Network Reliability

What this lecture covers

This recitation problem applies independence to a network reliability question: given a network of components that each work with probability p independently of one another, what is the probability that a working path exists between two points, A and B? It shows how to break a complex network into simpler serial and parallel pieces and recombine their probabilities.

After watching, you should be able to compute the probability that a serial or parallel arrangement of components is operational, and combine those results to analyze a more complex network by decomposing it into independent sub-collections.

Key ideas

• Serial structure: k independent components in series all need to work, so the success probability is p^k.
• Parallel structure: k independent components in parallel need only one to work, so the success probability is 1 - (1-p)^k.
• Independent collections: groups of components that share no elements behave as independent events, letting a large network be split into smaller, separately analyzable pieces.
• Divide and conquer: a complex network can be decomposed into a chain of serial and parallel sub-networks, each solved separately and then multiplied or combined according to the serial/parallel rules.

Walkthrough

Serial and parallel building blocks (1:01)

The lecture derives the success probability formulas for a serial chain (p^k) and a parallel bank (1 - (1-p)^k) of independent components, both following directly from independence.

Decomposing the full network (3:02)

The example network from A to B is split into independent segments (A-to-C, C-to-E, E-to-B). The C-to-E segment is itself further broken into nested serial and parallel pieces until every unknown reduces to a combination of the two basic formulas, which are then substituted back in to reach the final probability.

Before you watch

• Review independence of events and the multiplication rule from earlier lectures in this course, since the whole solution rests on treating separate components and collections as independent.

Check your understanding

1. Why does a serial connection require all components to work, while a parallel connection requires only one?
2. Why can two collections of components that share no elements be treated as independent?
3. How does breaking the network into smaller serial/parallel pieces make the overall probability calculation manageable?