Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL

Chapter 43. Reliability in Instrumentati... > 43.2Reliability Assessment - Pg. 742

742 paRt | Vi Automation and Control Systems 43.2 Reliability assessment 43.2.1 component Failure Rates An essential part of the design of reliable instrumentation and control systems is a regular assessment of reliability as the design proceeds. The design can then be guided towards the realization of the target MTBF as the finer details are decided and refined where necessary. if reliability assess- ment is delayed until the design is almost completed there is a danger that the calculated MTBF will be so wide of the target that much design work has to be scrapped and work restarted almost from the beginning. in the preliminary design phase only an estimation of component numbers will be available and an approximation to the expected MTBF is all that can be computed. As the design proceeds more detail becomes available, more pre- cise component numbers are available, and a more accurate calculation is warranted. in the final assessment we can examine the stress under which each component operates and allocate to it a suitable failure rate. The basis of all reliability assessment of electronic equip- ment is thus the failure rates of the components from which it is assembled. These rates depend upon the electrical stresses imposed on each component, the environment in which it functions, and the period of operation. The effect of the envi- ronment and the stress will vary widely, depending upon the type of component involved, but all components tend to show similar trends when failure rates are plotted against time. 43.2.2 Variation of Failure Rate with time The general behavior of much electronic equipment is shown in Figure 43.5 in which failure rate is plotted against time. in view of its shape this is often called the "bathtub" curve. The graph can be divided into three areas: an initial period when the failure rate is falling, usually called "infant mortality;" a longer period with an approximately constant failure rate corresponding to the normal working life; and a final period with an increasing failure rate, usually called the "wear-out" phase. This kind of behavior was first observed in equipment which used thermionic valves, but similar characteristics are also found in transistorized equipment, with a somewhat longer period of infant mortality. The fail- ure rate is subsequently approximately constant. The fault statistics from maintained equipment, however, sometimes show a slowly decreasing failure rate which is usually attrib- uted to an improvement in the quality of the replacement components. The argument is that as time passes the com- ponent manufacturer learns from failed components what are the more common causes of failure and can take steps to eliminate them. The manufacturer is thus ascending a "learning curve," and the reliability of the components is expected to increase slowly with time. The evidence from equipment fault recording and life tests thus either supports the assumption of a constant failure rate, or suggests that it is a slightly pessimistic view. in a complex item of equipment such as a digital com- puter there are inevitably a few components with less than average reliability and these are the ones that fail initially. Their replacements are likely to be more reliable, so giving an improvement in system reliability. As these weaker items are eliminated we reach the end of the infant mortality phase and the system failure rate settles down. Where high reliability is important, for example, in air- craft control systems, the customer usually tries to avoid the higher failure rate associated with the infant mortality phase and may specify that the equipment should undergo a "burn-in" period of operation at the manufacturer's factory before dispatch. The burn-in time is then chosen to cover the expected infant mortality phase. The wear-out phase is never normally reached with modern transistorized equipment, as it is almost always scrapped as obsolete long before any sign of wear-out occurs. Even in equipment designed for a long life, typi- cally 25 years, such as submarine repeaters and electronic telephone exchanges, the wear-out phase is not expected to appear. Any component which has moving parts such as a switch or a relay must, however, experience a wear-out phase which FiguRe 43.5 Failure rate and failure density func- tion variation with time.