Design For Reliability and Quality: IIT, Bombay
Design For Reliability and Quality: IIT, Bombay
IIT, Bombay
Lecture 3
Design for Reliability
IIT, Bombay
Instructional Objectives
The objective of this lecture is to outline the basic concepts of risk, reliability and safety and the methodologies to define reliability in a quantitative manner to aid to design for reliability.
Risk Assessment
A product is considered safe if the risks associated with the product are assessed to be acceptable. The risk assessment often involves the determination of the frequency of the occurrence of a specified hazard and a conclusion on the acceptability of the hazard. Alternately, the levels of risk can be classified as: Tolerable Risk This level of risk is usually accepted while a constant review of its cause and ways to reduce the same is the must. Acceptable Risk This level of risk is also acceptable and does not need immediate attention. Unacceptable Risk: This level of risk is deemed to be unacceptable.
Quantify Reliability
Engineering designs often neglect the stochastic nature of the material properties, the dimensions of the components and the externally applied load and usually consider the same by applying a factor of safety. However, a quantification of the uncertainty in design of parts is necessary in critical applications like in space, aircraft and nuclear applications.
IIT, Bombay
For example, consider a structural member subjected to a static load experiences a stress and exhibits the yield strength (of the material) as y . Figure 5.3.1 schematically depicts the nature of variation of stress experienced and the material yield strength. As shown schematically in Figure 5.3.1, the distribution of stress on the structural member can be envisaged as typical Gaussian in nature with a mean ( ) and standard deviation ( sd ). Similarly, the variation in the yield strength of the material can also be envisaged as Gaussian with a presumed mean ( y ) and
y standard deviation ( sd ). If we subtract the stress distribution from the strength distribution, we
get a distribution as ( y ), which is always positive. Interesting to note, however, is the fact that the frequency distribution of can overlap with that of y and can lead to a situation as
> y , which will lead to failure. So, the probability of failure will be given as Pf = P( > y )
and, the reliability will then be defined as R = 1 Pf (2) (1)
The mean and the standard deviation of the distribution ( y ) can be easily determined considering and y as two independent variables.
y y
Figure 5.3.1 Schematic presentation of the distribution of stress on a typical structural member and variation in the material yield strength
To avoid unforeseen failure as indicated above, the mechanical designs often undertake the concept of a safety factor to reduce the probability of failure and hence, increase the reliability. The safety factor may be estimated as the ratio of the mean capacity to the mean load or demand.
IIT, Bombay
Thus, reliability can be defined as the probability that a system, component, or device will perform without failure for a specified period of time under a given operating conditions. If R(t) and F(t) are respectively is the reliability and the probability of failure with respect to time t, and are mutually exclusive, we can write R ( t ) + F( t ) = 1 (3)
Furthermore, considering a typical test for reliability with N 0 number of parts, and N S ( t ) and N F ( t ) are respectively the number of parts surviving and failing till time t, we can write that N S (t) + N F (t) = N 0 We can write further that,
R (t) = N S (t) N (t) = 1 F( t ) = 1 F N0 N0
(4)
(5)
Hence, the instantaneous failure rate can be envisaged as the number of failures per unit time for the total number of parts exposed during time t and given as, h(t) = dN F ( t ) 1 dt N 0 ( t ) (6)
In a statistical form, the instantaneous failure rate, h(t), can be envisaged as the probability that a given test item has survived till time t 1 and would fail between time interval t 1 and t 1 + dt 1 , and can be expressed as h(t) = f (t) f (t) = = P( t 1 t t 1 + dt 1 | t t 1 ) 1 F( t ) R ( t ) (7)
where f(t) refers to the typical nature of the statistical frequency distribution of time to failure. Hence, f(t) can be written as
f (t) =
dF( t ) d[1 R ( t )] dR ( t ) = = dt dt dt
(8)
Substituting equation (8) in (7), we can write, h(t) = t dR ( t ) 1 dR ( t ) = h ( t )dt R ( t ) = exp h ( t )dt dt R ( t ) R (t) 0 (9)
For a given constant value of failure rate, equation (9) can thus be used to estimate the reliability of a test component. Figure 5.3.2 schematically shows the typical nature of the failure curves realized for general components.
IIT, Bombay
The failure curve usually consists of three parts. The first part is a decreasing failure rate, known as early failures that may occur due to design errors, manufacturing defects, or installations errors. These early failures can be minimized by improving the production quality control, and subjecting the parts to a proof test before their actual service with the customers. As time passes the system stabilizes and these early failures leave the system leading to a constant failure rate in the second part, which is also referred to as random failure zone. Here the failures occur because of random overloads or random faults having no pattern at all. Mechanical components usually do not exhibit a region of constant failure rate. After a stipulated long time, the mechanical components and materials begin to age and wear out and this accelerates the failure rate leading to an accelerated failure rate in the third (last) part. The failure curve is also referred to as bathtub curve due to its shape. However, the usual nature of the distribution of mechanical failures is slightly different from the electronic failures as shown in Figure 5.3.3.
