RAMS Kafka

Welcome to: 

Welcome to Reliability and Safety (R&S) Training Course P. Kafka, ESRA, Reconsult R&S Training Course CERN, February 2002

Content: 

Content Module 1: Basic Elements in Reliability Engineering Module 2: Interrelations of Reliability & Safety (R&S) Module 3: The ideal R&S Process for Large Scale Systems Module 4: Some Applications of R&S on LHC Module 5: Lessons Learned from R&S Applications in various Technologies R&S Training Course CERN, February 2002

Module 1:Basic Elements in Reliability Engineering: 

Module 1: Basic Elements in Reliability Engineering Content: Short R&S History Some Basic Terms A few Definitions and Formalisms From Components to Systems Important Methods Common Cause Failures Human Factor Issues Types of Uncertainties R&S Training Course CERN, February 2002

Module 1:Short History of R&S as Synonym of Risk: 

Module 1: Short History of R&S as Synonym of Risk Risk – very old Term (Perikles; 430 v.Chr) “the worst thing is to rush into actions before the consequences have been properly debated”, and “the Athenians are capable at the same time of taking Risk and Estimating before-hand” „Trial and Error“ Approach (‘00 – ‘40) „Worst Case - Safety Case“ Studies (‘40 –) Recognition of Stochastic Events (‘40) Development of Reliability Theory (’40 -) Reliability Studies for Complex Systems (’50 -) Comprehensive Risk Studies (’70 -) Global Risk Management: based on: Goal – Assignment – Proof (’90 -) „Risk Informed Decision Making“ (‘95 -) Risk Studies for Large Scale Test Facilities just in the beginning (’00 - ) R&S Training Course CERN, February 2002 Time Axis

Module 1:Some Basic Terms: 

Module 1: Some Basic Terms Reliability: The ability of an item to operate under designated operating conditions for a designated period of time or number of cycles. Remark: The ability of an item can be designated through a probability, or can be designated deterministic Availability: The probability that an item will be operational at a given time Remark: Mathematically the Availability of an item is a measure of the fraction of time that the item is in operating conditions in relation to total or calendar time R&S Training Course CERN, February 2002

Module 1:Some Basic Terms: 

Module 1: Some Basic Terms Maintainability: The probability that a given active maintenance action, for an item under given conditions of use can be carried out within a stated time interval when the maintenance is performed under stated conditions and using stated procedures and resources (IEC 60050)1) Remark: probabilistic definition Safety: Freedom from unacceptable risk of harm Remark: very vague definition RAMS: An acronym meaning a combination of Reliability, Availability, Maintainability and Safety R&S Training Course CERN, February 2002

Module 1:Some Basic Terms: 

Module 1: Some Basic Terms R&S Training Course CERN, February 2002 Dependability Reliability Safety Availability Maintainability Today‘s Understanding for Purists

Module 1:Some Basic Terms: 

Module 1: Some Basic Terms Hazard: A physical situation with a potential for human injury, damage to property, damage to the environment or some combination of these Individual Risk: The frequency at which an individual may be expected to sustain a given level of harm from the realisation of specified hazards Social Risk: The frequency with which a specified number of people in a given population, or population as a whole, sustain a specified level of harm from the realisation of specified hazards R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms For non-repaired items the reliability function: t ¥ R(t) = exp [ - ò l(x)dx] = ò f(x)dx 0 t where l(x) is the instantaneous failure rate of an item f(x) is the probability density function of the time to failure of the item when l(t) = l = constant, i.e. when the (operating) time to failure is exponentially distributed R(t) = exp(-lt) Example: For an item with a constant failure rate of one occurrence per operating year and a required time of operation of six month, the reliability is given by R(6m) = exp(- 1 x 6/12) = 0,6065 R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms R&S Training Course CERN, February 2002 Failure Rate l follows normally the so-called “bath tube curve” Time

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms R&S Training Course CERN, February 2002 Failure Rate l are often published in Date Books A few Examples Offshore Reliability Data; OREDA Handbook; 2nd Edition; distributed by Det Norska Veritas Industri Norge AS; DNV Technica 1992 ISBN 82 515 0188 1 Handbook of Reliability Data for Electronic Components; RDF 93 English Issue 1993; Copyright France Telecom – CNET 1993 Reliability Data of Components in Nordic Nuclear Power Plants; T-book 3rd Edition; Vattenfall AB; ISBN 91-7186-294-3 EUREDATA; Published by Joint Research Centre (JRC) Ispra, It Links for Data Informations see at ESRA Homepage

