There is a system of pipes from one point to another. Pipe A is the start point, and connects left to right to Pipe B, C & D, which are in parallel and connect to the end point. The pipes can flow if and only if there is a path A to the end point. In essence, pipe A must work and at least one of B, C & D must work.

The likelihood of any pipe being blocked in $t$ amount of years is modelled by a $\text{Exp}(3)$ distribution, and all pipes are blocked independently of one another.

I have been asked to find the mean time for the system to not work, and if it is known that the system is working, what is the probability that B is not blocked?

First, I would like to find an expression for the likelihood of this system working at time $t$, here is my working:

A, B & C are the event that these pipes are not blocked

$$\begin{align} &\text{Let $Y$ be that the system works at a time $t$} \\ \therefore\quad\mathbb{P}(Y) &= \mathbb{P}(A \cap B) + \mathbb{P}(A \cap C) + \mathbb{P}(A \cap D) \\[1ex]&\text{Since all pipes block independently, and are identically distributed} \\ &= \mathbb{P}(A)\mathbb{P}(B) + \mathbb{P}(A)\mathbb{P}(C) + \mathbb{P}(A)\mathbb{P}(D) \\ &= 3\,\mathbb{P}(A)\mathbb{P}(B) \\[1ex] &{\text{Since A & B are modeled by the remainder of}\\\text{ an exponential distribution with $\lambda$ = 3}} \\ &= 3\cdot(1-(1-e^{-3t}))^{2} \\ &= 3\cdot(e^{-3t})^{2} \\ &= 3e^{-6t} \end{align}$$

For the mean time to failure, I did the following calculation:

\begin{align*} \mathbb{E}(Y) = \int_{0}^{\infty} te^{-6t} dt = \frac{1}{36} \end{align*}

Now, I am having difficulty understanding how to calculate $\mathbb{P}(B \vert Y)$

Here is my working so far, but I am almost certain it is wrong:

\begin{align*} \mathbb{P}(B \vert Y) &= \frac{\mathbb{P}(B \cap Y)}{\mathbb{P}(Y)} \\ \mathbb{P}(B \cap Y) &= \mathbb{P}(B \cap ((A \cap B) \cup (A \cap C) \cup (A \cap D)) \\ &= \mathbb{P}(A\cap B) + \mathbb{P}(B \cap A \cap C) + \mathbb{P}(B \cap A \cap D) \\ &= \mathbb{P}(A)\mathbb{P}(B) + \mathbb{P}(B)\mathbb{P}(A)\mathbb{P}(C) + \mathbb{P}(B)\mathbb{P}(A)\mathbb{P}(D) \\ &= (e^{-3t})^{2} + (e^{-3t})^{3} + (e^{-3t})^{3} \\ &= e^{-6t} + 2e^{-9t} \\ &= e^{-6t}(1 + 2e^{-3t}) \\ \therefore \mathbb{P}(B \vert Y) &= \frac{e^{-6t}(1 + 2e^{-3t})}{3e^{-6t}} = \frac{1 + 2e^{-3t}}{3} \end{align*}

Please let me know where I have messed up, thankyou so much for any help.


3 Answers 3


You need to apply inclusion exclusion to find $P(Y)$. You're double counting as it is. Luckily you can just use the fact that $P(Y) = P(A)P(B \cup C \cup D) = P(A)(1 - P(B^c \cap C^c \cap D^c ))$, where $B^c$ denotes the complement of $B$. To make computing $P(B | Y)$ easier notice that as $A$ must work, $P(B | Y) = P(B |B \cup C \cup D) = P(B)/P(B \cup C \cup D)$.


Here is another approach to find the mean time that the system doesn't work, in case you're interested.

Let $T_a,T_b,T_c,T_d\sim \exp(3)$ denote the times that pipes $A,B,C,D$ become blocked and $T$ as the time that the system shuts down.

Notice $X=T_a\sim \exp(3)$ and $Y=\max\{T_b,T_c,T_d\}\sim f_Y$ where $$f_Y(y)=9e^{-3y}\left(1-e^{-3y}\right)^2\cdot 1_{[0,\infty)}(y)$$ Since $X,Y$ are independent $f_{X,Y}\equiv f_X \cdot f_Y$ and we have $$\begin{eqnarray*}\mathbb{E}(T) &=& \mathbb{E}(T|T_a>\max\{T_b,T_c,T_d\})\mathbb{P}(T_a>\max\{T_b,T_c,T_d\})+\mathbb{E}(T|T_a\leq\max\{T_b,T_c,T_d\})\mathbb{P}(T_a\leq\max\{T_b,T_c,T_d\}) \\ &=& \mathbb{E}(Y|X>Y)\mathbb{P}(X>Y)+\mathbb{E}(X|X\leq Y)\mathbb{P}(X\leq Y) \\ &=& \int_0^{\infty}\int_0^xyf_{X,Y}(x,y)\mathrm{d}y\mathrm{d}x + \int_0^{\infty} \int_0^y xf_{X,Y}(x,y)\mathrm{d}x\mathrm{d}y \\ &=& \frac{1}{4}\end{eqnarray*}$$


The system fails by time $t$, if either pipe A fails, or all of pipes $B,C,D$ fail.

That is the event $\min\{T_A,\max\{T_B,T_C,T_D\}\}\leq t$.

So the cumulative distribution function for the time until system failure is is:

$$\begin{align}F_T(t) &= \mathsf P(\min\{T_A,\max\{T_B,T_C,T_D\}\}\leq t)\\&=1-\mathsf P(\min\{T_A,\max\{T_B,T_C,T_D\}\}>t)\\&=1-\mathsf P(T_A>t)\,\mathsf P(\max\{T_B,T_C,T_D\}>t)\\&=1-\mathsf P(T_A>t)\,(1-\mathsf P(\max\{T_B,T_C,T_D\}\leqslant t))\\&=1-\mathsf P(T_A> t)~(1-\mathsf P(T_B\leqslant t)\,\mathsf P(T_C\leqslant t)\,\mathsf P(T_D\leqslant t))\\&=(1-\mathrm e^{-3t}~(1-(1-\mathrm e^{-3t})^3))\,\mathbf 1_{0\leqslant t}\\&=(1-3\mathrm e^{-6t}+3\mathrm e^{-9t}+\mathrm e^{-12t})\mathbf 1_{0\leqslant t}\end{align}$$

The expected time until system failure may be found using that when random variable $T$ is most certainly non-negative, then we have: $\mathsf E(T)=\int_0^\infty 1-F_T(t)\,\mathrm d t$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.