💻 computer science

A New System Function for Maximum Processable Flow in Process Plants and Application to Reliability Assessment

This paper proposes a novel linear programming-based system performance function that integrates structural topology and process sequencing constraints to enhance process plant reliability assessment, demonstrating through benchmark applications that optimizing equipment and pipeline layouts based on the resulting component importance measures can reduce system failure probability by up to 20%.

Original authors: Ji-Eun Byun, Se-Hyeok Lee

Published 2026-03-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ji-Eun Byun, Se-Hyeok Lee

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a massive, complex factory where raw materials (like gas or chemicals) flow through a maze of pipes and machines to become a finished product. The goal is simple: get as much product out the other end as possible, as fast as possible.

But what happens if a pipe bursts or a machine breaks? Does the whole factory stop? Or does the material just find a different route?

This paper introduces a new, smarter way to answer that question. Instead of just guessing or using old, rigid checklists, the authors created a digital "traffic controller" that simulates exactly how much material can still flow through the factory when things go wrong.

Here is the breakdown of their idea using simple analogies:

1. The Problem: The "Traffic Jam" vs. The "Broken Bridge"

Traditional ways of checking safety (called Fault Trees) are like looking at a map and saying, "If any bridge on this route is broken, the trip fails." They treat all bridges as if they are equally important and interchangeable.

The Flaw: In reality, some bridges are critical, while others are just side streets. If you break a side street, traffic might just reroute. If you break the main highway, you're stuck. Old methods often miss this nuance. They might think a factory is safe because it has "enough" pipes, even if those pipes are arranged in a way that creates a dead end.

2. The Solution: The "Flow Simulator"

The authors built a new mathematical tool (a System Performance Function) that acts like a super-smart GPS for fluids.

How it works: Imagine water flowing through a series of rooms (Stages). To get from Room 1 to Room 4, the water must pass through specific checkpoints.
The Simulation: The tool uses a computer program (Linear Programming) to ask: "If I block these specific pipes and break these specific machines, what is the absolute maximum amount of water that can still get from start to finish?"
The Result: It doesn't just say "Yes" or "No." It gives you a number. Maybe the factory can still run at 100% capacity, or maybe it drops to 50%, or maybe it stops completely.

3. The "Aha!" Moment: Topology Matters

The paper uses a simple example to show why this matters.

Scenario A: You break a pipe that has a backup route nearby. The water flows around it. Result: The factory keeps running at 100%.
Scenario B: You break a different pipe that is the only way to the next room. Even though you broke the same number of pipes as in Scenario A, the factory is now stuck at 50%.

Old methods would say both scenarios are the same. The new method sees the difference because it understands the layout (topology) of the pipes, not just the count of them.

4. The Superpower: Finding the "Golden Spots"

Once the tool can calculate how much flow is possible, it can also tell you which parts of the factory are the most critical.

Think of it like a game of Jenga. Some blocks are holding up the whole tower; if you pull one, the tower falls. Others are just decorative; you can pull them, and the tower wobbles but stays up.

The authors used their tool to calculate a "Criticality Score" (called Birnbaum's Measure) for every single pipe and machine.

High Score: This pipe is a "Golden Spot." If it breaks, the whole system fails.
Low Score: This pipe is "Redundant." It's nice to have, but the system can survive without it.

5. The Real-World Win: Rearranging for Safety

The authors tested this on two real-world examples: a gas supply plant and a pressure system.

The Experiment: They took a gas plant that had a 23% chance of failing.
The Fix: They didn't buy new equipment. Instead, they used the "Criticality Scores" to rearrange the existing pipes. They moved the "Golden Spot" pipes to more secure locations and removed the "low score" (redundant) ones that weren't helping much.
The Result: By simply changing the layout, they dropped the failure risk by 20%.

The Big Takeaway

This paper teaches us that how you arrange your equipment is just as important as the equipment itself.

If you are building a factory (or even planning a city's water supply), you shouldn't just throw in extra pipes hoping for the best. You need a smart system that simulates the flow, identifies the weak links in the chain, and helps you design a layout where the system is robust enough to handle accidents without stopping production.

It's the difference between building a house with a weak foundation and a strong roof, versus building a house where the foundation is reinforced exactly where the weight is heaviest. This new tool helps engineers find that perfect balance.

1. Problem Statement

Process plants (e.g., chemical, energy, and gas facilities) are critical infrastructure subject to diverse hazards. Traditional reliability assessment methods, such as Fault Trees (FT) and Event Trees (ET), often fail to accurately model the operational reality of these systems.

The Limitation: Existing models typically treat system components as symmetric or rely on abstract logical gates (e.g., series/parallel, $k$ -out-of- $n$ ). They frequently neglect flow continuity and topological dependencies.
The Consequence: In process plants, throughput is dictated not just by the availability of individual assets but by the ability of material to flow sequentially through interconnected stages via a specific network topology. Ignoring these constraints can lead to misleading reliability assessments (e.g., failing to distinguish between a failure that blocks flow and one that merely reduces redundancy).