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms For non-repaired items: If observed failure data are available for n non-repaired items with constant failure rate, then the estimated value of l is given by n l = n / S TTFi i=1 where TTFi is time to failure of item i Example: For 10 non-repaired items with a constant failure rate, the observed total operating time to failures of all the items is 2 years. Hence l = 10/2 = 5 failures per year R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms For non-repaired items: ¥ MTTF = ò R(t)dt ………Mean Time To Failure 0 When time to failure is exponentially distributed, MTTF = 1 / l Example: For a non-repaired items with a constant failure rate of two failures per four years of operating time, MTTF = 1 / 2 / 4 = 2 years = 17.520 h R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms For repaired items with zero time to restoration the reliability function is given t1 R(t1,t2) = R(t2) + ò R(t2 – t) × z(t)dt 0 where R(t2), represents the probability of survival to time t2, and the second term represents the probability of failing at time t(t< t1) and, after immediately restoration, surviving to time t2 z(t) is the instantaneous failure intensity (renewal density) of the item, i.e. z(t)dt is approximately the (unconditional) probability that a failure of the item occurs during (t, t + Dt) Example: For a repaired items with a constant failure rate of one failures per operating year and a required time of operation without failure of six months, the reliability is given by R(t, t + 6) = exp (-1 x 6/12) = 0,6065 R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms For repaired items with zero time to restoration the Mean Time To Failure is given Ù ¥ MTTF = ò R(t)dt 0 When observed operating time to failures of n items are available, then an estimate of MTTF is given by Ù MTTF = total operating time / kF Example: For a repaired items with a constant failure rate of 0,5 failures per year MTTF = 1/0,5/1 = 2 years = 17.520 h R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Consider: If a repaired item with zero time to restoration operates continuously, and if the times to failure are exponentially distributed three often used terms are equal MTTF = MTBF = MUT = 1/l MTTF Mean Time To Failure MTBF Mean Time between Failure MUT Mean Uptime R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Repaired items with non-zero time to restoration The reliability of a repaired item with non-zero time to restoration for the time interval(t1, t2) may be written as t1 R(t1, t2) = R(t2) + ò R(t2 – t)n(t)dt 0 where the first term R(t2) represents the probability of survival to time t2, and the second term represents the probability of restoration (after a failure) at time t(t < t1), and surviving to time t2 n(t) is the instantaneous restoration intensity of the item When the times to failure are exponentially distributed, then R(t1, t2) = A(t1)exp(-l × (t2 – t1)) where A(t1) is the instantaneous availability at time t1, and lim R(t, t + x) = [MTTF / (MTTF + MTTR)] exp(-lt) t ®¥ R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Repaired items with non-zero time to restoration When times to failure and times to restoration are exponentially distributed, then, using either Markov techniques or the Laplace transformation, the following expression is obtained: R(t1, t2) = (µR/(l + µR) + l/(l + µR)exp[-(l + µR) t1]exp[-l × (t1 – t2)] and lim R(t, t + x) = µR /(l + µR) exp(-lx) t ® ¥ Example: For a item with l = 2 failures per operating year and a restoration rate of µR = 10 restorations per (restoration) year, and x = 1/4 lim R(t, t + 1/4) = 10/12 exp(-2 x 1/4) = 0,505 t ® ¥ R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Repaired items with non-zero time to restoration _ We can define a asymptotic mean availability A of an item _ _ A = lim A (t1, t2) = A = MUT / (MUT + MTTR) t2 ® ¥ where MTTR… Mean Time to Repair Example: For a continuously operating item with a failure rate of l = 2 failures per operating year and a restoration rate of µR = 10 restorations per year then _ A = (0, ¼) = 10/12 + 2/144 {[(exp(-12 x 0) – exp(-12 x ¼)] / ¼ - 0} = 0,886 = (0, 1) = 0,833 R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Additional Formulas see e.g. in the following Textbooks (random sample of useful books) Birolini, A., Quality and Reliability of Technical Systems; Springer 1997 2nd Edition; ISBN 3-540-63310-3 Hoyland A., & Rausand, M., System Reliability Theory; John Wiley & Sons; 1994; ISBN 0-471-59397-4 Modarres, M., Reliability and Risk Analysis; Marcel Dekker, Inc. NY; 1993, ISBN 0-8247-8958-X Schrüfer, E., Zuverlässigkeit von Meß- und Automatisierungseinrich- tungen; Hanser Verlag, 1984, ISBN 3-446-14190-1 Knezevic, J., Systems Maintainability, Chapman & Hall, 1997, ISBN 0 412 80270 8 Lipschutz, S., Probability, Schaums Outline Series, McGraw-Hill Book Company, 1965, ISBN 07-037982-3 IEC 61703, Ed 1: Mathematical Expressions for Reliability, Availability, Maintainability and Maintenance Support Terms, 1999 http://www.dke.de R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Types of Maintenance R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Maintainability Measures Probability of Task Completion: PTCDMT = P(DMT  Tst) = ò0 Tst m(t)dt Tst ….stated time for task completion m(t)…probability density function of DMT Mean Duration of Maintenance Task: MDMT = E(DMT) = ò0¥ t x m(t)dt E(DMT)… expectation of the random variable DMT R&S Training Course CERN, February 2002

Module 1:A few Definitions and Formalisms: 

Module 1: A few Definitions and Formalisms Maintenance and the Exponential Distribution m(t) = (1 / Am ) · exp – (t / Am) , t > 0 In case of exponential probability distribution: m(t) = P(DMT  t) = 1 – exp – (t / Am) DMT….Duration of Maintenance Task Am…….Scale parameter of the exp. distribution = MDMT Example: On average it takes 10 days to restore a specific machine; find the chance that less than 5 days will be enough to successfully complete the restoration: Solution: m(t) = (1/10) · exp - (t / 10) and P(DMT)  5 = M(5) = 1 - exp – 5/10 = 1 - 0,61 = 0,39 R&S Training Course CERN, February 2002

Module 1:From Components to Systems: 

Module 1: From Components to Systems We have to recall some Basic Laws of Probability A and B are mutually exclusive events than the probability that either of them occurs in a single trial is the sum of their probability Pr{A + B} = Pr{A} + Pr{B} If two events A and B are general, the probability that at least one of them occurs is: Pr{A + B} = Pr{A} + Pr{B} – Pr{AB} Two events, A & B, are statistically independent if and only if Pr{AB} = Pr{A} × Pr{B} Bayes Theorem Pr{Ai½B} = PR{Ai} × Pr{B½Ai} / [ Si Pr{B½Ai} × Pr{Ai}] More see in e. g. Schaum’s Outline Series [Seymour Lipschutz]: “Theory and Problems of Probability”, McGRAW-HILL Book Company R&S Training Course CERN, February 2002

Module 1:From Components to Systems: 

Module 1: From Components to Systems We know R(t) = e – l . t = 1 – Q(t) Q(t) = 1 – R(t) Qav ~ l . t / 2 l = 1 / MTBF [h-1] MTBF = Operational Time / Number of Stops MTTR = Sum of Repair Time / Number of Repairs For the System we yield: lS = S l = 0,0125 + 0,0125 = 0,025 1/h MTBFS = 1/(1/MTBF + 1/MTBF) =1/(1/80 + 1/80) = 40 h RS = R x R = 0,9 x 0,9 = 0,81 QS = Q + Q – (Q x Q) = 0,1 + 0,1 – 0,01 = 0,19 = 1 - 0,81 R&S Training Course CERN, February 2002 Serial System MTBF = 80 l = 1/80 = 0,0125 R = 0,9 MTBF = 80 l = 1/80 = 0,0125 R = 0,9 RS = Ri n

Module 1:From Components to Systems: 

