Introduction to Reliability Predictions
Performing a reliability prediction analysis begins by gathering data on your system components and its operating conditions. You can begin your reliability prediction at any time, even prior to your product being built or manufactured, and then update failure related data information as your system design stabilizes. One advantageous use of reliability prediction analysis is to evaluate your product while in the design phase in order to look for potential unreliable areas and mitigate them prior to manufacture.
We recommend going through Getting Started with Relyence Reliability Prediction as a starting point for learning Relyence Reliability Prediction. From there, you can proceed to building your own analyses. The following process is intended to be a starting point; you can adapt it as required for your needs.
1. Determine which reliability prediction standard to use
There are several widely used reliability prediction standards including MIL-HDBK-217 Part Stress, MIL-HDBK-217 Parts Count, ANSI/VITA 51.1, ANSI/VITA 51.1 Parts Count, Telcordia, 217Plus Parts Stress, 217Plus Parts Count, Mechanical NSWC-11, GJB/z 299 Part Stress, and GJB/z 299 Parts Count. All provide a large set of equations for computing failure rate. Relyence Reliability Prediction also supports IEC 61709, which provides stress models to compute component failure rates by adjusting the reference failure rates to account for different operating conditions. You can choose which standard best suits your needs, or you may have contractual requirements to use a specific standard. In very broad terms, the most common usages of these standards are:
MIL-HDBK-217 Part Stress: Initially designed for use in defense and military applications, MIL-HDBK-217 is a well-known standard for reliability predictions. It is widely used in all industry sectors. The Part Stress version of the standard is the most complete failure rate analysis process and uses a large number of parameters to describe the parts in your system including rated values as well as operating values.
MIL-HDBK-217 Parts Count: A simplified version of the Part Stress version of MIL-HDBK-217 requiring less data parameters and assuming default operating conditions and stresses when performing analysis. Typically used in early design stages or when all data parameters are not known.
ANSI/VITA 51.1: VITA 51.1 is an industry standard that provides modifications to MIL-HDBK-217 F Notice 2. VITA 51.1 is a consensus-based update for MIL-HDBK-217 to provide updates that reflect more recent technology trends.
ANSI/VITA 51.1 Parts Count: A simplified version of the Part Stress version of ANSI/VITA 51.1 requiring less data parameters and assuming default operating conditions and stresses when performing analysis. Typically used in early design stages or when all data parameters are not known. ANSI/VITA 51.1 Parts Count is based on MIL-HDBK-217 F Notice 2 Parts Count and provides modifications to reflect more recent technology trends.
Telcordia: Initially developed for use in the telecommunications industry, Telcordia is also a well-known and widely accepted standard across a broad range of industries. Telcordia supports including laboratory test data and/or field data to adjust computed failure rates. It also can help assess early life product characteristics.
217Plus Part Stress: Originally named PRISM, 217Plus is a reliability prediction standard developed by Quanterion Solutions. 217Plus includes different factors than MIL-HDBK-217, such as process grades and cycling factors. The Part Stress version of the standard is more detailed and includes more data to achieve a more accurate failure rate assessment.
217Plus Parts Count: A simplified version of the Part Stress version of 217Plus requiring less data parameters. Typically used in early design stages or when all data parameters are not known.
Mechanical NSWC-11: Based on the HANDBOOK OF RELIABILITY PREDICTION PROCEDURES FOR MECHANICAL EQUIPMENT, NSWC-11, this standard supports predicting failure rates for various types of mechanical parts.
GJB/Z 299 Part Stress: Developed in China to support Chinese manufacturing, similar to MIL-HDBK-217. The Part Stress version of the standard is the most complete failure rate analysis process and uses a large number of parameters to describe the parts in your system including rated values as well as operating values.
GJB/Z 299 Parts Count: A simplified version of the Part Stress version of GJB/z 299 requiring less data parameters and assuming default operating conditions and stresses when performing analysis. Typically used in early design stages or when all data parameters are not known.
IEC 61709 provides stress models to compute component failure rates by adjusting the reference failure rates to account for different operating conditions. IEC 61709 requires you have a predicted failure rate from one of the described calculation standards or a specified failure rate from a source like NPRD or EPRD or your own source.
When using Relyence Reliability Prediction, you are not required to use just one standard. Relyence Reliability Prediction supports mixing models to provide ultimate flexibility when performing your reliability prediction analyses.
2. Define your Analysis Tree
To begin your Reliability Prediction analysis, create the hierarchical representation of your product. As in the case of our Example drone used in the tutorial, we broke the drone into the overall Quadcopter Drone, and the subsystems Motherboard, GPS, and Ground Controller.
By defining your product in this hierarchical manner, you can see the contributions of the separate subsystems to the overall failure rate and more effectively spend your time working on areas that provide the most reliability improvement.
In Relyence Reliability Prediction, use the Analysis Tree to define your products and its subsystems.
3. Add your parts
Once your subsystems are defined, you add all the parts that make up each subsystem. For example, if you are looking at a Motherboard, you add in each and every component on that Motherboard. You must know the types of all your parts (resistor, capacitor, IC, etc.) and enter these in order to perform a reliability prediction.
For each part type, there are a number of specific data parameters that are used to compute the device failure rate. Relyence Reliability Prediction automatically asks for the set of parameters needed to compute a failure rate based on the standard or standards employed. You may not know every value needed, but you can leave the value blank and Relyence Reliability Prediction will use an acceptable, average value as a default.
If you know the manufacturer Part Numbers, you can enter those, and Relyence Reliability Prediction will look up the number in its database and, if found, pull in available specific data values. You can also create your own Parts Library so that you only need to enter a part one time and then reuse it.
4. Calculate failure rate and MTBF
When all your data has been entered, a reliability calculation can be performed. Relyence Reliability Prediction will compute a failure rate for every part in your system, each subsystem, and an overall system failure rate. MTBF values will also be computed.
5. Evaluate failure rates
Once calculations are complete, you can review the failure rates to see if there are areas of high failure rate that are of concern. If so, go back and evaluate those areas and try out different design options to lower the failure rate. The more you lower the overall failure rate of your product, the better your reliability will be.
By looking at individual parts and their associated pi factors, you can determine how effective particular design changes will be. For example, if the Pi T factor for a particular part is very high, that means that lowering the temperature (the determination for Pi T) will bring down the failure rate. Perhaps adding a fan to this component can lower its operating temperature, resulting in a lower failure rate.
6. Continue to refine your analysis
As your product evolves, update your reliability analysis to reflect those changes and rerun your calculations. Also, if you have left data parameters you did not know blank, go back and enter known values as they become available. The more accurate the data in your analysis, the better your overall prediction analysis will be.