You are on page 1of 6

Title:

Advanced Methods for Assessment of System Reliability and Residual Life

Applicant Details: Ankit Rana | Indian Institute of Technology, Bombay | 3rd Year, Mechanical Engineering

Question 1
How would you implement a function that would generate a random number from a probability distribution? How would you do that for a normal distribution?

Solution
Sampling random numbers from various probabilistic distributions are necessary for all stochastic simulations. Given problem is to generate samples for a normal distribution to be used as an input for simulation. It is difficult to generate samples from arbitrary distributions. They are generated after generating IID Uniform random numbers between 0 and 1. Thus the very first step is to have a good Random Number Generator (RNG) ~ IID U[0,1] Let us assume that we have a good RNG ~ IID U[0,1]
Random Number Generator and its properties: Numbers produced should be distributed uniformly on [0,1] and be uncorrelated. Generator should have sufficiently long cycle Must be fast and use less memory Must be able to reproduce a given stream of random numbers exactly Most common is Linear Congruential Generator (LCG) Zi = (a.Z i - 1 + c) mod m where a, c, m are known constants and Z0 is the seed. Ui = Zi / m Another is combination of two MRGs as proposed by LEcuyer in 1999 which has period of approximately 2191. The initial 10,000 seed vector is determined empirically which passes all the tests of uniformity and independence (Runs, K-S, chi-square, etc.) (Tae-Soo Kim, 2006)

Next step is to transform these random numbers into variates from arbitrary input distributions. Some of the general purpose techniques are as follows: (Jerry Banks, 2004) Inverse-Transform Convolution Composition Acceptance-Rejection Based on special properties Inverse-Transform is most useful when the cumulative distribution function (cdf) is known F(x) and is of the form such that its invers can be calculated F-1(x). Example: Exponential Convolution is used when the desired random variable X can be expressed as a sum of other random variables that are IID and can be readily generated that direct generation of X. Implementation Example: Erlang(k,)

Composition technique is useful when the distribution function F, from which we want to generate X, is a convex combination of other distributions functions F1, F2, and sampling from these Fjs is easier as compared to F. Example: double exponent function Acceptance-Rejection is used when other techniques fail or are insufficient. Random number is generated by the rejection method. Polar method which is used to sample random variates from normal distribution using uniform numbers is discussed in the later part of this document. Normal distribution: The normal distribution is a continuous probability distribution that has a bell-shaped probability density function, known as the Gaussian function or informally the bell curve

where parameter is the mean and 2 is the variance

Inverse Transform Approach: Normal distribution is a continuous distribution function and no close form inverse exists. Thus to apply the inverse-transform method one might wish to approximate the inverse of the cdf by numerically integrating and searching for cdf. Although approximations are available for many distributions they are used depending on the accuracy required and the field in which it is being implemented. One of the Inverse approximations, proposed by Schmeiser is as follows

for 0.0013499 R 0.9986501 which matches the true normal distribution with one digit after decimal point.
Direct Approach (Box Mueller): 1. Generate random numbers U1 & U2 ~ IID U[0,1] 2. Let V1 = and V2 = We get two random variates V1 & V2 ~ N[0,1] (IID) Advantages: More accurate when compared with indirect method Disadvantages: Computational load as it involves trigonometric functions Polar Approach: 1. Generate random numbers U1 & U2 ~ IID U[0,1] 2. Let V1 = 2U1 1 and V2 = 2U2 1 3. 4. If W > 1 go back to step 1 else Y=

5. X1 = V1.Y and X2 = V2.Y Advantage: Polar method is computationally faster than direct approach as it replacing trigonometric function with division. Disadvantage: Some random numbers generated are wasted. Though not too many as acceptance is with probability /4

Question 2
Your task is to predict how many miles a car can drive on 10 liters of fuel. What factors would you take into account?

Solution
Considering the high cost of fuel these days along with huge amount of fuel consumption and its impact on the environment it is necessary to make vehicles that are fuel efficient. Not just the specifications of the engine but several human and environment factors also affect the fuel consumption of the car. We can divide the factors into three major groups:

Vehicle parameters

This included factors such as mass, idle fuel or emission rates, fuel or emission efficiency rates, rolling resistance factor, frontal area and the aerodynamic drag coefficient. A better aerodynamic profile will encounter less drag and thus less fuel consumption. An increase in mass will lead to increase in the inertia and thus more power will be required to accelerate the vehicle thus leading to low fuel economy. Better design of the drivetrain can prevent transmission losses. Manual transmission is better than Automatic transmission and a direct drive transmission is better than overdrive in terms of preventing transmission losses. Also losses in four wheel drive will be less than that of two wheel drive vehicles

Rolling resistance is influenced by the air pressure and deflated tyres will experience high resistance nut too much of pressure will also lead to wear of tyres. It also depends on the weight and the speed. Air-fuel ratio and the compression determine the engine efficiency. Air-fuel ratio near the stoichiometry ratio of 14.7 has maximum efficiency. Higher compression promote high peak temperature and lower exhaust temperature thus leading to better fuel heat energy conversion to mechanical energy.