Module 1: From Components to Systems We know R(t) = e – l . t = 1 – Q(t) Q(t) = 1 – R(t) Qav ~ l . t / 2 l = 1 / MTBF [h-1] MTBF = Operational Time / Number of Stops MTTR = Sum of Repair Time / Number of Repairs R&S Training Course CERN, February 2002 Parallel System For the System we yield lS = 2 l / 3 = 0,0083 1/h MTBFS = 80 + 80 – 1/(1/80 + 1/80) = 120 h RS = 1 - [(1 - R) x (1 - R)] = 1 – (1 - 0,9) x (1 – 0,9) = 0,99 RS = R + R – R x R = 0,9 + 0,9 – 0,9 x 0,9 = 0,99 QS = Q x Q = 0,1 x 0,1 = 0,01 RS = 1 - (1 - Ri)n

Module 1:From Components to Systems: 

Module 1: From Components to Systems R&S Training Course CERN, February 2002 Mixed System For the System we yield RS = 1– [(1– 0,95)(1– 0,99)] x 0,98 x {1– [(1– 0,99) x 0,97 x (1- 0,90)]} = 0,9995 x 0.98 x 0,99603 RS = 0,97562 ~ 0,97 The Unreliability QS = 1 – R = 0,02438 ~ 0,03

Module 1:From Components to Systems: 

Module 1: From Components to Systems R&S Training Course CERN, February 2002 Nowadays we calculate Reliability Characteristics by the means of commercial PC programs like: Cafta (USA) Care (Israel) Item Software (UK) Isograph (UK) Relex (USA) Risk Spectrum (S) Saphire (USA) For further information look for Software presentations at ESREL Conference Sites, e.g. ESREL99; ESREL2001, ESREL2002

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Failure Mode of Item Nowadays a Semi Quantitative Procedure using a three parameter grading system RPZ FMEA Principle: it represents a qualitative structure “what” can be happen “why”

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 large FTs consist of 5.000 Function Elements Fault Tree Principle: A qualitative structure “how” the system fails Using Failure Rates we can perform the Fault Tree quantification Cooling fails

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 TOP Event TE = A + BD + BE + CD + CE Simplified Cut Set Example IF A, B, C, D, E = O,1 than TE = 0,045

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Large Event Trees consist of dozen‘s of branches Event Tree Principle: it represents a qualitative structure “what” can be happen Using Probabilities we can perform the Event Tree Quantification Event Sequence Condition ESk1 Event Sequence Condition ESk2 Plant Damage State, PDSj Initiating Event IEi e.g. Cooling Pipe Break

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Pump d.n.st. Valve d.n.op. IEs Initating Events 1 - i System Functions 1 - j Fault Trees 1 - k Consequence: Type Frequency Amount Basic Events Function Failure PSA Model An Integrative Model of Event Trees and Fault Trees Large PSA Models consist of a fifty Event Trees and a hundred of Fault Trees OR

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Markov Modelling / Chains Three Types: Homogeneous Continuous Time Markov Chain Non-Homogeneous Continuous Time Markov Chain Semi-Markov Models Pros - very flexible capability good for repair good for standby spares good for sequence dependencies Good for different type of fault coverage, error handling and recovery Cons can require large number of states modelling is relative complex model often different from physical or logical organisation of the system

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Markov Modelling / Chains Simple Example Control System Two processors; 1 active, 1 hot backup Fault coverage may be imperfect c = pr {fault detected and recovery is successful given processor fault occurs} 1 – c = pr {fault is not detected or recovery is unsuccessful given processor fault occurs} l = Failure Rate µ = Repair Rate 1 F 2 l µ 2cl 2(1-c)l

Module 1:Important Methods: 

Module 1: Important Methods R&S Training Course CERN, February 2002 Stress Strength Safety Factor N/mm² pdf Structural Reliability (simplified one dimensional case) a measure for probability of failure

Module 1:Common Cause Failures: 

Module 1: Common Cause Failures Type of Failures of Items R&S Training Course CERN, February 2002

Module 1:Common Cause Failures: 

Module 1: Common Cause Failures The Boolean representation of a three component system considering Common Cause Failures (CCF) shows as following: AT = Ai + CAB + CAC + CABC AT….total failure of component A Ai…..failure of component A from independent causes CAB..failure of component A and B (and not C) from common cause CAC..equivalent R&S Training Course CERN, February 2002

Module 1:Common Cause Failures: 

Module 1: Common Cause Failures The simple single parameter model called b factor model looks like Qm = b . Qt = e.g. 0,1 that means in other words 10% of the unavailability of a system would be caused by common cause failures Some other models are shown in the next copy R&S Training Course CERN, February 2002

Module 1:Common Cause Failures: 

Module 1: Common Cause Failures R&S Training Course CERN, February 2002

Module 1:Human Factor Issues: 

Module 1: Human Factor Issues Human Factor Issues are massive involved in the R&S Technology Human Operator Reliability in control rooms Human Reliability in maintenance work Human Reliability in abnormal, accidental and emergency conditions Man – Machine Effectiveness Human Operators in control loop systems Ergonomics for control, supervision and maintenance of systems R&S Training Course CERN, February 2002

Module 1:Human Factor Issues: 

Module 1: Human Factor Issues HR Models of the first generation THERP (Techniques for Human Error Program) HCR (Human Cognitive Reliability Model) PHRA (Probabilistic Human Reliability Analysis) SLIM (Success Likelihood Index Method) Within THERP the so called HRA Action Tree represents the procedure used for estimating probabilities R&S Training Course CERN, February 2002 Ptot = A + (a×B) + (a×b×C×D) + (a×b×C×d×E) +(a×b×c×E)

Module 1:Human Factor Issues: 

Module 1: Human Factor Issues HR Models of the second generation ATHENA (NRC) CREAM (Halden) MERMOS (EDF) FACE (VTT) and many others. These models are more cognitive oriented as the first generation models The challenge nowadays is the estimation of HEPs for “Errors of Commission” For “Errors of Omission” a soundly based tool box and validated data are available R&S Training Course CERN, February 2002

Module 1:Software Issues: 

