Preface

Lev M. Klyatis and Edward L. Anderson

When Lev Klyatis began his engineering career in 1958 as a test engineer at the Ukrainian State Test Center for farm machinery, he was surprised to learn that, even after extensive testing by this center, the testing was not accurately predicting the reliability of the products as used by farmers. This test center would conduct farm machinery field testing during one season of operation, and make the recommendation to manufacture the new product based on results of this single-season testing.

Neither the designers, nor test engineers, nor the researchers, nor other decision-makers involved knew what would happen after the first season. The test center was not accurately predicting true product reliability during the life cycle of the machines. Later, Lev Klyatis realized that this situation was not unique to farm machinery, but was related to other areas of industry and other countries over the world, even when they claimed to be doing accelerated reliability testing.

Why are we writing this book? As will be seen, it is the author's observation that the developments of technology, methodologies, hardware, and software are advancing at an unprecedented rate. But, in the same time, we find that reliability testing and prediction are advancing much more slowly; and in many cases it is common to find reliability testing and prediction methodologies that have changed little in the past 60-70 years. As product complexity increases, the need for near-perfect product reliability, which is founded on the ability to accurately predict reliability prior to widespread production and marketing, becomes a company's critical objective. Failure to predict and remedy failures can result in human tragedy, as well as serious financial losses to the company. Consider the two following recent examples.

On May 31, 2009, Air France's flight AF447 departed Rio de Janeiro en route to Paris carrying 228 passengers and crew; several hours into the flight it crashed into the Atlantic Ocean, killing all on board [1,2]. A contributing factor in the accident was pitot tubes, which were believed to have iced, resulting in the loss of accurate airspeed and altitude information. The pitot tubes were known to have a problem with icing and had been replaced by several other airlines. Following the accident, The European Aviation Safety Agency (EASA) made compulsory the replacement of two out of three airspeed pitot's on Airbus A330s and A340s AD (204-03-33 Airbus Amendment 3913-447. Docked 2001- NM-302-AD), and the FAA followed with a near-identical requirement in promulgating Docket No. FAA-2009-0781 AD 2009-18-08 Final Rule Airworthiness Directive AD concerning Airbus A330 and A340 airplanes. It is profoundly troubling that in age of state-of-the-art fly-by-wire jet aircraft, we would be encountering problems with pitot tube icing [3].

In February 2014, General Motors issued a recall for over 2.6 million vehicles to correct an ignition switch defect responsible for at least 13 deaths, and possible more than 100, and this does not include those seriously injured. The ignition switch could move from the "On" position to the "Acc" position; and, when this happened, safety systems, such as air bags, anti-lock brakes, and power steering, could be disabled with the vehicle moving. The problem was initially uncovered by GM as early as 2001, with continued recommendations to change the design through 2005, but this recommendation was rejected by management.

By the end of March of 2015 the cost to GM for the ignition switch recalls was $200 million and was expected to reach as much as $600 million [4-6]. Add to the financial loss the personal tragedy of those killed or injured and to their families, and the true cost of failed reliability prediction becomes evident. By the end of the next decade it is almost a certainty that you will be sharing the road with some type of autonomous vehicle [7-9]. Consider the degree of reliability prediction that will be needed to provide the level of confidence needed. Whether you are driving an autonomous vehicle or merely sharing the road with them, you are literally betting your life on the adequacy and accuracy of the reliability testing for each critical component and decision-making process. Considering that, today, we are having difficulties with ignition switches and pitot tubes, this will be a major undertaking.

This is particularly so when the testing will need to account for such varied environmental conditions as heat, cold, rain, snow, roadway salt, and various other expected and unexpected contaminants. Couple this with the 10 years plus life of the average automobile [10], and reliability assurance against a wide variety of degradations is necessary, and all life failure modes must default to a fail-safe mode. These are only examples from many real-life problems that are connected with inadequate reliability prediction and testing methods.

Unfortunately, too often these costs for failed reliability prediction and testing are never factored into an organization's decision-making processes. While the human and financial impacts of responsible new product development should be foremost in an organization's (including research and pre-design, and testing) activities and concerns, too often they are overlooked or assumed to be someone else's responsibility, especially in a large organization. But if we are to remain a civilized society, such responsibility cannot and should not be delegated up the chain of command.

One of the major concerns is the development of higher speed and processing power of electronic developments occurring at a much greater rate than other areas of people's activity. In fact, Moore's law-the expectation that microprocessor power doubles every 18 months-is widely accepted in the industry. How often do we consider the effects and the implications of electronic developments, especially in new software development, that transfers thinking resulting in real brain development to what may be termed virtual thinking? Virtual thinking is surrendering human thinking and mental development to a system of control that provides answers automatically, with a minimal role for thought. Consider how calculators and electronic cash registers have reduced people's ability to do basic mathematics; or how GPS navigation has diminished the average person's ability to read a map and plot the course to their destination-it is so much easier to just type in the destination address and let the machine direct you turn by turn to the destination. But the development of thinking skills and using them for the advancement of society and civilization are the basic differences between humans and animals.

Unfortunately, people often do not understand that electronic systems are only part of a system of controls containing real physical limits, processes and technologies, and that there are limits to what can be accomplished with software and corresponding hardware. The most advanced automotive stability system cannot allow the vehicle to corner at an unsafe speed. While the system may enhances a person's abilities as a driver, it cannot violate the rules of physics. And frequently, the enhancement provided by technology is accompanied by a reduction in the skill of the operator as they become increasingly dependent on the technology.

A common example of this occurs when predictions are based on abstract (virtual) processes which are different than the real (actual) processes. Too often, prediction reliability is based on only virtual (theoretical) understanding and does not account for the real situations in people's real life. Because of this, many reliability prediction approaches are based only on theoretical knowledge, and the testing used is not a real interconnected process, but relies on secondary conditions expected in their virtual world. Therefore, testing development, together with prediction development, is not developing as quickly as needed and is moving forward very slowly, much more slowly than design and manufacturing processes (Figure 3.6). Reliability (mostly accelerated reliability) testing needs technology, equipment, and corresponding costs as complicated as the new products they are testing. But too often should be key concerns the management of many companies prefer to pay as little as they can for this technology and the development of necessary testing equipment. They want to save expenses for this stage of product development. Phillip Coyle, the former director of the Operational Test and Evaluation Office (Pentagon) said in the US Senate that if, during the design and manufacturing of complicated apparatus such as a satellite, one tries to save a few pennies in testing, the end results may be a huge loss of thousands of dollars due to faulty products which have to be replaced because of this mistake. This relates to other products, too.

As a result, product reliability prediction is unsuccessful, which is reflected in increasing and unpredictable recalls, decreasing actual reliability of industrial products and technologies, decreasing profit and increasing life-cycle cost, in comparison with that planned during design. Lev Klyatis came to realize that the reliability prediction approaches utilized at the time (and even frequently now) did not obtain accurate or adequate initial information to successfully predict the...