Maximum Processable Flow for Process Plant Reliability

Tags: publication process-plants linear-programming

Process plants must keep flow moving in order — unloading, storage, vaporisation, supply. Any reliability assessment that ignores this sequencing and topology misses how plants actually fail.

Conventional fault-tree models treat assets within the same stage as symmetric and cannot tell apart two failure scenarios with very different consequences for throughput.

We propose a new system performance function that computes the maximum processable flow across interconnected stages, formulated as a linear program and integrated into reliability analysis [Byun & Lee, 2026].

Applied to a 57-node, 102-edge gas supply plant, the function evaluates in fractional seconds — fast enough to drive 10⁷ Monte Carlo samples and Birnbaum-based component importance ranking.

Reconfiguring the plant by swapping six low-importance assets for six high-importance ones reduced the system failure probability from 22.9% to 18.4% — about a 20% reduction — with no change in pipeline or equipment count.

If you’d like to try it:

• Datasets: https://github.com/jieunbyun/network-datasets/tree/main/datasets/process_plants