Module 1: Software Issues Why Software Reliability Prediction (SRP) is needed? Amount & Importance of software is increasing Software accounts for approximately 80 % of switch failures Software reliability is not improving fast Software is costly to fix Motivation, pressure and number of experts for doing SRP is limited Basic Questions in SRP: At what rate do failures occur ? What is the impact of these failures ? When will faults be corrected ? R&S Training Course CERN, February 2002

Module 1:Software Issues: 

Module 1: Software Issues Important Definition Failure…. An event in which the execution of a software system produces behaviour which does not meet costumer expectation (functional performance) Fault……The part of the software system which must be repaired to prevent a failure. R&S Training Course CERN, February 2002

Module 1:Software Issues: 

Module 1: Software Issues If we have an observed data example we can calculate l(t) (failure intensity/rate) if a Logarithmic Poisson Distribution is suitable: l(t) = a / (b · t + 1) The parameters to be estimated are a and b For that we need the likelihood function or the probability that the observed data occur: L(data) = j Pr{yj failure in period j} R&S Training Course CERN, February 2002

Module 1:Software Issues: 

Module 1: Software Issues Example: R&S Training Course CERN, February 2002

Module 1:Software Issues: 

Module 1: Software Issues Example: Parameter estimates: a = 2,93; b = 0,016 l(t) = 2,93 / (0,016 · t + 1) Thus: Estimates of failure intensity at 1.000 system month: l(t) = 2,93 / (0,016 x 1000 + 1) = 0,17 failures per system month Estimate the mean cumulative number of failures at 5.000 system month: 2,93 / 0,016) · ln (0,016 · t +1) = 2,93 / 0,016) · ln (0,016 x 5.000 +1) = 805 failures Today’s References [IEC 61508; Belcore Publications plus Handout] R&S Training Course CERN, February 2002

Module 1:Types of Uncertainties: 

Module 1: Types of Uncertainties Within the process of R&S we have to be aware about - at least - three type of uncertainties Parameter uncertainties (aleatory uncertainties) Model uncertainties (epistemic uncertainties) Degree of completeness Problems and unresolved issues performing an uncertainty assessment increases as this sequence But “some information about uncertainties is better than nothing” R&S Training Course CERN, February 2002

Module 1:Some Standards: 

Module 1: Some Standards IEC 300 Dependability Management IEC 605 Equipment Reliability Testing IEC 706 Guide to the Maintainability of Equipments IEC 50(191) Procedure for Failure Mode and Effect Analysis (FMEA) IEC 1014 Programmes for Reliability Growth IEC 1025 Fault Tree Analysis (FTA) IEC 1070 Compliance Test Procedure for Steady State Availability IEC 1078 Reliability Block Diagrams IEC 1123 Reliability Testing IEC 1160 Formal Design Review IEC 1146 Reliability Growth Models and Estimation Methods IEC 1165 Application of Markov Methods IEC 61508 Functional safety of electrical/electronic/programmable electronic safety related systems Others for Reliability Issues: CENELEC, IEEE, ISO, MIL, ASME, etc. R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach Content: Systems Reliability towards Risk Informed Approach Anatomy of Risk Some Definitions Living Models Reliability Growth Management Risk Monitoring How Safe is Safe Enough? R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach Deterministic in System Reliability Design Process: based on pre-defined rules and criterions derived from experiences Calculation Process: based on determined laws and formulas, calculating point values Review Process: check of the compliance with rules and standards Decision Making Process: yes / no - go / no go answers based on rule compliance R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach Probabilistic in System Reliability Design Process: based on pre-defined rules and criterions based on experiences plus probabilistic goals and targets Calculation Process: based on determined laws and formulas plus uncertainties and random variables, calculating distribution functions Review Process: check of the compliance with rules and standards plus check of the compliance with the goals and targets Decision Making Process: yes / no - go / no go answers based on risk insights R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

In the Deterministic Approach we use formalisms derived from best practice and fitted with single point values as a first guess The „Real World“ do not follow that formalisms based on single point values. Practical all values required show spreads (uncertainties) and / or a stochastic behaviour Therefore exists a challenge for modern analysis techniques and numerical solutions (e.g. Simulations) I advocate for the extension from deterministic approach towards probabilistic models to consider the stochastic behaviour and the uncertainties Module 2: Systems Reliability towards Risk Informed Approach R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach Probabilistic towards Risk Informed Approach PROS it is an extension of the deterministic basis it is supported quantitatively by historical experiences it models determined, random and uncertain elements it is quantitative and therefore appropriate for sensitivity, importance and optimisation studies it integrates design, manufacturing and operational aspects it integrates various safety issues and allows rankings it shows explicit vagueness and uncertainties CONS relatively new, more complex, and not well understood larger projects, harder to get financial support harder transformation of results into “yes or no” decisions R&S Training Course CERN, February 2002

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach R&S Training Course CERN, February 2002 „Bow – Tie“ Logic Initiating Event

Module 2:Systems Reliability towards Risk Informed Approach: 

Module 2: Systems Reliability towards Risk Informed Approach R&S Training Course CERN, February 2002

Module 2:Anatomy of Risk: 

Module 2: Anatomy of Risk R&S Training Course CERN, February 2002 Risk of a Plant Type, Amount and Frequency of Consequences Consequences Frequencies Release Parameter Receptor Parameter IEs Frequencies Conditional Probabilities Classical Decomposition

Module 2:Some Definitions: 

Module 2: Some Definitions Reliability Insights generated by Importance Measures Fussel-Vesely = [PR{top} – Pr{topïA = 0}] / Pr{top} Weighted fraction of cut sets that contain the basic event Birnbaum = Pr{topïA = 1} – Pr{topïA = 0} Maximum increase in risk Associated with component A is failed to component A is perfect Risk Achievement worth = Pr{topïA = 1} / Pr{top} The factor by which the top probability (or risk) would increase if component A is not available (not installed) Risk Reduction Worth = Pr{top} / Pr{topïA = 0} The factor by which the risk would be reduced if the component A were made perfect R&S Training Course CERN, February 2002

Module 2:Living Models: 

Module 2: Living Models In R&S we have to learn permanently from the past; that means it is an ongoing, never ending process, we call it Living Process It is strongly recommended to establish and the store all the models and data with the means of computerised tools This helps to manage in a more efficient way three important issues System Changes Personal Changes Increasing State of Knowledge R&S Training Course CERN, February 2002

