A Process for Developing Highly Reliable and Safe Devices
As we enter the third millennium, product reliability issues are becoming more complex. No longer does the manufacturer need to be concerned just about whether the product performs as advertised but also, is it safe for the user and free of adverse effects on non-users. Customer expectations spurred on by consumer advocate groups and a preponderance of product liability lawsuits present manufacturers with a tremendously challenging situation how to produce a highly reliable, durable, and safe product and yet keep the cost low enough so not to eliminate the market for it. As competitors scratch for ever-increasing market shares, product improvement is clearly a fertile area of opportunity. The concept of product improvement itself is fuzzy and often conveniently treated as an aspect of product quality. There is an enormous difference between reliability and quality. Quality does not ensure that a product is reliable or that it will meet its design life, or that it is durable and can survive extraordinary usage and application stress, or that it is safe.
To develop a product with a specific characteristic such as a designated life, that characteristic must be designed into it from the very first. It must be incorporated into the product specifications, then into the conceptual design, the preliminary design, final design, prototype test articles and finally the marketed product itself. As simple and obvious as this seems, it is rarely done this way in the world of commercial products. Most of the time, the practice is to turn the product design engineers loose to arrive at concepts, preliminary design, final design and prototypes with minimal consideration of certain desirable features and characteristics. By this time it is too late. An old axiom states that reliability (durability, robustness, safety) must be designed in, it cannot be added on later as an afterthought. But this is what happens in many cases.
Wyle Laboratories has been in the testing and engineering business for 55 years involving defense, space, naval, railroad, mining, energy, communications, and pharmaceutical. In this broad span of experience, we have witnessed the extremes of aerospace and defense on one end with meticulous processes for assuring that the system (product) does have all of the features/characteristics that it was specified to have, to some commercial producers who bring their finished product to us to find out what they have. For the latter case, again, it's too late.
Some industries are held to higher standards. The aircraft industry must meet specific standards by showing that certain Federal Aviation Administration regulations have been met. The same is generally true for the medical and pharmaceutical devices industry with the Food and Drug Administration.
A reliability program for a product must be uniquely tailored to its customers' expectations, which are market-dependent. That is, it would be irrational to apply defense system reliability specifications to a commercial electronic product for the obvious reason that customers generally recognize that a product's price is proportional to features, characteristics, and attributes, such as reliability. Manufacturers should realize that any product is transparent to the reliability methods, processes, and tests employed, whether they are space satellites or pharmaceutical devices.
The Product Development Process
The process shown in Figure 1 is generic but indicates major tasks to be accomplished in the development of a product. At every task in the development process there are corresponding reliability engineering tasks that should be integrated into it (Reference 1). Some, but not all, of the more practical reliability engineering methods and techniques are shown. Ronald E. Giuntini, PhD, ASQ CRE, Wyle Laboratories
Figure 1. Reliability Integrated Into the Product Development Process (Click to Zoom)
Product Conceptual Design Phase
Everything should begin with the design requirements, specifications, features, characteristics, and the nature of the usage environment and applications the product is expected to see (Reference 2). Simultaneously, reliability, maintainability, safety, and human factor requirements and specifications should be entered into the decision process. Why? During the product conceptual design phase, several candidate concepts usually are developed. The selection of the best candidate to carry into the preliminary design phase is or should be based on a "systems effectiveness modeling" approach (better than tradeoffs), where every attribute and characteristic is analyzed and calculated (Reference 3). If these reliability-related features are not part of the conceptual design phase, then the selection of the best concept will have been based on an incomplete specifications package.
In the selection of the best candidate concept, a Failure Mode and Effects Analysis (FMEA) is one of the techniques that should be used. Applying the FMEA to the candidate concepts will provide a means of locating their respective weaknesses and can be a key engineering process to help assure that the selected concept is, in fact, the best.
The FMEA can be characterized as a systematic method of cataloging failure modes starting at the lower level of assembly. The FMEA can be performed utilizing either actual failure modes from field data or hypothesized failure modes derived from design analyses, reliability prediction activities and experiences relative to the manner in which components fail. The FMEA provides insight into failure cause and effect relationships. It provides a disciplined method to proceed part-by-part through the system to assess failure consequences. In its most complete form, failure modes are identified at the part level. Each identified part failure mode is analytically induced into the system, and its failure effects are evaluated and noted, including severity and frequency (or probability) of occurrence. Probabilities and severities enable a third factor to be computed called criticality. Criticality provides a means of ranking the failure modes. In this form, the FMEA is upgraded to the Failure Mode, Effects, and Criticality Analysis (FMECA) (References 4,5).
