あのあるる
成都军事夏令营Results of Statistical Analysis of
Pressure Relief Valve Proof Test Data Designed to Validate a Mechanical Parts Failure
Databa
by
Julia V. Bukowski, Ph. D.
别丢掉 林徽因for
< L.L.C.
September 30, 2007
TABLE OF CONTENTS
Executive Summary 1
Introduction 2 1.
2. Notation 3
Background 3
3.
collected4. FMEDA Analysis 4
A
4
Set
5.
Data
Analysis
of
B
9
Set
Analysis
Data
6.
of亚洲大学排名
C
11
Data
Set
7.
Analysis
of
青少年在线英语8. Discussion 13
References 15
Acknowledgements
16
16 Appendix
EXECUTIVE SUMMARY
The purpo of this document is to report on our successful efforts to validate statistically certain random equipment failure rate data ud in a mechanical parts failure rate and failure mode databa and, by extension, to validate the techniques ud to derive the data. To accomplish this, a Failure Modes, Effects, and Diagnostic Analysis (FMEDA) is initially ud to predict the uful-life failure rate for the fail-to-open condition of a particular pressure relief valve (PRV) using the failure rates from the mechanical parts databa. Next, this prediction is statistically tested against three independent data ts consisting of proof test data for PRV provided by Fortune 500 operating companies. The data ts all meet the intent of the quality assurance of proof test data as documented by the Center for Chemical Process Safety (CCPS) Process Equipment Reliability Databa (PERD) initiative. By applying the quantal respon method to the results of the PRV proof tests, it is demonstrated that the proof test data are consistent with the predictions of the FMEDA. Specifically, all of the data ts support the FMEDA result at a 95% confidence level. All analys lead to a uful-life PRV failure rate between 10-8 and 10-7 failures/hour.
It is very important to note that the results of this study cannot be ud to justify extension of proof test intervals beyond the uful life of the PRV. The small value of the failure rate derived from the FMEDA applies only to the uful life of the PRV which depends not only on the equipment's specific
ations but also on other factors, such as the ambient and process environment in which the PRV is ud and the levels and frequency of any on-line maintenance performed. Data analys place uful life in the range of 4 to 5 years.
Finally, we note that the results of the statistical analys of the three independent data ts predict an initial failure probability of approximately 1% – 1.6%. This initial failure probability is extremely significant as it accounts for the vast majority of failures obrved in proof test. This emphasizes the value of careful installation and thorough commissioning procedures. When commissioning testing cannot be done after installation, as is the ca with a PRV, both the initial probability of failing to open, as well as the PFD bad upon the random failure rate must be taken into account in the risk analysis. .
英语写信格式范文
1. INTRODUCTION
provide的用法
best memorySafety instrumented systems (SIS) are automatic systems designed for the purpo of taking action
to avoid danger or to reduce the conquences of a potentially dangerous event. International performance-bad standards [1, 2] require that designers of the systems u probabilistic analysis for equipment failures classified as “dangerous” to determine if any given design meets risk reduction goals. This is generally accomplished through an unavailability analysis. The analysis must incorporate all equipment needed for the automation system to protect against pre-identified hazards. Typical equipment includes electronic nsors, electronic signal conditioning modules, microcomputer controllers, relays, solenoids, pneumatic actuators and valves. The unavailability analysis requires, at a minimum, the uful-life failure rates and failure modes data for all subsystems.
For the electrical/electronics equipment, FMEDA techniques [3, 4] have been ud to provide failure rates, failure mode distributions, and diagnostic lf-test capability measures for subsystems bad on extensive component failure rate and failure mode databas [5, 6]. In esnce, the techniques compute a subsystem failure rate bad on the failure rates and failure modes of the components which compri the subsystem. The techniques rely on the existence of, and regulatory authorities' acceptance of, the part-level databas of failure rate and failure mode data that have been collected (over many years) from field failure data for a wide variety of electrical/electronics parts.
While it could be argued that a variety of mechanical failure models to predict failure rates exists, it is also true that most work on mechanical part failure models is focud on mechanical failure due to aging or wear-out. As a result, some published uful life failure rates for mechanical equipment are actually bad upon data points more reflective of wear-out than random failures. This report deals with failures that principally reprent random failures during the uful life of the equipment.
Until recently, a databa of uful-life failure rates and failure modes for mechanical components comparable to tho for random failures of electrical/electronic components has not existed. In [7] a technique is described for constructing a mechanical component random failure rate and mode databa bad on a combination of field failure data and expert knowledge. This databa [8], if adequately validated, and coupled with end of uful life bounding limit data, would provide the mechanical component counterpart to the electrical/electronic component databas and allow FMEDA techniques to be applied to SIS containing both electrical and mechanical components in order to generate the information required at the subsystem level to comply with the new standards [1, 2].
In this report we describe the results of a FMEDA analysis of a particular PRV to determine the uful-life failure rate of the fail-to-open condition. (The fail-to-open condition occurs when a PRV re
mains clod when test pressures (TP) reach or exceed 1.5 times the PRV t pressure.) We then u three
independent sources of proof test data to validate the predictions made by the FMEDA analysis. While this does not validate the entire mechanical part databa, it lends strong support to its validity at least with respect to the component failure modes responsible for the PRV fail-to-open condition and, by extension, to the techniques ud to create the databa.
2. NOTATION
distribution
function for time to failure
F(t) cumulative
failure/109 operating hours
FIT 1
FMEDA failure modes, effects, and diagnostics analysis
PRV pressure relief valve(s)
q i estimate for F(T i) bad on failures in i th data interval
氛围英语R(0) initial reliability; may be less than 1
R(t) reliability function for t > 0
SIS safety instrumented system(s)
SP t pressure – pressure at which PRV should open
T i equivalent failure time associated with q i
TP test pressure – pressure required during proof test to cau PRV to open
λuful-life failure rate, a constant
λ(t) failure rate as a function of time; λ may be a constant
3. BACKGROUND
In most SIS, some failure modes cannot be detected while the SIS is in operation. For example, in a PRV, the valve would normally be clod and would open only in the ca of an overpressure event. If the valve were stuck in the clod position, this would be undetectable in operation unless an overpressure event occurred and the valve failed to open. While the ensuing injuries/damages would reveal the valve failure, it is preferable to discover the failure before an overpressure event occurs. The only way to uncover the otherwi undetectable failures is through proof tests. The PRV is removed from the process and pressurized on a test bench until the valve opens; the pressure needed to open the valve is the "test pressure" (TP). Each PRV has a "t pressure" (SP), a pressure above which the valve should open in normal operation. The ratio of TP/SP is recorded. If TP/SP > 1.5, i.e., if the pressure required to open the PRV during testing is 50% or more above its t pressure, the valve is deemed to have "failed-to-open".
Proof tests are normally conducted during periodic inspection and maintenance. Thus, when an equipment failure is discovered during proof test, the actual time of the failure is not known. All that can be determined for certain