Module 2:Reliability Growth Management: 

Module 2: Reliability Growth Management Basic Structure Management Testing Failure Reporting, Analysis and Corrective Action System (FRACAS) During Test we observe Type A modes (not fixed) Type B modes (fixed) At beginning of the test operation li = lA + lB Effectiveness Factor EF linh = lA + (1 - EF) lA (more details for growth models see MIL-HDBK-189) R&S Training Course CERN, February 2002

Module 2:Risk Monitoring: 

Module 2: Risk Monitoring R&S Training Course CERN, February 2002

Module 2:Risk Monitoring: 

Module 2: Risk Monitoring R&S Training Course CERN, February 2002 Operational Time Reliability Characteristic Abnormal Event Relining Unknown Level without calculation Base Line „Risk Profil“ of a plant

Module 2:How Safe is Safe Enough?: 

Module 2: How Safe is Safe Enough? Typical way of Thinking R&S Training Course CERN, February 2002 Broadly Acceptable Intolerable ALARP Risk unjustifiable Tolerable only if reduction impracticable or cost grossly disproportionate Tolerable if cost of reduction exceeds improvement Maintain assurance that risk is at this level Benchmark

Module 2:How Safe is Safe Enough?: 

Module 2: How Safe is Safe Enough? R&S Training Course CERN, February 2002 List of important qualitative Risk Characteristics related to Tolerability of Risk Qualitative Characteristics Direction of Influence Personal Control Increase Risk Tolerance Institutional Control Depends on Confidence Voluntariness Increase Risk Tolerance Familiarity Increase Risk Tolerance Dread Decrease Risk Tolerance Inequitable Distribution Depends on Individual Utility Artificiality of Risk Source Amplifies Risk Awareness Blame Increase Quest for Social and Political Response

Module 2:How Safe is Safe Enough?: 

Module 2: How Safe is Safe Enough? R&S Training Course CERN, February 2002 Risk Contours in Land Use Planning (z-Zone) [Okstad; ESREL01]

Module 2:How Safe is Safe Enough?: 

Module 2: How Safe is Safe Enough? R&S Training Course CERN, February 2002 Type of Exposure Cancer Pneumonia Mining Suicide Motorcar traffic Industrial work Chemical industry work Electrical current Lightning Individual Risk (D) number per mio persons and year

Module 3:The ideal R&M Process for Large Scale Sys: 

Module 3: The ideal R&M Process for Large Scale Sys R&S Training Course CERN, February 2002 Content: The ideal Process Anatomy of Risk From R&S Goals via the Implementation into the System to the Proof of the Compliance Constrains and Problems Implementing an ideal Process

Module 3:The ideal Process: 

Module 3: The ideal Process R&S Training Course CERN, February 2002 The ideal R&S process consists (simplified) of four main elements: Establishment of the Risk Policy Evaluation and Assessment of the Risk Concerns Performing Risk Control To do Decision Making The process is highly intermeshed and iterative! and multi-disciplinary

Module 3:The ideal Process: 

Module 3: The ideal Process R&S Training Course CERN, February 2002 To make this ideal process useful for application we need quantitative Safety Risk / Goal witch is tolerable by the society. There is trend to use as orientation for this Goal the so called Minimal Endogen Mortality (MEM Value) which is the individual risk for young people to die per year This MEM value is given in most of the countries at al level of 2 x 10-4 per person year Based on this number some experts advocate for a Global Individual Risk Goal for Hazardous Installations at a level of 10-5 per person year.

Module 3:The ideal Process: 

Module 3: The ideal Process R&S Training Course CERN, February 2002 List of important qualitative Risk Characteristics related to the Tolerability of Risk Qualitative Characteristics Direction of Influence Personal Control Increase Risk Tolerance Institutional Control Depends on Confidence Voluntaries Increase Risk Tolerance Familiarity Increase Risk Tolerance Dread Decrease Risk Tolerance Inequitable Distribution Depends on Individual Utility Artificiality of Risk Source Amplifies Risk Awareness Blame Increase Quest for Social and Political Response

Module 3:The ideal Process: 

Module 3: The ideal Process R&S Training Course CERN, February 2002 The ideal process integrates design, construction, and operational parameters from the system, the operator and the environment. The process is plant wide and comprehensive As a consequence we need for at least the analysis of hardware, software, paperware and the operator behavior The analysis of hardware is reasonably established The analysis of operator behavior is reasonably established The analysis of paperware is reasonably established The analysis of software is not well established

Module 3:Anatomy of Risk: 

Module 3: Anatomy of Risk R&S Training Course CERN, February 2002 Three main Elements (Anatomy) of Risk: what can go wrong ? how frequent is it ? what are the consequences ? Consensus across Technologies these elements describe in a most complete form the “real world” as larger the consequences as smaller the frequencies should be Unresolved issue across Technologies how safe is safe enough - tolerability of risk

Module 3:From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002

Module 3:From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 The allocation of local targets derived from a global goal for LSS is analytically not possible. It is multi parameter problem. Therefore some simplifications of the problem were developed. One of them is the so-called AGREE Allocation [US MIL HDBK-338]. It works primarily for serial systems lj = nj · [ - log(R·(T))] / (Ej·tj·N) R(tj) = 1 – {1 – [R·(T)]nj/N} / Ej with R·(T) system reliability requirement nj , N number of modules in (unit j, system) T time that the system is required to operate tj time that unit j is required during T

Module 3: From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 Proof / Review Allocation

Module 3: From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 For the allocation of local targets a linear partition to all the considered initiating events (IEs) should be used as a first approximation The allocated targets to the IEs should be subdivided also linear for all the system function modules relevant for that IE This liner allocation could be realised by spread sheet programming Commercial programs use a Simulation procedure applied to the system topology

Module 3: From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 For the proof of the global target all the frequencies calculated for similar consequences have to be summed up Commercial programs realise fault tree linking based on the identified event trees to do these summation process computerised

Module 3: From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 Allocation of MTTR [British Standard 6548] for New Designs MTTRi = (MTTRs x S1k ni • li ) / kni • li where MTTRi is the target mean active corrective maintenance time (or mean time to repair) for the a system with k consisting items The Linear Programming Method proposed by Hunt (92, 93) using different constraints produces more realistic MTTRs. The method permits better system modelling, different repair scenarios, trade offs, data updating. etc.

