Failure rate prediction is a method that is applicable mostly during the conceptual and early design phases, to estimate equipment and system failure rate. It can also be used in the manufacturing phase for product improvement.
失效率預估方法大多應用在概念定義與設計初期，推定設備與系統失效率，亦可在製造階段作為產品改善之用。
Three basic techniques can be adopted:
可以採用三種基本技法：
- failure rate prediction at reference conditions, also called parts count analysis;
- failure rate prediction at operating conditions, also called parts stress analysis;
- failure rate prediction using similarity analysis.
零件計數法：根據零件之參考條件，預估失效率。
零件應力法：根據零件之操作條件，預估失效率。
相似分析法：比較新設計設備與既有設備之性能資料，預估可靠度。
The choice of which technique to use depend on the available level of knowledge of the system at the moment the reliability prediction is performed and also on the acceptance degress of approximation.
A.1.1.2 參考條件失效率預估及操作條件失效率預估 (failure rate prediction at reference conditions and failure rate prediction at operating conditions)
In the first two cases, the analyst needs to know the number and type of components that constitute the system. The analyst also needs to know the operating conditions for which the failure rate prediction is being performed. If the operating conditions are the same as the reference conditions for the components, then no account of the operating conditions needs to be made. However, when the failure rate prediction is for operating conditions that differ from the reference conditions, then the specific application conditions of the component are taken into account (electric, thermal, environmental) using models developed for the purpose. For accurate predictions, a reliable failure rate database is needed. IEC 61709 gives recommendations on how failure rates can be stated at so-called "reference condtions" in such a database, but it does not contain failure rate data. Several failure rate data handbooks have been developed and some of them are commercially available. However, reliability calculations can be time-consuming and therefore commercial software tools are available to perform these calculations.
Failure rate prediction is based upon the following assumptions:
- components are logically connected in series (i.e., each one is necessary for the system);
- component failure rate are constant over time;
- component failure rate are independent.
失效率預估法係基於下列假設條件：
零件之間彼此為邏輯串聯（亦即，每一項都是系統所必需的）。
零件失效率對時間而言為常數。
零件失效率彼此為獨立。
These assumptions need to be discussed with reference to the system under study since they can lead to a worst-case estimate when redundancies at the higher levels of assembly are present.
A.1.1.3 相似性分析失效率預估 (failure rate prediction using similarity analysis)
Similarity analysis includes the use of fielded (in-service) equipment performance data to compare new designed equipment with predecessor equipment for predicting end item reliability.
Comparisons of similar equipment may be made at the end item, sub-assembly, or component levels usingthe same field data, but applying different algorithms and calculation factors to the various elements. Elements to be compared may include:
- operating and environmental conditions (measured and specified);
- design features;
- design processes;
- reliability assurance processes;
- manufacturing processes;
- maintenance processes;
- components and materials.
相似性分析法考量因素：
操作與環境條件
設計性質
可靠度保證過程
製造過程
維護過程
組件與材料
For each of the above elements, a number of sub-elements should be compared. As examples, operating and environmental conditions may include steady-state temperature, humidity, temperature variations, electrical power, duty cycle, mechanical vibration, etc.; equipment design features may include number of components (separated according to major component family), number of circuit card assemblies, size, weight, materials, etc.
Similarity analysis should include necessary alogrithms or calculation mehtods used to quantify similarities and differences between the equipment being assessed and the pdedecessor equipment.
Element similarity analysis is used when a similarity analysis is not possible because no predecessor equipment is sufficiently similar or available for a one-to-one comparison with the newly designed equipment being assessed. Element similarity analysis is the structured comparison of elements of ths new equipment with similar elements of a number of different predecessor equipment, for which reliability data are available.
A.1.1.4 利益 (benefits)
- Time and cost of analysis are very low, provided reference data and models are available.
- The necessary input information and data are small and therefore adapted to the situation in the early design and development phase.
- Basic information on component reliability is gained in the early design and development phase.
- Adapted to manual and computerized calculations.
- Little training is necessary.
只需要提供參考資料及模型，分析所需時間與成本低。
必要的輸入資訊與資料少，可用於設計與發展初期階段。
A.1.1.5 限制 (limitations)
- The functional structure (e.g., lower level redundancies) of a system cannot be considered, and therefore only simple structures lend themselves to parts count analysis.
- The precision level of the predictions may be low, especially for small sub-systems and limited run productions, since published or collected data are valid only statistically, i.e., they require large samples.
- The evaluation of failure modes and mechanisms and their effects is not possible.
無法考量系統功能架構(亦即較低層級複聯)，只能使用零件計數法處理簡單結構。
預估的精確程度不高，特別是小型次系統或限量生產，因為出版或蒐集得到資料，需要較大樣本才具統計意義。
無法評估失效模式與機制，及其效應。
A.1.1.6 標準 (standards)
The Applicable IEC standard is IEC 61709.
A.1.1.7 Example for an integrated circuit (as given in IEC 61709)
For a bipolar random access memory, the failure rate is stated as λ[sub]ref[/sub] = 10[sup]-7[/sup] h[sup]-1[/sup] in a trust-worthy database based on the following reference conditions stated in IEC 61709:
- reference ambient temperature: θ[sub]amb, ref[/sub] of 40 °C;- reference self-heating: 20 °C.
What is the value of the failure rate at an ambient temperature θ[sub]amb[/sub] = 70 °C with the same self-heating?
Step 1: The failure rate model at operation conditons is stated in IEC 61709 as λ = λ[sub]ref[/sub] × π[sub]T[/sub]
Step 2: From Figure A.1 (taken from IEC 61709), the factor for temperature influence follows to π[sub]T[/sub] = 3.4,
- using the reference virtual junction temperature,
θ[sub]1[/sub] = θ[sub]amb, ref[/sub] + ΔT[sub]ref[/sub] = 40 °C + 20 °C = 60 °C
- and the actual virtual junction temperature
θ[sub]2[/sub] = θ[sub]amb[/sub] + ΔT[sub]ref[/sub] = 70 °C + 20 °C = 90 °C