Abstract
In chemical process simulation, the computational load of solving recycle loops can be significantly reduced by assigning weights to process streams. This paper proposes a two-tiered tearing criterion: first, minimizing the number of cut streams (cycle intervals), and second, selecting alternative streams with the minimum weight for each interval. We present an efficient decomposition algorithm that utilizes a weighed reachable vector method for cycle detection and a three-step "searching, judging, and regulating" procedure to determine the optimal cut-set. Evaluations on benchmark problems demonstrate that the algorithm consistently achieves the minimum weighing sum under the minimum number of cut streams.
Introduction
Simulating complex chemical plants often involves solving systems of interconnected units with recycle loops. In a sequential modular approach, these loops must be "torn"—broken at specific streams—to allow for iterative calculation. The set of chosen streams is known as the "cut-set."
Historically, tearing algorithms focused primarily on minimizing the number of cuts. However, not all streams are equal; some carry more variables or exhibit higher sensitivity, leading to a greater computational "weight." A cut-set with fewer streams might still be computationally more expensive than a slightly larger set of "lighter" streams. This paper addresses the challenge of finding an optimal cut-set that minimizes both the number of cuts and their total computational weight.
Methodology
The proposed approach models the chemical process as a weighed directed graph (digraph), where nodes represent units and edges represent streams with associated weights.
The algorithm operates in several key phases:
- Initial Sequencing and Matrix Construction: Unlike standard methods, this algorithm constructs a weighed adjacency matrix (see Figure 4). This matrix is used to establish an efficient initial sequence of nodes, prioritizing paths that likely lead to optimal solutions.
- Cycle Interval Detecting: Using a "weighed reachable vector method," the algorithm identifies recycle loops through matrix operations. This systematic approach avoids the complexities of manual graph tracing.
- Searching, Judging, and Regulating: For each identified cycle interval, the algorithm searches for the stream with the minimum weight. It then "judges" whether cutting this stream successfully breaks the cycle. If not, a "regulating" step modifies the local weighed adjacency matrix to find the next best candidate. This iterative refinement ensures the selection of the absolute minimum weight cut for that interval.
- Precedence Determination: Finally, Boolean algebra determines the optimal calculation order for the simulation.
Experimental Evaluation
The algorithm was tested against several standard benchmark digraphs from chemical engineering literature:
- Forder & Hutchison (1969): Used as a step-by-step working example.
- Sargent & Westerberg (1964): A 19-node, 31-stream problem.
- Christensen & Rudd (1969): Two distinct flowsheets.
- Sulfuric Acid Plant (1971): A complex real-world plant model with 44 units and 69 streams.
Results and Discussion
The proposed algorithm was compared with the Basic Tearing Algorithm (BTA), the K & S algorithm, and the M & H-2 algorithm.
In the Sargent & Westerberg problem, while most algorithms identified a 6-cut solution, the proposed method achieved a total weight of 8, significantly better than M & H-2 (weight 9) and BTA (weight 16).
The superiority was most evident in the Sulfuric Acid Plant case. The proposed algorithm found a solution with only 5 cuts and a total weight of 9. Competing methods required 6 to 9 cuts, with weights ranging from 10 to 13.
Conclusion
By integrating stream weights into the tearing criteria and utilizing vector-based matrix operations, this algorithm provides a robust and efficient solution for process decomposition. It strictly enforces the minimum number of cuts while optimizing for computational load, outperforming traditional methods on both simple and complex chemical plant flowsheets.
Summary updated