FlightGlobal.com
Home
Premium
Archive
Video
Images
Forum
Atlas
Blogs
Jobs
Shop
RSS
Email Newsletters
You are in:
Home
Aviation History
1961
1961 - 1431.PDF
CORRECT FUNCTIONING WITHIN THIS CURVE FLIGHT, 5 October 1961 Aviation Electronics . . . 535 control practices. The third and fourth types of failure requiredetailed review of both design and operating conditions, in which "post-mortem" analysis of fault reports can be helpful. In the process of assessing the effect of defective parts, theprobability of a faulty unit is given by: P - I - e-n«where: P is the probability of a faulty unitn is the number of parts in that unit q is the average quality level of those parts. The quality of parts received from a manufacturer can be estab-lished by a sampling inspection plan or by complete inspection of the parts. Similarly, the incidence of breakdowns in storage, testand assembly can be assessed. A plot of this function is given on p. 534. As an example it can be seen that a unit containing 250components requires an average quality level of better than one faulty part per thousand to have an 80 per cent chance of workingwhen switched on. Time-dependent Failures For the initial prediction of the number of failures which willoccur during the operating life of a unit a probability model is appropriate. A prediction of ultimate obtainable reliability can bebased on known reliabilities of individual components from which the non-time-dependent failures have been eliminated. Thisassumes correct circuit design, safety margins and environment, and accurately known failure rates for the component parts. Itmust also ignore the "infant mortality'" known to exist with some components. Such a probability model assumes a catastrophic failure, i.e.,that the component is either good or bad, and gives the probability of failure as: P = I ^ e1" where: and: where: P is the probability of failuret is the operating time in hours T is the mean time between failure in hours 7f f is the sum of the failure rates of all theindividual components. Typical failure rates range between 0.01 per cent per 1,000 hoursfor metal oxide resistors to an expected life of a few hundred hours for some small servo motors. Although a transistor mayhave an expected life of the order of one million hours—a failure rate of 0.1 per cent per 1,000 hours—a unit containing, say. 800such transistors stands a 2 per cent chance of failing in just one day's operation as a result of their unreliability. It is very necessary to appreciate the difference between time-dependent and non-time-dependent failures and of the relative effects of the reliability of individual components. It has been saidthat the ideal racing car would cross the finishing line of a race and then fall to pieces. Similarly, the reliabilities of electroniccomponents and equipments should be balanced to give the desired system performance. In an equipment containing one servo motorand 1,000 transistors, the transistor reliability is important, but if the ratio were only one to ten the transistors would play a farsmaller part in the reliability of the whole. Again, it is useless to strive to improve components already adequately reliable whenthey are regularly damaged by the operators during assembly or test, or when the system contains interface problems between units.resulting in faulty operating conditions. The American ApproachThe Americans have realised in the last few years that reliability costs money. To evaluate components involves life tests on manythousands of them, often for long periods. Collection of statistical data on equipment involves large numbers of people, and largesums of money are often needed to buy environmental test facilities: and that is only the beginning. Once the reliability of a product ismeasured, an improvement in its reliability involves tighter and more detailed specifications for components and consequentlyhigher cost, closer inspection and more rigorous testing. On one project alone, an American company has spent S35 millionfor reliability. This is not private money, but government money and it is bringing a high return. For the first time, too, the militaryauthorities are demanding that the manufacturers guarantee a minimum failure rate for their equipment. Responsibility forreliability is thus being placed once again where it belongs, with the equipment designers: and large quality-control organizations arebeing established to meet the problems. The field of the quality-control engineer then covers not only thecompilation of statistics, as it does in this country, but also the evaluation of design writing component specifications, test specifica-tions, production engineering and all stages of inspection. The BEST XWORKIN POINT NOMINALVALUE OF A NOMINAL VALUE OF B PARAMETER B The irregular "Shmoo" plotted to show the spread of conditions unde- which a circuit will operate effectively. The best working point ij shown, but the circuits still work with the parameters plotted American component industry co-operates in the quest for relia-bility. Competition is far more intense there than it is in this eountr> and military projects account for a higher proportion of the totaloutput. Designers of aircraft equipment are therefore able to impose specifications covering not only the parameters of the finishedarticle, but also the standard of workmanship, the manufacturing materials and the processes used. In many factories, the buyer place*-his own inspectors to inspect his goods during the manufacturing stages. It seems that the customer is telling the manufacturer howto do his job; but remarkably enough, the method works. Built-in Reliability Both in the USA and in this country the military authoritiesattempt to exert some control over materials and manufacture of equipment for their use. They publish sets of standards such as theAmerican MIL Standard series, and have compiled lists of approved parts. Their own inspection departments are scattered throughoutthe industry. But the methods differ in the two countries. In Britain, the finished articles are inspected whereas the Americaninspectorate approves the manufacturing processes and the com- petence of the operators. The finished article is not normallysubjected to detailed inspection. Although these methods can ensure the overall standard of the product, they are not capable ofdealing with the special requirements of certain equipment. Where specially high reliability or critical component parameters arerequired, or when some new process is being used, it is necessary for the American system of individual specifications drawn up in con-junction with the design authority to be applied, together with corresponding inspection methods. Here lies the origin of the complicated procedure designed tobuild into the finished product the desired degree of reliability A factory engaged in such a programme mav employ one qualitv-control specialist for as few as every five members of the design and production team. Components are usually subjected to 100 percent inspection and testing on receipt and may require special storage conditions. In the interests of accuracy, assembly is made assimple as possible and wiring aids are used. A wireman who has 20 sheets of diagrams, each showing three wires, will make fewermistakes than if he were to use a conventional one-sheet wire list covering all 60 wires. Evaluation of a design requires a different approach. Everycomponent must operate well within its rating, not only within applied voltage and power dissipation limits, but also within theexpected tolerances for parameters of the component itself. A circuit must function correctly when the tolerances of associatedcircuits vary—a factor which is more difficult to assess. A number of techniques have been evolved, including "worst case" analysis inwhich all components are assumed to be on their worst margin of tolerance and very few circuits will work. It is then necessary toapply a process similar to that known to mathematicians as relaxation. The possible values of each parameter in turn areexplored and circuit values adjusted to obtain correct operation. If the circuit contains more than two or three components the problemoften becomes very complicated and digital computers may be used in its solution. A much simpler process is that of drawing Shmoo diagrams wherejust two parameters are considered. The correct function of the circuit is plotted against both parameters giving a Shmoo-likearea within which the circuit works (see above). The best operating point is the point furthest from the boundary of the Schmoo in anydirection. Some other aspects of design affect the expected life of components. The life of tantalum capacitors, for instance, dependsvery critically on their working voltage. It is at this stage that a nev.
Sign up to
Flight Digital Magazine
Flight Print Magazine
Airline Business Magazine
E-newsletters
RSS
Events