Module 3: From Goals towards Compliance : 

Module 3: From Goals towards Compliance R&S Training Course CERN, February 2002 Allocation of MTTR [British Standard 6548] for New Designs Example: MTTR based on BS 6584 versus LP (MTTRs 30min; MTTRmin 5 min; MTTRmax 120 on average)

Module 3:P&Cs Performing the ideal Process : 

Module 3: P&Cs Performing the ideal Process R&S Training Course CERN, February 2002 Trade-off for selecting methods: Simplicity versus Flexibility The Place of Various Modelling Techniques for LSSs

Module 4:Some Applications of R&S on LHC: 

Module 4: Some Applications of R&S on LHC Content: Where We Are Similarities and Differences in R&S Master Logic Anatomy of Risk Decomposition and Aggregation of the System Cause - Consequence Diagram R&S Training Course CERN, February 2002

Module 4:Where We Are: 

Module 4: Where We Are ACCELERATOR SYSTEMS RELIABILITY ISSUES  Burgazzi Luciano  ENEA, Bologna  Via Martin di Monte Sole, 4  40129 Bologna  Tel. 051 6098556 Fax 051 6098279  Email: burgazzi~bologna.enea.it  ABSTRACT  In the lastyears lt has been recognized the needfor investigation into the reliability of accelerator systems. This requirement results ftom new applications of accelerators (e.g. High Power Proton Accelerators for Accelerator Production of Tritium and Accelerator Transmutation of Wastes, International Fusion Materials Irradiation Facility) demanding high availability and reliability. Atpresent, although a sign~ficant history ofaccelerator operation has been accumulated over the past 50 or so years, there is a deficiency in re/iability estimates of accelerator systems due to thefact that the reliability is not a major topic asfar as most existing acceleratorsfor scient~fic experiments, in thefield ofhigh energy physics, are concerned. At the moment, despite the fact that standard reliability tools are suitable for accelerator reliability mode/, no formal reliability database for major accelerator components (such as Ion source, RE systems, etc.) 15 available, being evident that the only available data (in terms ofmean time between failures and mean time between repairs) may be inferred by the analysis of existing facilities operational experience information, leading consequently to a large uncertainty in the results (i.e. high EE, iflognormal distributions are assumed). Therefore an activity aimed at continued data collection, continued statistical inference analysis and development of mode/ing approaches for accelerators is envisaged in the next future. The present paper intends to highlight the main issues concerning the reliability assessment of accelerator machin es, focusing on the state of the art in this area and suggesting future directions for addressing the issues. In particular the topic is discussed referring mainly to Acce/erator-Driven Reactor System concept, on which the effort ofseveral research organization isfocused aiming at its development. Continued research and methodology development are necessary to achieve the future accelerator system design with characteristics satisfring the desired requirements, in terms of availability and safety. R&S Training Course CERN, February 2002

Module 4:Where We Are: 

Module 4: Where We Are R&S Training Course CERN, February 2002 REFERENCES   [1] F. E. Dunn, DC. Wade “Estimation of thermal fatigue due to beam interruptions for an ALMR-type ATW“ OFCD-NEA Workshop on Utilization and Reliability of High Power Proton Accelerators, Aix-en-Provence, France, Nov.22-24, 1999 [2] L.C. Cadwallader, T. Pinna Progress Towards a Component Failure Rate Data Bank for Magnetic Fusion Safety International Topical Meeting on Probabilistic Safety Assessment PSA 99, Washington DC (USA), August 22-26 1999 [3] C. Piaszczyck, M. Remiich, “Reliability Survey of Accelerator Facilities“, Maintenance and Reliability Conference Proceedings, Knoxville (USA), May 12-14 1998 [4] C. Piaszczyck, ‘Operational Experience at Existing Accelerator Facilities“, NEA Workshop 011 Utilization and Reliability of High Power Accelerator, Mito (Japan), October 1998 [5] VI. Martone, “IFMIF Conceptual Design Activity“ Final Report, Report ENEA RT­ERG-FUS-96-1 1(1996) [6] C. Piaszczyck, M. Rennieh “Reliability Analysis of IFMIF“ 2nd International Topical Meeting on Nuctear Applications of Aceelerator Technology ‚ ACCAPP ‘98, Gatlinburg (USA), September 20-23 1998 [7] L. Burgazzi, “Safety Assessment of the IFMIF Facility“, doc. ENEA-CT-SBA-00006 (1999) [81 C. Piaszczyck, M. Eriksson “Reliability Assessment of the LANSCE Accelerator System“ 2‘d International Topical Meeting on Nuelear Applications of Accelerator Technology ‚ ACCAPP ‘98, Gatlinburg (USA), September 20-23 1998 [9] L. Burgazzi,“Uncertainty and Sensitivity Analysis on Probabilistic safety Assessment of an Experimental Facility“ 5th International Conference on Probabilistic safety assessment and Management“ Osaka (Japan) Nov. 27-Dec 1,2000.

Module 4:Where We Are: 

Module 4: Where We Are R&S Training Course CERN, February 2002 Component [from Burgazzi, ESREL2001] Ion Source rf Antenna 6,0 E-3 Ion Source Extractor 1,0 E-5 Ion Source Turbomech Vac Pump 5,0 E-5 LEPT Focussing Magnet 2,0 E-6 LEBT Steering Magnet 2,0 E-6 DTL Quadrupole Magnet 1,0 E-6 DTL Support Structure 2,0 E-7 DTL Drive Loop 5,0 E-5 DTL Cavity Structure 2,0 E-7 High Power rf Tetrode 1,0 E-4 Circulator 1,0 E-6 Rf Transport 1,0 E-6 Directional Coupler 1,0 E-6 Reflectometer 1,0 E-6 Resonance Control 1,0 E-5 Solid State Driver Amplifier 2,0 E-5

Module 4:Where We Are: 

Module 4: Where We Are R&S Training Course CERN, February 2002 Results of Reliability Studies at LANSCE Accelerator [from Burgazzi, ESREL2001]