The FMECA is used throughout the product development cycle and can also be applied to the manufacturing process itself to identify failure modes and weak links (Reference 6). The FMECA is an immensely valuable and versatile tool when properly applied. A single reliability technique such as the FMECA is a special purpose tool. It cannot do everything nor answer all the questions. Suppose, the issues arise as to the expected life or the mean-time-between-failures of the product. This requires another technique, the reliability prediction. The prediction can only have relevance during the preliminary design phase and the final design phase.
In the conceptual design phase, an essential precursor step to the prediction should be performed. The reliability goals should be allocated to the various subsystems and lower level functional assemblies and to generic components. The allocation process itself can become involved but is vital since the prediction will show how close to the allocation the design comes so that changes in the design and parts selections can be made with minimal costs, both initial and life-cycle. The allocation process begins with a reliability numerical goal usually expressed as a probability of success such as 0.95 for some specified time. This number is then allocated to the subsystems. If there were "n" subsystems, then the average allocated to each would be the "nth" root of 0.95. Of course, the allocation process is much more complicated than this, but this example shows that the lower one goes into the product or system, the higher the reliability (allocation) at each lower level must become. If "n" were 5, then the average allocated probability would be 0.9898. Component level reliability must be very high. Both the concept FMEA and the reliability allocation are performed in the product conceptual phase.
Product Preliminary Design Phase
The best conceptual design has been selected and in the preliminary design phase, that concept will be "fleshed out". Functions and functional blocks will be replaced with generic hardware elements such as valves, pumps, motors, resistors, micro-processors, etc. The exact part type will not have been selected, but the part will have been designated. Selection of the exact brand with the exact capabilities is usually accomplished in the final design phase.
In addition to performing an FMECA on the preliminary design, a reliability prediction should be made. The prediction will indicate how close the design is to the allocated probabilities. If necessary, tradeoffs can be made to provide higher or lower component reliabilities to achieve the overall subsystem or system level goals.
For electronics, there are two general methods of prediction: (1) parts count reliability prediction method (Preliminary Design Phase) and (2) Parts Stress Reliability Prediction Method (Final Design Phase) (Reference 7).
The parts count reliability prediction method is applicable during the early (preliminary) design phases when the information regarding the product's components is insufficient to use the parts stress analysis models. For this method, only the generic part types including complexity for microcircuits, the quantities, the part quality level and the usage environment are needed.
Since the physics of failure is different for electronic and non-electronic and mechanical parts, the reliability models are different but the basic process is identical.
The reliability prediction provides the quantitative baseline needed to have knowledge of the goodness of a design very early in the development process. Costly weaknesses can be found and improvements made, long before full-scale production begins. It is much cheaper to find the flaws during design and correct them, than to redesign and correct after customers find the problems. The reliability prediction process is a relatively inexpensive method of assessing the quality of the design. It identifies the highest contributors to failure and enables the designer to select other parts or make changes that produce a more reliable and durable product (Reference 8). Predictions may be used to evaluate the need for environmental controls, to employ redundancy, or to trade off other reliability enhancing techniques against cost, space or volume, and other resource limitations.
Reliability prediction enables the designer to examine a number of factors affecting the rate of failure and to test various options for reducing the failure rate by performing sensitivity analysis (i.e. varying the factors and experimenting with various schemes).
Product Final Design Phase
During this phase, all components, parts, assemblies, etc. are to be designated and incorporated into breadboards and engineering models for form, function, and fit testing. Product prototypes are to be fabricated for life stress testing and other types of reliability tests. Everything possible should be done to identify flaws, weak links, failure modes and mechanisms. Which of these that will be corrected is a managerial decision, not an engineering one?
At this phase, another more detailed FMECA and another reliability prediction are needed to parallel the design. The parts stress reliability prediction method is much more detailed and requires unique parts data and the usage of a variety of failure rate prediction equations requiring greater precision. However, accuracy of the prediction is not where its value lies, but in its ability to identify the relative reliability so that prudent parts selections can be made commensurate with the budget and, therefore, the market for the product. It should be emphasized that the idea is not necessarily to produce a product with the highest reliability but the highest affordable reliability that still allows the product to be sold competitively. So the excuse that reliability is too expensive and will force the product out of the competitive range is totally incorrect. It should and can be tailored to the consumer's pocketbook rather than some abstract, unrealistic and arbitrary program. It is much better that the manufacturer finds the bulk of the problems rather than the customer finding them. Realism suggests that a customer's expectations are geared to the product price (e.g. Volkswagen vs. Rolls Royce).
Within this phase are development of prototypes and testing of prototypes. The idea is to test the prototypes to find the weak components or design features. Accelerated life testing and particularly HighlyAccelerated Life Testing (HALT) is a combined environments step stress methodology successfully used by Wyle on customers' products. Obviously, HALT identifies problems that must be corrected. HALT without corrective action would serve no useful purpose (Reference 9).
