Reliability Growth

Observing and Predicting Trends in Reliability

Today's complex products require that companies focus on establishing detailed plans and procedures for developing and manufacturing products that meet specified reliability, maintainability, and other performance requirements. Reliability growth refers to a well-defined process for identifying and correcting reliability problems early in the design process so that the reliability of a product increases or "grows" as the product goes through successive development stages. Reliability growth programs should be established for all new products and for all existing products undergoing major redesigns so that the improvement due to changes in design and manufacturing processes can be easily tracked.

During the early design stage, a reliability goal is set for a product. Because failure data from prototype testing is not yet available, the initial reliability goal is often based on either the failure data for similar products or the failure data for the subcomponents of the product. During the development stage, product prototypes generally undergo extensive testing so that deficiencies in respect to design, engineering, and manufacturing can be identified and corrected.

A typical reliability improvement test consists of operating product prototypes for several weeks in the same types of environments in which customers will eventually operate the product. A team of project engineers and technicians analyze every failure that occurs, determining root causes for failures and developing design and manufacturing improvements that are to either eliminate or reduce the recurrences of these failures. As the testing continues, the improvements developed by the team are incorporated into the prototype so that product reliability continues to improve throughout the testing period.

Achieving Reliability Growth

The three most important steps in the iterative process for achieving reliability growth are depicted in the feedback loop that follows:

Most of the problems encountered during testing are likely to be component failures and manufacturing deficiencies that could not be foreseen in the early design phase. Because various performance requirements can conflict, optimizing a design to meet one requirement can cause the design to fail to meet another requirement. Thus, iterations of designs are often needed to correct all of the component selection and manufacturing deficiencies that are found during prototype testing.

In addition to improving reliability, early implementation of a reliability growth program minimizes the impact on production scheduling and total product cost, especially since the costs associated with either redesigning a product late in the development cycle or retrofitting products already in the field are extremely high. Developing a product that meets reliability requirements also ensures that the product ultimately meets user needs and has an acceptable total life cycle cost. Setting interim reliability goals that are to be achieved during testing ensures that resources are allocated efficiently.

Reliability Growth Data

The basic principle of any reliability growth model is to apply the testing results and data points to determine if the reliability of the product is growing sufficiently to meet the reliability requirements for the product. The types of data used for predicting reliability growth are:

• Reliability Data. The reliability of the product is recorded at different points in time. The reliability is a ratio of the number of units still functioning and the number of units that entered the stage.
• Success/Failure Data. The item is tested and can either succeed or fail. Success data can consist of a code indicating the outcome, such as S for Success or F for Failure; or, it can consist of a code indicating the failure mode for each failure. Success/failure data might also indicate the number of failures that occurred when several units are tested together.
• Failure Time Data. The time to failure for an item is tracked using either cumulative or non-cumulative operating times. Failure time data is the most commonly used type of data in reliability growth. This data can be for one failure or system, or it can be for multiple failures or multiple systems, which is commonly called interval or grouped data.

The failure data collected during prototype testing is used to determine whether the reliability goal for the product is likely to be met or exceeded by the time the product is scheduled to be put into full-scale production. Although many methods exist for modeling the reliability growth process, the Duane/Crow-AMSAA model is considered the best practice. In 1964, J. T. Duane, an engineer at the Aerospace Electronics Department of General Electronic Company, published a paper demonstrating how a learning curve approach could be used to monitor the continuing reliability improvements in the early stages of developing complex electromechanical and mechanical systems. Duane explained how this was because the lessons learned from failures were used to refine designs.

According to Duane, a graph of the cumulative MTBF versus the cumulative operating time plotted on log-log paper fell close to a straight line. Graphing this learning curve provides a means of measuring and predicting reliability during a period of product change. Dr. Larry Crow later added powerful statistical capabilities to Duane's postulate for learning curve modeling to create the model so widely respected and implemented today. An example of a reliability growth chart generated by the Duane/Crow-AMSAA model follows:

Reliability growth is generally quantified by graphing any of the following three measures over time:

• The increase in the mission success probability (reliability).
• The increase in MTBF.
• The decrease in failure rate as a function of time.

Reliability growth charts depict trends that are used to forecast failures as a function of additional test time or calendar time, thereby making planning for redesign and test resources easier. In addition to allowing you to determine whether reliability requirements will be achieved, reliability growth charts can help you to determine the time needed to meet these requirements and the associated costs. Extrapolating a growth curve beyond the currently available data shows what reliability a program can be expected to achieve providing that the conditions of the test and the engineering effort to improve reliability are maintained at their present levels.

If the reliability growth graph indicates that the reliability goal is not going to be met or exceeded, then the product design must be improved. This might require the use of more reliable components or redundancy. Additional resources might also have to be devoted to designing and manufacturing a more reliable product to meet the required delivery date.

Other Uses for Reliability Growth Charts

In addition to analyzing the results from testing newly developed or redesigned products, the Duane/Crow-AMSAA model can be used to:

• Predict future failures.
• Track fleets of repairable systems.
• Provide accurate trending of significant events for management.
• Forecast safety incidents that can be controlled.

If you would like additional information about how reliability growth is implemented in the Relex Reliability Software Suite, please email info@relexsoftware.com.

Copyright © 2004, Relex Software Corporation

Close Window