Module 4:Similarities and Differences in R&S: 

Module 4: Similarities and Differences in R&S R&S Training Course CERN, February 2002 It makes a difference analysing for „Reliability“ of LHC or for „Safety“. But many elements and parts of analysis are common For “R” we look mainly for failures in operational systems For “S” we look after occurring an initiating event for failures in stand-by (safety) systems In other words: R….what is the probability of loss of function of LHC S….what is the probability of a given damage (consequence) at the LHC

Module 4:Similarities and Differences in R&S: 

Module 4: Similarities and Differences in R&S R&S Training Course CERN, February 2002 R for Reliability of LHC is mainly involved in Systems Operational Effectiveness Function Performance Attributes Reliability Maintainability Supportability Functionability Availability Maintenance Operation Logistics Technical Effectiveness Operational Effectiveness

Module 4: Similarities and Differences in R&S: 

Module 4: Similarities and Differences in R&S R&S Training Course CERN, February 2002 Initiating Events Domain of „R“ Domain of „S“

Module 4:Master Logic: 

Module 4: Master Logic R&S Training Course CERN, February 2002 Analysing R we have to look first which system functions are needed for the function of the entire LHC The opposite of the function R answers for the malfunction Q (unavailability Q = 1 - R) of the LHC To answer the question: “which system functions are needed” the so-called Master Logic is an appropriate tool and a way of thinking In the next slide a simplified example, but for training we should expand it using an excel sheet

Module 4:Master Logic: 

Module 4: Master Logic R&S Training Course CERN, February 2002 IT type of graphics (trees) are very convenient to design and to show a large Master Logic

Module 4:Anatomy of Risk: 

Module 4: Anatomy of Risk R&S Training Course CERN, February 2002 Analysing S we have to look first which Type of Risks we have to evaluate. This is strongly dependent from the so-called hazard potential To answer the question: “which type of risks we have to evaluate” the so-called Anatomy of Risk is an appropriate tool and a way of thinking In the next slide a simplified example, but for training we should expand it using an excel sheet

Module 4:Anatomy of Risk: 

Module 4: Anatomy of Risk R&S Training Course CERN, February 2002 Risk of the LHC Plant Type, Amount and Frequency of Consequences Consequences Frequencies Release Parameter Receptor Parameter IEs Frequencies Conditional Probabilities Classical Decomposition

Module 4:Decomposition and Aggregation of the System: 

Module 4: Decomposition and Aggregation of the System Decomposition Down to Component Level R&S Training Course CERN, February 2002 Aggregation Up to System Function Level Different way of thinking

Module 4:Cause – Consequence Diagram: 

Module 4: Cause – Consequence Diagram R&S Training Course CERN, February 2002 „Bow – Tie“ Logic is appropriate for S Initiating Events

Module 4: Identification von IEs: 

Module 4: Identification von IEs Task: Identification of Initiating Events (IEs), which can lead at the end of an event sequence to a plant damage state. Method: Master Logic Diagram Operational Experience Analysis Logic: Bast Seminar 23.10.2002 Bergisch Gladbach

Module 4: Event Sequence Analysis: 

Module 4: Event Sequence Analysis Task: Identification of Event Scenarios and the related techn/physical parameters which can lead at the end of the event sequences to a plant damage state. Method: Event Tree, System Response Analysis Operational Experience Analysis Logic: Bast Seminar 23.10.2002 Bergisch Gladbach

Module 4: PDS Frequencies: 

Module 4: PDS Frequencies Task: Evaluation of the plant damage states frequencies at the end of all the different event sequences Method: Event Sequence Analysis, Fault Tree Analysis Operational Experience and Data Generation Formalism: f(PDS) = f(IE) . p(IE --> PDS) Analysis Logic: Bast Seminar 23.10.2002 Bergisch Gladbach

Module 4: Source Term Analyse: 

Module 4: Source Term Analyse Task: Evaluation of type, amount and frequency of possible releases of harmful material and classification into release categories (STGs) Method: Event Sequence Analysis, Fault Tree Analysis System Response Analyse Operational Experience and Data Generation Formalism: f(STG) = f(PDS) . p(PDS --> STG) Analysis Logic: Bast Seminar 23.10.2002 Bergisch Gladbach

Module 4: Consequence Model: 

Module 4: Consequence Model Task: Evaluation of type, amount and frequency of the various possible consequences around a plant Method: Event Sequence Analysis, Fault tree Analysis, Source Term Analysis, Dispersion Modelling Operational Experience and Data Generation Formalism: f(C) = f(STG) . p(STG --> C) Analysis Logic: Bast Seminar 23.10.2002 Bergisch Gladbach

Module 4: Risk Model: 

Module 4: Risk Model Task: Evaluation of the Risk Parameters for the involved Persons Method: Dose Response Modelling, Population Modelling Data Analysis Formalism: R(C) = f(STG) . C(STG) R(C)……Vector of the Risk Parameters per Year f(STG)…Vector of the Frequency of a Source Term C(STG)..Matrix of Consequence Parameter under the Condition of a Release Category Bast Seminar 23.10.2002 Bergisch Gladbach

Module 5:Lessons Learned from Various Technologies: 

Module 5: Lessons Learned from Various Technologies Content: Success Stories and Pitfalls Constraints in Data and Methods Limitations per se Technologies such as Aviation, Space, Process, Nuclear, Offshore, Transport R&S Training Course CERN, February 2002

Module 5:Success Stories and Pitfalls: 

Module 5: Success Stories and Pitfalls R&S Training Course CERN, February 2002 There is a consensus across technologies that we should know the main elements (Anatomy) of Risk: what can go wrong ? how frequent is it ? what are the consequences ?, and we should consider: as larger the consequences as smaller the frequencies should be These elements describe in a most complete form the “real world” It exist the unresolved issue across technologies “how safe is safe enough ?” – the tolerability of risk

Module 5:Constraints in Data and Methods: 

Module 5: Constraints in Data and Methods To model the Real World we have the transform the historical experience via methods and data into a prognosis for the future The data base is often sparse and limited We have to start with generic data, statistically improved by Bayesian technique, if more and more plant specific date will be available Methods should be tested by Benchmarks between indepen- dent expert teams Formal Expert Judgement procedures should be used if the evidence from the past related to the methods and the data is very limited Remember: as longer you would search in potential date bases as more reliable date you would identify R&S Training Course CERN, February 2002