IIT, Bombay
Figure 5.3.3 (a) Failure curve of a typical electronic equipment (b) Failure curve more typical of mechanical equipment
With respect to the reliability of components, mean life of a part or component refers to the average life a specified number of components which is measured over the entire span of their life. The mean time to failure (MTTF) refers to the sum of survival time (up time) for all the components put under test divided by the number of components, which are failed. For example, assume the testing of 3 identical systems starting from time 0 until all of them failed. The first system failed at 23 hours, the second failed at 26 hours and the third failed at 20 hours. The MTTF is the average of the three failure times, which is 23 hours. The mean time between failure (MTBF) refers to the mean or average time between two successive component failures. MTBF is usually applied to a group of similar equipment, for example all the pumps in a refinery. If there are 40 operating pumps with a total of 23 failures over one month, the MTBF during that month is calculated as (40 / 23) or 1.74 months. For a single item, MTBF is just the time period / number of failures. For example, if a pump fails thrice in one month, the MTBF would be 1/3 = 0.33 months
System Reliability
The overall reliability of a typical mechanical or any system will be governed by the reliability of the individual components. If the components are so arranged that the failure of any component will lead to the total failure in the system, the same is envisaged to be in series. In contrast, a
IIT, Bombay
better arrangement of components would be one that would require failure of all components for the total failure of the system. Such an arrangement exhibits parallel reliability. The system reliability in series and parallel arrangement of components can be given as equations (10) and (11), respectively. R system = R A R B R C ........ R n (10) (11)
R system = 1 (1 R A ) (1 R B ) (1 R C ) ........ (1 R n )
IIT, Bombay
Customer requirements Demand from customers for an effective reliability program is ever increasing based on the benefits of such programs. Competitive advantage Companies often publish the product reliability numbers to enhance competitiveness.
The demand to achieve desired performance level in an efficient and optimized manner has led to a growing movement towards increasing applications of design for reliability and its spread to industries where it had not been used in the past. Previously design practices tend to focus on mainly on functionality and robustness or product integrity. In developing the design for the products, the following characteristics are usually missed: - (a) key failure modes and failure rate of the product, (b) key failure mechanisms that may be present in the service environment, (c) usable life of the product, (d) cost of maintenance required to maintain the inherent reliability, (e) availability, and (f) rigorous testing.
IIT, Bombay
As to address all of the above issues, the process of design for reliability has been developed to ensure a predictable and reliable life of usual components. Some of the key elements for design for reliability is as follows. Concurrent Engineering Concurrent engineering is a feature that ensures the design is not completed before reliability requirements are identified and dealt with. Configuration Design The physical configuration is the key important characteristic that determines the reliability of an asset. Depending on the severity of the product service and the maximum economic reliability of available components present in the product, it may be necessary to build redundancy into some locations. Component Selection The second important characteristic that determines reliability is the choice of components that make the product. Components with better load bearing ability rather than cheap components should be considered as better option. Design and Build It is possible to create a strong configuration and select robust components, and still produce a product that is unreliable. There are design and assembly practices like use of protective grommets at points of wear, use of strain relief at bends, or changes in direction that ensure the configuration and components deliver the desired reliability. Verification and Performance Testing The final assembled product may not always perform as expected. Interactions between dynamic components can produce unexpected effects. As a result, it is necessary to verify that the assembled product functions as expected. It is also essential to simulate the wear and tear that represents an entire life using accelerated testing. Customer Needs The product must be designed not only based on functionality but also considering the customer needs.
IIT, Bombay
IIT, Bombay
increase service life and reduce the maintenance cost. Such decisions can be justified by using techniques like life cycle costing and so on. Ease of inspection It is very essential that cracks or any other flaws are identified when they occur in the product. Ideally it should be possible to perform various visual methods for detections of such cracks or other flaws, but special design features may have to be provided in order to ensure this. The product should be designed in such a way that it is capable for ready inspection. Simplicity Simplification of the component and the product reduces the chances for failure and errors and increases the reliability. Specificity The greater the level of specificity, the higher will be the inherent reliability of the design. Specifying standard components for the product increases the reliability. It means the component being used has a history and their reliability is known. Replacement Whenever it is required to use components with high failure rates the design should specifically take care for the ease of replacement of such component.
IIT, Bombay
Exercise
1. If a device has a failure rate of 5x10-6 failures/h, what is the reliability for an operating period of 1000 h?
References
[1] 1. David M. Anderson and David M. Anderson, Design for Manufacturability and Concurrent Engineering, CIM Press, 2004. [2] G Dieter, Engineering Design - a materials and processing approach, McGraw Hill, NY, 2000.
IIT, Bombay