Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CHAPTER Probability, Statistics, and Reliability for Engineers and Scientists Second Edition PROBABILITY DISTRIBUTION FOR CONTINUOUS RANDOM VARIABLES • A. J. Clark School of Engineering •Department of Civil and Environmental Engineering 5d Probability and Statistics for Civil Engineers Department of Civil and Environmental Engineering University of Maryland, College Park CHAPMAN HALL/CRC CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 1 Simulation and Probability Distribution Q Need for Simulation – Estimating the probability of failure for both explicit or implicit limit state functions without knowing analytical techniques such as the FORM method. – Simulation provides an unique opportunity to understand several important elements related to probability distributions and probabilistic analysis. 1 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 2 Simulation and Probability Distribution Q Need for Simulation (cont’d) – Simulation is used to verify the accuracy of structural reliability methods with little background in probability and statistics. – Measured data are often very limited, and making decision with small sample sizes increases the risk of incorrect decision. CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 3 Simulation and Probability Distribution Q Monte Carlo Simulation – Monte Carlo simulation has six essential elements: 1. Defining the problem in terms of all the random variables, 2. Quantifying the probabilistic characteristics of all the random variables (i.e., mean, COV, distribution type), 3. Generating the values of these random variables 2 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 4 Simulation and Probability Distribution Q Monte Carlo Simulation (cont’d) 4. Evaluating the problem deterministically for each set of all the random variables, 5. Extracting probabilistic information from n observations. 6. Determining the accuracy and efficiency of the simulation. CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 5 Simulation and Probability Distribution Q Formulation of the Problem – Consider a simply supported beam as shown w P L/2 L 3 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 6 Simulation and Probability Distribution – Assume both w and P are random variables. – Thus, the design bending moment M at the midspan of the beam is also a random variable. – The task now is to evaluate the probabilistic characteristics of the design bending moment using simulation. CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 7 Simulation and Probability Distribution – If the span of the beam is 30 feet, the expression for the design moment can written as wL2 PL + M= 8 4 = 112.5w + 7.5 P – W and P in this case are called basic random variables. 4 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 8 Simulation and Probability Distribution Q Generation of Random Variables – Computer software packages are available (e.g., Excel, Quattro Pro, etc.) – The generated random numbers from these packages are called pseudo random numbers – These numbers are generated from a welldefined and predictable process CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 9 Simulation and Probability Distribution Q Midsquare Method – – This method illustrates the problems associated with deterministic procedures The general procedure is as follows: 1. Select at random a four-digit number (seed) 2. Square the number and write the square as an eightdigit number using preceding (lead) zeros if necessary 3. Use the four digits in the middle as the new random number. 4. Repeat steps 2 and 3 to generate as many numbers as necessary 5 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 10 Simulation and Probability Distribution Q Example 1: Midsquare Method – Consider the seed number 2189. This value would produce the following: 04791721 62678889 46076944 00591361 34963569 92833225 69422224 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 11 Simulation and Probability Distribution Q Transformation of Uniform Random Numbers – The Uniform Distribution for x < 0 0 x −a FX ( x ) = b − a 1 for a ≤ x ≤ b for x > b where a < b. The mean and variance are given by µX = a+b 2 and σ 2X = (b − a )2 12 6 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 12 Simulation and Probability Distribution Q Inverse Transformation Technique or Inverse CDF Method – In the in inverse transformation technique or inverse CDF method, the CDF of the random variable is equated to the generated random number ui, that is , FX(xi) = ui, and the equation can be solved for xi as follows: xi = FX−1 (ui ) CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 13 F X (x ) F U (u ) 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 1 0.8 0.6 0.4 0.2 0 1 0 0 0.2 0.4 0.6 0.8 1 X U f X (x ) f U (u ) xi = FX−1 (u i ) U 1 ui 0 xi X 7 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 14 Simulation and Probability Distribution Q Example: Normal Distribution If X is normally distributed, that is X ~ N(µ,σ2), then Z = (X- µX)/σX is a standard normal variate, that is, Z ~ N(0,1). It can be shown that x −µX ui = FX ( xi ) = Φ( zi ) = Φ i σX or x −µX zi = i σX Thus, xi = µ X + σ X zi = µ X + σ X Φ −1 (ui ) CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 15 Simulation and Probability Distribution Q Example: Lognormal Distribution If X is lognormally distributed, that is X ~ LN(µ,σ2), then Z = (lnX- µY)/σY is a standard normal variate, that is, Z ~ N(0,1). It can be shown that ln xi − µ Y ui = FX ( xi ) = Φ (zi ) = Φ σY or ln( xi ) = µ Y + σY Φ −1 (ui ) Thus, xi = e [µY + σY Φ −1 (u i )] 8 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 16 Simulation and Probability Distribution Q Example: Simply Supported Beam The simply supported beam is subjected to the external loading w and P as shown in the figure. The probabilistic characteristics of the basic random variables are as follows: Random Variable L Mean COV 30 - Standard Deviation - Distribution Type Deterministic w 2 0.10 0.2 Normal P 20 0.15 3.0 Lognormal CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 17 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam w P L/2 L wL2 PL M= + = 112.5w + 7.5 P 8 4 9 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 18 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam 1. Simulate the design moment M for 10 values. 2. Also, Find the mean, variance, standard deviation, and coefficient of variation of M using the simulated sample values. wL2 PL M= + = 112.5w + 7.5 P 8 4 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 19 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam Mean (w ) = Stdev (w ) = 2 0.2 u1 0.388248947 0.082540402 0.891083258 0.607604281 0.682506093 0.316169559 0.949955696 0.430819593 0.697860999 0.331143892 u2 0.874573 0.840615 0.540006 0.492971 0.103666 0.312569 0.726221 0.290549 0.904197 0.375488 Mean (P ) = Stdev (P ) = w 1.94322 1.72236 2.24646 2.05462 2.09494 1.90431 2.32889 1.96514 2.10365 1.91265 M = 112.5w + 7.5 P P 23.47394542 22.95014085 20.07731226 19.72681305 16.38747866 18.38852828 21.63514737 18.21599244 24.0322073 18.8642452 Mean (M ) = Variance (M ) = Stdev (M ) = COV (M ) = 20 3 M 394.6671 365.8919 403.3068 379.0954 358.5872 352.1491 424.2632 357.6985 416.9024 356.6548 380.9 727.5 27.0 0.071 kip-ft 2 kip-ft kip- ft 10 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 20 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam – Sample Calculations Consider the second row in the table: a) w: is normal w − µW u1 = FW (w) = Φ (z ) = Φ σW x − µW z= or σW Therefore, w = µ W + σW z = µ W + σW Φ −1 (u1 ) = 2 + 0.2 × Φ −1 (0.08254 ) = 2 + 0.2 × −Φ −1 (1 − 0.08254 ) = 2 − 0.2 × Φ −1 (0.91746 ) = 2 − 0.2(1.39) ≈ 1.722 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 21 Simulation and Probability Distribution – Sample Calculations σ B) P: is lognormal σY = ln 1 + X µ X 1 1 2 µ Y = ln (µ X ) − σ 2X = ln(20) − (0.149) = 2.9846 2 2 ln P − µ Y u 2 = Φ σY or P = e [µ Y + σY Φ −1 (u 2 )] or ln P = µ Y + σY Φ −1 (u2 ) = e [2.9846+ 0.149×Φ = e [2.9846+ 0.149×Φ −1 (0.84061)] 2 3 2 = ln 1 + = 0.149 20 −1 (0.84061)] = e[2.9846+ 0.149×1] = 22.957 11 CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 22 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam – Sample Calculations • Consider the the second row in the table C) M: M = 112.5w + 7.5P = 112.5(1.722) + 7.5(22.957 ) = 365.90 kip - ft CHAPTER 5d. PROBABILITY DISTRIBUTION FOR CONTINUOUS RAND. VARIABLES Slide No. 23 Simulation and Probability Distribution Q Example (cont’d): Simply Supported Beam Mean (w ) = Stdev (w ) = 2 0.2 u1 0.388248947 0.082540402 0.891083258 0.607604281 0.682506093 0.316169559 0.949955696 0.430819593 0.697860999 0.331143892 u2 0.874573 0.840615 0.540006 0.492971 0.103666 0.312569 0.726221 0.290549 0.904197 0.375488 Mean (P ) = Stdev (P ) = w 1.94322 1.72236 2.24646 2.05462 2.09494 1.90431 2.32889 1.96514 2.10365 1.91265 M = 112.5w + 7.5 P P 23.47394542 22.95014085 20.07731226 19.72681305 16.38747866 18.38852828 21.63514737 18.21599244 24.0322073 18.8642452 Mean (M ) = Variance (M ) = Stdev (M ) = COV (M ) = 20 3 M 394.6671 365.8919 403.3068 379.0954 358.5872 352.1491 424.2632 357.6985 416.9024 356.6548 380.9 727.5 27.0 0.071 kip-ft 2 kip-ft kip- ft 12