a DM of 0.42, offers a potential improvement in life of 42% with the elimination of one failure mode, whereas the product in Figure 2 (DM = 0.02) has a potential for improvement of only 2%. Clearly, the second design is more mature; that is, it has less room for improvement.

Figure 3 shows the failure-mode progressions of a group of products that are assembled into a system. Although all of these system components have a significant potential for improvement (37 to 60%), they do not share a route to achievement of maturity. Product 3, for example, has a relatively large improvement potential of 49%, but it can achieve a lifetime in the accelerated stress environment of more than 300 minutes if just one failure mode is fixed. By comparison, product 4, which has a 60% potential for improvement, would only reach a life in the accelerated environment of around 175 minutes if one failure mode were fixed. Therefore, a quantification of the limit of a system's design potential is also needed. Such quantification gives the technological limit (TL) of the product.

Figure 3. Plot of the failure-mode progressions of four devices assembled into a system. The connecting line indicates the technological limits of the system components. If all these devices were iterated to achieve their TLs, the system DM would be 0.13.

The TL calculation involves removing failure modes in the order of their occurrence and each time recomputing the DM until it is less than 0.1. Once the DM drops below 0.1, the time of the first remaining failure mode—that is, the point at which it appears in the FMVT process—is the technological limit. In the case of product 1 in Figure 3, eliminating the first failure mode would result in a DM of 0.06. Therefore, the TL of product 1 is the time of its second failure mode, or 250 minutes. The TLs of components 1 through 4 of the system are 250, 240, 310, and 220, respectively.

Regarding TL as a function of DM, then, as a product's design is iterated, its technological limit is the best it is expected to become. In refining an electronic system, it is worthwhile only to address remaining failure modes that are below the system technological limit. How well balanced are the functional lifetimes of the components of a system if all the components are brought up to their technological limit? The component TL can be used to calculate the system design maturity. In the case of the system in Figure 3, it would be 0.13, meaning that replacing the component with the worst TL (product 4) with one not determined to be the first system element to fail would potentially improve the life of the system by 13%, assuming that all of the products constituting the system were first iterated to their technological limit.

Another example involves a network hub that had passed all required validation tests but nevertheless had some inherent design weaknesses that were exposed in the field. Running the hub through the FMVT process reproduced those failure modes. FMVT turned up the following key facts about a product that had already been released to the market:

The design maturity of the network hub is 31.8. This indicates that the design could potentially improve by 3180% if one failure mode were fixed. This calculation is based on all of the failure modes found, both severe and minor. Two of the failure modes were soft failures, however (failures causing a temporary interruption of proper functioning without damaging the product). Taking just the hard failure modes into account, the DM of the hub is 63.5, meaning that fixing one failure mode offers a potential of improving the design by 6350%. The DM value is high because the first failure mode appeared only 187 seconds into the test. The assessment of risk relating to this first failure mode was low; although a hard failure, a simple change in the design of a power connection would eliminate the failure mode and make it easy to avoid in the field. If this source of failure were fixed or ignored, the DM of the network hub, calculated on hard failures, drops to 0.16; the design has a potential of improving by 16% as a result of fixing one failure mode.

An important fact to note about the FMVT trial of the hub is that its hard failure modes were all mechanical (the two soft failures were electronic); the real damage to the unit was sustained by its larger mechanical parts—the power connector, network cable connector, and clamshell. How could such design flaws have been missed during development testing? The answer is because the solid-state electronics of the product were screened using air hammer technology. Air hammer vibration does not produce significant low-frequency energy. Also, because the main objective of using an air hammer vibration machine on an electronic component is to find any failure modes in the electronics, the circuit board is typically mounted directly to the table, without the plastic housing. For this reason, failure modes associated with low-frequency resonance—involving items such as connectors, clips, and cases or clamshells— would not be found. It is not surprising then that an electronic product that was developed with the use of air hammer technology had significant mechanical failure modes but no hard electronics failures.

Using FMVT to iterate a product design before doing certification testing can avoid repeating that testing. Because products often fail some of the tests, such repetition is common, which can be an expensive way to iterate a design. FMVT allows a manufacturer to plan on one pass through the certification testing.

FMVT has been used on a variety of products ranging from medical electronics to automotive components to kitchen utilities. In addition to offering flexibility and speed, FMVT has been able to reproduce failure modes from warranty and field incidents that other tests could not reproduce.