Part III: Which Model To Choose

Note: This is the third part of a three-part article. It provides guidelines for making a sound judgment when deciding which reliability prediction standard to apply to a particular analysis. Part I identifies the major factors in reliability prediction models that contribute to predicting component failure. Part II explores the various reliability prediction standards that are available, including MIL-HDBK-217, Telcordia, NSWC-98/LE1, PRISM, RDF 2000, CNET 93, HRD5, and GJB/z 299B.

Choosing a Reliability Prediction Standard

There is no correct answer when choosing the reliability prediction standard to apply to your analysis. However, there is a best answer based on sound reasoning, customer requirements, project goals, and model metrics.

Customer Requirements

A particular standard may be specified by the customer in the work contract. If this is the case, your decision is an easy one. If a standard is not specified, then you may want to consider the nationality of the customer. Certain standards may be more generally accepted or preferred in the customer's country of origin. When conducting a reliability prediction for a U.S. military program, for example, one would most likely use MIL-HDBK-217.

Model Coverage and Environment Support

Other very important factors in deciding which standard to use include the coverage of component types and environments supported. Certain standards may not have the part types or operational environments that are specified in the project. To assist you with this decision, we have provided a comprehensive list of covered part types and environments for the four major standards at www.relex.com/resources/art/art_predmodels.asp.

Reliability Goals

The reliability goals of the project may be a consideration when making a decision on a standard. Is the project implementing a reliability growth program? If so, you may want to plan ahead by using a standard that allows for the incorporation of lab and field data to refine the prediction results. Does the project require that the analysis consider manufacturing or management processes? Then perhaps PRISM with process grades is the way to go. As you will see in the next section, associating these features with a particular standard is not as critical when the Relex Reliability Prediction package is used.

Project Design Stage

If the project is in the early design stage, not all model parameters may be known. This situation may dictate taking a parts count approach. On the other hand, the project may be a second or third iteration of a previous product, and you may have predecessor data available.

Industry of Application

A certain standard may be more accepted in a particular industry. For example if the project is in the telecommunications industry, you would most likely narrow your choices down to Telcordia, HRD5, RDF 2000, or PRISM.

Relex as a Complete Solution

Whichever model you decide to use, the aid of an automated software package is highly recommended. Using reliability prediction software can save you substantial time, which essentially translates to saving you money. The Relex Reliability Prediction module is tailored specifically for this purpose, and it extends the advantages and features unique to individual models to all prediction models.

The Relex Reliability Prediction module can bring all of the above prediction standards to your fingertips. Furthermore, Relex provides numerous bonus features that give you unbounded flexibility. A hallmark of Relex's claim to fame is that the standards, along with their added features, can be mixed throughout the system. In addition, Relex has thousands of Commercial-Off-The-Shelf (COT) parts in its extremely large Parts Library. Systems, assemblies, and parts can be directly linked into Relex RBD, OpSim, FMEA, Fault Tree, and Event Tree modules for seamless analyses at any level. For additional Relex product information, refer to www.relexsoftware.com/products/index.asp.

Mixing Models - an Ability UNIQUE to Relex!

The Relex Reliability Prediction module allows you to use various models in the same system on a part-to-part or assembly-to-assembly basis. This can prove to be very beneficial in many circumstances. For example, suppose PRISM is being used as the basis for the analysis. The system being analyzed has many switches and connectors. However, PRISM does not include a model for switches or connectors. With Relex, this is not a problem. For those parts, MIL-HDBK-217 or Telcordia can be used to predict a failure rate.

Additionally, there are a number of other benefits available when using Relex:

• Telcordia I calculation methods for incorporating burn-in, laboratory, and field data and calculating first year multipliers can be applied to other standards.
• Predecessor data, process grades, Bayesian analysis, and NPRD/EPRD data from PRISM can be used with any model.
• Mechanical parts from NSWC-98/LE1 can be mixed with electrical parts modeled by other standards.
• Any analysis can use mission profiles (operational cycling and thermal variation information), regardless of the model.

The only exceptions to model mixing are with RDF 2000 and all model environments. Due to its unique calculation methodology, there is no approach to mixing RDF 2000 with other model features. In addition, environments are specifically defined in the standards, and there is no logical way to translate an environment defined in one standard into a different standard.

Relex Consulting Services

Sometimes a unique analysis is necessary to accurately model a system. Other times, the work coming into a reliability organization exceeds the resources available. In either of these cases, Relex consulting services can help. The experienced consultants at Relex can assist with reliability prediction analyses, building custom parts libraries, conducting electrical stress analyses, and much more. For additional information, refer to www.relexsoftware.com/services/consulting.asp.