Road parameters

includes factors such as speed, acceleration rate and traffic conditions.

Speed determines the drag force acting on the vehicle as described above. Higher acceleration requires higher power input. Fuel consumption is less on a highway. This is because inertia component (which is very high in city continuous braking and accelerating) of the power required is not present.

Environment parameters

Handling of the vehicles significantly determines the fuel consumption.

Other than the human factor, ambient conditions also affect the economy. On a cold day the economy would be less as the density of air would be higher and thus more drag force would be experienced by the vehicle.

When a car moves it has to overcome certain resistance offered by its own inertia, drag and friction by ground along with drive train losses and other accessories (like Air conditioning). It is the engine that must provide power to overcome these resistances. These, according to (Richardson, 1982), can be calculated as Zt = Zd + Zr + Za + Ze + Zm where Zd , Zr , Za , Ze and Zm are, respectively, the power required to overcome vehicle drive-train resistance, tyre rolling resistance, aerodynamic drag, inertial and gravitational resistance, and for the accessories (such as air-conditioning) Zd = 2.36 10-7 v2M Zr = (3.72 10-5v + 3.09 10-8v2) M Za = 1.29 10-5 Cd Av3 Ze = 2.78 10-4 (a + g sin) Mv where M (kg) is the inertia mass of vehicle, v (km/hr) the vehicle speed, a (m/s2) the vehicle acceleration, Cd the aerodynamic drag coefficient, A (m2) the vehicle frontal area, and the road gradient.

Estimation of fuel economy: Assumptions: The vehicle is moving at a constant speed of 80 kmph. Mass of the vehicle is 1000kg Ambient Temperature = 20oC Fuel used is petrol whose calorific value is 35 MJ/litre Methodology adopted

Power required estimation Drivetrain Zd = (2.36 10-7 ) (v2M) = 2.36 10-7 x 802 x 1000 = 1.5 kW Rolling Zr = (3.72 10-5v + 3.09 10-8v2) M = 3.17 kW Drag (can also be calculated using 0.5 v3 A CD) Za = 1.29 10-5 Cd Av3 = 6.6 kW Assuming no other accessories and vehicle on flat ground Zt = Zd + Zr + Za = (1.5 + 3.17 + 6.6) kW = 11.27 kW Thus total power required is 11.27 kW Energy from fuel combustion: x calorific value of fuel ..(1)

This is the total energy and there are losses due to frictional parts, efficiency restriction of engine, drivetrain losses, energy carried by the exhaust gases, etc. Otto engine efficiency is defined as where r is the compression ratio For r = 10 and =0.3 for combustion reactants we get = 0.498 Note this is theoretical efficiency. Practically this will be lower because of inefficient compressors and turbines and heat loss from combustion chamber. Thus actual = 0.498 x 0.7 = 0.3486 This is further combined by the mechanical efficiency which is generally 70% Thus the net thermal efficiency t = 0.3486 x 0.7 0.25 Thus the energy rate by combustion of the fuel would be t x Equating eq. (1) and eq. (2) 11.27 kW = t x x calorific value of fuel = 0.35 x x 35 MJ/litre x calorific value of fuel (2)

Therefore, = 11.27 kW / (0.25 x 35 MJ/litre) = 1.288 x 10-3 litre/s = 4.64 litre/hr As per the assumption that the car runs at uniform speed of 80kmph (50 mph) in 10 litres of fuel the vehicle will cover 80km x (10/4.64) = 171.4 km 110 miles Fuel economy 11 miles/litre

Works Cited
Jerry Banks, B. L. (2004). Discrete-Event System Simulation. Prentice Hall. Radmi Akcelik, M. B. (n.d.). Operating cost, fuel consumption, and emission models in aaSIRDA and aaMOTION. 25th Conference of Australian Institute of Transport Research (CAITR 2003), University of South Australia, Adelaide, Australia, 3-5 December 2003. Rajput, R. K. (2005). Internal Combustion Engines. New Delhi: Laxmi Publications. Richardson, A. (1982). Stop-start fuel consumption rates. Second Conference on Traffic, Energy and Emission. Melbourne.

You might also like