Often, it is desirable to perform a cycle test or a duration test to determine the failure rate or the mean-time-between failures (MTBF). Such tests have to be designed around the product. Each test will be unique. In this type of test, a quantity of products is run until failure, and the MTBF is statistically calculated for some level of statistical confidence.
The product life can be estimated by specifically engineered accelerated aging tests that can be correlated to the design life of the product.
Other tests of a developmental nature that could be performed during this phase are electrostatic discharge (ESD), radiated emissions, and radiated susceptibility. For some products, other suitable tests such as high pressure or low pressure, or even drops from various heights and orientations could be valuable.
After prototype testing and design modifications have been implemented, the product is ready to be released to manufacturing (Step 7). In Step 8, the product is manufactured. A reliability engineering task that has received some attention the last few years is referred to as the "Process FMEA." The Process FMEA examines the process by which the product itself is built. For some products it is very valuable that the Process FMEA be extended to the analysis of how the product is maintained and used (Reference 6). The Maintenance and Use FMEA should be performed during the design phases so that any features that affect the maintenance or the use could be appropriately dealt with.
Highly Accelerated Stress Screening (HASS)
HASS is a screening process invented by Dr. Gregg K. Hobbs, the inventor of HALT. Both were conceived in 1988. HASS is an environmental stress screen using the highest possible stresses to attain time compression in the screen (Reference 10). Knowledge of the destruct limits (found in HALT) is required to define an effective HASS process to be used as a part of the manufacturing test cycle. HASS is applied to every product coming down the manufacturing line. It is able to prevent products with latent defects from getting into the field. HASS should be strong enough to detect latent failures but not so strong as to take significant life out of the product (Reference 11). HASS is widely used on very high value electronics and may not be appropriate for many medical or pharmaceutical devices. HASS, unlike HALT, is a pass or fail process where only those products that pass get to the marketplace.
Safety and reliability of a product overlap only when the unreliability can result in a risk to the user. A product that fails at an unexpected or crucial usage time can be hazardous (Reference 12). One aspect of reliability is to determine the expected product life so that it will not be used beyond a designated time without preventive maintenance or replacement of the low reliability parts that would render the product prone to potential failures.
Reliability engineering is often a much overlooked discipline in the development of medical and pharmaceutical devices. Reliability should be an integral part of the development process from concept to design to testing. Most device and product developers tend to underestimate the value of reliability engineering as a design and development discipline for producing reliable products in the shortest time and at the least cost. When finally recognized as being essential to getting competitive products to the marketplace, the device often has been designed and built. Trying to add reliability at this point requires costly redesigns and delays. There is no substitute for a thorough product reliability process that parallels the product development process.
- Blanchard, Benjamin S., Logistics Engineering and Management, Prentice-Hall, Inc., Englewood Cliffs, NJ 07632, pages 22-23.
- Chase, Wilton P., Management of System Engineering, John Wiley & Sons, New York, NY, pages 23, 68-69.
- Giuntini, Ronald E., Design Reliability Seminar, John F. Kennedy Space Center, Florida, Four-day seminar for the Kennedy Space Center reliability engineers, 1992, 1993, 1994.
- MIL-STD-1629A, Military Standard, Procedures for Performing a Failure Mode, Effects and Criticality Analysis, November 20, 1980, U.S. Dept. of Defense.
- Giuntini, Ronald E. and Martin, Christopher Scott, Failure Mode, Effects, and CriticalityAnalysis (FMECA) of the SPIROSTM Drug Delivery System, TaskA Part 1, for Dura Pharmaceuticals, December 3, 1996, pages 5-10.
- Onodera, Katsushige, Effective Techniques of FMEA at Each Life-Cycle Stage, in 1997 Proceedings Annual Reliability and Maintainability Symposium, Philadelphia, PA, pages 50-56.
- MIL-HDBK-217F, Military Handbook, Reliability Prediction of Electronic Equipment, December 2, 1991, U.S. Department of Defense, page 3-2.
- Giuntini, Ronald E., Reliability Prediction of the Electronic Systems of the SPIROSTM Drug Delivery System, Dura Pharmaceuticals, TaskA-Part 2, Revision January 30, 1997, pages 12-14.
- Hobbs, Gregg K., HighlyAccelerated Stress Screens HASSSM, publication of Hobbs Engineering Corporation, Westminster, CO, copyright 1992, pages 1-10.
- Hobbs, Gregg K., HighlyAccelerated Stress Screens HASSSM, publication of Hobbs Engineering Corporation, Westminister, CO, copyright 1992, pages 1-7.
- Hobbs, Gregg K., Development of Stress Screens, Proceedings of the Annual Reliability and Maintainability Symposium, Philadelphia, PA, January 27-29, 1987, pages 115-118.
- Firenze, Robert J., The Process of Hazard Control, Kendall/Hunt Publishing Company, Dubuque, IA, page 184.