Students will learn how reliability allocation (or apportionment) can help manage system and machine development where high-level reliability targets must be interpreted into specific goals for subsystems and components.

One of the easiest ways to save time and money when developing complex systems is to get the reliability targets for subsystems and components right. Giving a ‘complex’ or ‘technologically challenging’ subsystem and an easier target than others makes it easier for the design and development to meet its targets. Students will learn how helpful and valuable reliability (and quality) allocation can be.

Students will learn about reliability basics, its key definitions, and how it relates to other metrics such as the ‘mean time to failure (MTTF)’ and ‘mean time between failure (MTBF).’

Some systems involve redundant components. Understanding how much redundancy plays a role in system reliability allocation is important. Students will learn how to estimate system reliability basic component reliability characteristics. Students will also learn how simply allocating system reliability targets to subsystems and components without modification ensures that system reliability requirements will not be met.

Reliability performance requirements are not arbitrary. They must be based on business, technology, market, and several other factors. Further, the targets provided to the design or production as part of a larger reliability allocation framework must include reliability design margin. Students will learn about preparing system-level requirements (with margin) in preparation for

The ‘easiest’ way to allocate goals is to give each component the same goal as the others. This makes it easy to find mathematically. However, this creates the problem that reliability allocation is seeking to avoid. A more helpful approach involves ‘allocation factors’ allowing the more ‘complex’ or ‘challenging’ components to achieve the ‘easier’ goals.

The ‘theoretical’ part of reliability allocation involves taking system reliability goals and targets (once margin has been set aside) and using allocation factors to create tailored goals that allow the more ‘challenging’ components to get the ‘easier’ goals. The more ‘basic’ components get, the more ‘challenging’ their goals must be.

Not all failures are created equal. For example, some product features might only be used by a smaller fraction of users. Perhaps only 75 % of smart lock users and customers will use the keypad. So if the keypad fails for the other 25 %, it won’t matter. This reduces the ‘importance’ of the keypad. Taking into consideration ‘importance’ allows reliability targets to be easier, so you don’t make something more reliable than it needs to be. ‘Importance factors’ can combine with ‘allocation’ factors to consider the smart lock example above, failures with different levels of ‘severity,’ and component failures that don’t always cause system failures.

The activities described in the previous lessons must be combined coherently as part of an overall strategy. This comprises ‘8 steps’ that start with gathering information to help set system reliability goals and targets, all the way to managing (and re-negotiating) reliability goals throughout the production process.

Most products and machines involve software, which can be challenging to characterize in terms of ‘reliability challenges.’ Software cannot be assumed to be perfectly reliable to make

the reliability allocation equations work.’ Software’s defect density and code complexity can be used to create a relative comparison to the hardware components.

Redundancy involves additional components that provide identical (or very similar) functions as other components to improve reliability. However, redundancy comes at a considerable cost beyond money. Redundancy also involves additional size, mass, heat, and power consumption. Following the ‘8 steps of reliability allocation’ means quickly determining if you need redundancy and how that influences other component goals.

Not all design and production allow ultimate flexibility with component design and selection. Some components are ‘mandated’ in that they have already been selected or are so unique that there are no options. This lesson teaches students how to allocate goals and targets only to the components whose reliability performance can still be influenced.

Products with relatively mature technologies might allow ‘indicative’ reliability estimates to be made. This lesson teaches students how to take these estimates and apportion goals so that the failure probabilities of these components remain proportional. This can offer more ‘accurate’ goals based on actual data, but does involve more challenging numerical calculations.

Some systems must have redundancy. This can be due to standards, regulations, or legislation. This is particularly relevant for safety-critical systems. It can also apply to repairable systems where there is a desire to have redundant components to allow concurrent maintenance. This lesson teaches students how to allocate goals for systems that aren’t simple series systems without redundancy.

One of the most overused assumptions in reliability engineering is that components and systems have ‘constant hazard rates.’ This implies that a system never ages. A one-hundred-year-old component would be as likely to fail in the next hour as a brand-new one. Most components experience some wear-in (or infant mortality), some wear-out, or both. This makes the ‘constant hazard rate’ assumption fundamentally incorrect. Students will learn about the ‘constant hazard rate’ and how, while inappropriate, it is often used to simplify reliability challenges, including reliability allocation.

One of the most well-known ‘constant hazard rate’ allocation methods was developed in the 1960s by the US Advisory Group on Reliability of Electronic Equipment (AGREE) for the Office of the Assistant Secretary of Defense. It has the inherent weakness of assuming all the parts of the system have constant hazard rates (they don’t) and uses a basic approach where these hazard rates are proportionally allocated based on the number of ‘subassemblies’ each subsystem or component has. Students will learn about the limitations of the ‘AGREE method’ and how to apply it.

The ‘mean time to failure’ or ‘MTTF’ is a commonly used statistic in reliability engineering, but it isn’t a reliability metric or statistic. Depending on how your products wear in or wear out, anywhere between 50 and 80 percent of them will fail by the ‘MTTF.’ We are usually interested in reliabilities where 10 percent or fewer fail. Further testing for the ‘MTTF’ often means you need to test for a lot longer than the timeframe of interest (like warranty period or service life). Students will learn how the ‘MTTF’ relates to the ‘constant hazard rate’ and reliability, and how dangerous it can be when confused with reliability.

Another well-known ‘constant hazard rate’ allocation method was developed by Aeronautical Radio Incorporated (ARINC) and is now known as the ARINC method.’ It has the same weaknesses associated with assuming a ‘constant hazard rate’, but is based on failure rate estimates for specific types of components. Typically, these estimates are drawn from standard ‘reliability prediction’ handbooks that are rudimentary at best. Students will learn how to apply the ‘ARINC method of reliability allocation’ and its shortfalls.

The third well-known ‘constant hazard rate’ allocation method is the ‘feasibility of objectives’ method, which uses a combination of four factors to allocate reliability goals: complexity, ‘state of the art,’ ‘operating time,’ and ‘environment.’ Students will learn how to use the ‘feasibility of objectives’ method, which is mathematically similar to the other two ‘constant hazard rate’ methods.

A very numerical and analytical approach to reliability allocation where goals are optimized against other characteristics such as weight, cost, and size. However, these approaches are based on creating assumed functions or ‘curves’ that relate costs to reliability. These curves are

notoriously hypothetical and removed from actual performance. Students will learn how to undertake optimization-based reliability allocation using techniques found in MIL-HDBK-338.

The ‘mean time to failure’ or ‘MTTF’ is associated with non-repairable products or systems (such as consumer products). The ‘mean time between failure’ or ‘MTBF’ is associated with repairable products or systems (like a manufacturing plant). The ‘MTBF’ is mathematically easier to allocate and can reflect meaningful performance metrics. Students will learn about the ‘MTBF’ and how it can relate to things like maintenance regimes (in a positive way).

Another approach to dealing with repairable systems is to allocate ‘availability goals.’ This takes into consideration how often machines fail and how long they remain offline. The mathematics behind allocating reliability is identical to that for reliability. Students will learn how to allocate availability goals to machines and systems that are routinely repaired.

Manufacturing quality refers to the likelihood of parts or components being defective. Quality can be measured using what’s called ‘capability indices’ that allow us (amongst other things) to anticipate the ‘parts per million’ or ‘PPM’ defect rates. The underlying mathematics is similar to that for reliability. Students will learn how to allocate manufacturing quality goals for specific products ‘critical to quality’ or ‘CTQ’ characteristics.

Students will review the key characteristics and features of reliability, MTTF, MTBF, availability, and manufacturing quality allocation. Students will also review the pros and cons of allocation and when it should (or should not) be used.