2. Methodology

The authors propose a novel System Performance Function based on Linear Programming (LP) to address these gaps.

A. System Representation

The process plant is modeled as a multi-stage network:

Nodes ( $N$ ): Represent equipment or pipeline joints. They are categorized as station nodes (where flow is held for processing/storage) or non-station nodes.
Edges ( $E$ ): Represent pipelines connecting nodes between stages ( $m$ to $m+1$ ).
Constraints: The model incorporates stage sequencing, node capacities ( $c_k$ ), and edge capacities ( $c_{ij}^m$ ).

B. The Optimization Formulation

The core of the methodology is an LP problem designed to calculate the Maximum Processable Flow ( $u$ ).

Objective: Maximize the total flow $u$ originating from the first stage.
Constraints:
1. Flow Conservation: Ensures inflow equals outflow at non-station nodes.
2. Stage Continuity: Ensures that for every station node in stages $2$ to $M$ , the total inflow matches the target flow $u$ .
3. Capacity Limits: Flow on any edge cannot exceed the edge's capacity or the capacities of its connected nodes.
Uncertainty Integration: Asset states (functional vs. failed) are modeled as binary random variables ( $X_n$ ). When an asset fails, its capacity is set to zero. The system failure probability is estimated by running Monte Carlo simulations where the LP is solved for each realization of asset states.

C. Component Importance Analysis

To support layout planning, the framework integrates Birnbaum's Importance Measure:
$BI_n = P(S=1 | X_n=1) - P(S=1 | X_n=0)$
This quantifies how much the system's survival probability changes if a specific component $n$ is guaranteed to work versus guaranteed to fail. This allows engineers to rank assets by their criticality to the system's topological flow.

3. Key Contributions

Topology-Aware Performance Function: Unlike traditional FTs, this method explicitly models the physical network structure and sequential process constraints, capturing the "flow continuity" requirement.
Computational Efficiency: By formulating the problem as a Linear Program, the method remains computationally efficient even for large-scale systems (hundreds of nodes/edges), enabling rapid Monte Carlo reliability assessments.
Differentiation of Failure Scenarios: The method can distinguish between scenarios that traditional models treat as identical. For example, it can differentiate between a failure that blocks a critical path (reducing flow) and a failure in a redundant path (maintaining flow), whereas a standard FT might classify both as simple component failures.
Layout Optimization Framework: The integration of importance measures provides a quantitative basis for reconfiguring plant layouts to maximize reliability.

4. Results and Validation

The authors validated the framework using two benchmark examples:

Case 1: Didactic Example (Small Scale)

Comparison: The LP-based function was compared against a Fault Tree model.
Finding: In a scenario where one node and one pipeline failed, the Fault Tree predicted the system survived in both topological variations. However, the proposed LP function correctly identified that in one variation, the flow dropped to 50% (system failure), while in the other, it remained at 100%. This proved the LP method's superior ability to capture topological nuances.

Case 2: Pressure Regularisation System (Real-world Layout)

Setup: A system with 25 nodes (original) vs. 44 nodes (expanded).
Result: Despite the expansion having a similar maximum throughput, the failure probability dropped from 23.3% to 18.3%. This highlighted that topology changes, not just capacity additions, drive reliability.

Case 3: Gas Supply Plant (Large Scale)

Setup: 57 nodes, 102 edges, 87 random variables. Target flow: 0.5.
Baseline: System failure probability was 22.9%.
Optimization: Using Birnbaum's measures, the authors identified the top 6 critical assets and the bottom 6 non-critical assets. They proposed a reconfiguration: adding the 6 critical assets and removing the 6 least critical ones.
Outcome: The new topology reduced the system failure probability to 18.4%, a ~20% reduction, using the same number of total components.

5. Significance

Risk-Informed Design: The framework shifts reliability management from a component-centric view to a system-centric, flow-based view. It allows engineers to optimize plant layouts specifically for flow continuity under failure conditions.
Cost-Effective Reliability: The study demonstrates that significant reliability improvements (e.g., 20% reduction in failure probability) can be achieved through strategic reconfiguration rather than simply adding more equipment.
Scalability: The use of Linear Programming makes the approach viable for industrial-scale plants with hundreds of components, overcoming the computational bottlenecks of complex state-space models.
Future Applications: The authors suggest this framework can be coupled with reliability-based optimization for automated topology design and integrated with structural health monitoring for real-time reliability updates.

In conclusion, this paper provides a robust mathematical tool that bridges the gap between network flow theory and process plant reliability, offering a more accurate and actionable approach to safety and risk management in industrial facilities.