Module 5:Limitations per se: 

Module 5: Limitations per se Within the R&S process we have to be aware about - at least - three type of uncertainties Parameter uncertainties (aleatory uncertainties) Model uncertainties (epistemic uncertainties) Degree of completeness Problems and unresolved issues performing an uncertainty assessment increases with this sequence But “some information about uncertainties is better than nothing” Remember: in the Deterministic Approach we generate point values only R&S Training Course CERN, February 2002

Module 5:Situation in different Technologies: 

Module 5: Situation in different Technologies Process Industry: large differences; from “yes” or “no” to risk based Offshore Industry: small differences; primarily risk-based Marine Structures: small differences; primarily risk-based Aviation: small differences; primarily risk-based Civil Engineering: differences; for specific structures risk-based Nuclear Industry: differences;tendency towards risk-based Transport: differences: tendency towards risk-based Motor Car Industry: differences; tendency towards risk-based Space Industry: strong tendency towards risk-based R&S Training Course CERN, February 2002

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 Why Events Occur (in 352 LERs, NPP; USA)

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 System Model versus Historical Experience (INEL; USA) Outage Frequency per Year for different Grid Systems

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 Informations related to the split of different causes of failures and their identification are useful

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 Hazard Rate from Test Runs [Campean, ESREL01] hj = Number of failures in current mileage band / mileage accumulated by all vehicles in current mileage band

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 The Volume and Importance of Maintenance in the Life Cycle of a System, e. g. Boeing 747; N747PA [Knezevic: Systems Maintainability, ISBN 0 412 80270 8; 1997]

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 The Volume and Importance of Maintenance in the Life Cycle of a System, e. g. Civil Aviation [Knezevic: Systems Maintainability, ISBN 0 412 80270 8; 1997] Between 1981 and 1985 19 maintenance-related failures claimed 923 lives Between 1986 and 1990 27 maintenance-related failures claimed 190 lives

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 Example Civil Aviation [Knezevic: Systems Maintainability, ISBN 0 412 80270 8; 1997] Safety demands expressed through the achieved hazard rates (1982 – 1991) for propulsion systems required by CAAM Hazard Hazard Rate High energy non-containment 3,6 x 10-8 per engine hour Uncontrolled fire 0,3 x 10-8 per engine hour Engine separation 0,2 x 10-8 per engine hour Major loss of trust control 5,6 x 10-8 per engine hour

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 If we have good (hard) statistical data then we should use it e.g. for traffic accidents normally exist good statistics. Thus, for RIDM we should use these data base [bast Heft M95; Risikoanalyse des GGT für den Zeitraum 87-91 für den Straßengüternahverkehr (GVK) und für den Benzintransport”, D]

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 If we have good (hard) statistical data in Handbooks then we should use it (see also [Birolini; Springer 1997, ISBN 3-540-63310-3]) MIL-HDBK-217F, USA CNET RDF93, F SN 29500, DIN 40039 (Siemens, D) IEC 1709, International EUREDA Handbook, JRC Ispra, I Bellcore TR-332, International RAC, NONOP, NPRD; USA NTT Nippon Telephone, Tokyo, JP IEC 1709, International T-Book (NPP Sweden) OREDA Data Book (Offshore Industry) ZEDB (NPP Germany)

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies Societal Risk of reference tunnel; RT, RT no ref.doors, RT ref. doors 50m from [D.de.Weger, et al, ESREL2002, Turin] R&S Training Course CERN, February 2002

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 There is a consensus across technologies that we should know the main elements (Anatomy) of Reliability: what can go wrong ? how frequent is it ? what are the consequences ?, and we should consider: as larger the consequences (e.g. costs) as smaller the frequencies should be These elements describe in a most complete form the “real world” It exist the unresolved issue across technologies “how reliable is reliable enough ?” – what is the most beneficial plant over time?

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 R&S has a long and successful story in industrial application The Deterministic Approach is a good basis for Safety Cases Nowadays new need an extension towards the Probabilistic Approach to model the ‘”Real World” in a more realistic manner A Risk Informed Decision Making Process (RIDM) should take place for all the safety concerns in the society Matured methods, tools and experienced experts, working since years in this field, are available and willing to help for dissemination of this RIDM process into practice The RIDM process can be used for all type of facilities

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 In the following Periodicals examples are published from different technologies Reliability Engineering & System Safety (RESS) Elsevier; http://www.elsevier.com/locate/ress IEEE Transactions on Reliability, published by IEEE Reliability Society ISSN 0018-9528 Qualität und Zuverlässigkeit; published by DGQ, Germany Carl Hanser Verlag; http://hanser.de

Module 5:Examples from different Technologies: 

Module 5: Examples from different Technologies R&S Training Course CERN, February 2002 At the following Conference Series you would get plenty R&S Informations ESREL Annual Conference Series PSAM Conference Series (every two years) RAMS Annual Conference Series SRA Annual Conference Series ICOSSAR Conference Series (every 4 years) OMEA Conference Series NASA & ESA Conferences on Risk and Reliability Plus specific Human Factor and Software Reliability Conferences, e.g. IFAC and ENCRESS

Some Key Words: 

Some Key Words Availability Verfügbarkeit Case Fall Cause Ursaache Consequence Auswirkung Event Ereignis Event Tree Ereignisbaum Example Beispiel Failure Mode Fehlerart Failure Rate Ausfallrate Fault Tree Fehlerbaum FMEA Fehler-Möglichkeits- und Auswirkungsanalyse Initiating Event Auslösendes Ereignis Maintainability Instandhaltbarkeit Maintenance Instadhaltung Minimal Cut Set Minimale Schnittmenge Probability Wahrscheinlichkeit Reliability Zuverlässigkeit Result Ergebnis Risk Risko Safety Sicherheit Solution Lösung Time Zeit R&S Training Course CERN, February 2002

That’s All: 

That’s All Thank you very much for your attention and the patience to follow all my presented issues R&S Training Course CERN, February 2002