A fluid network topology self-identification method based on depth-first search

Abstract

The application discloses a fluid network topology self-identification method based on depth-first search, which comprises the following steps: S1, classifying the states of elements and nodes according to the geometric parameters and flow parameters of the elements, and establishing a fluid network topology; S2, taking the inlet node i of an element e as a starting point, performing upstream traversal search on an upstream flow path, and synchronously updating the states of related elements; S3, taking the outlet node j of the element e as a starting point, performing downstream traversal search on a downstream flow path, and synchronously updating the states of related elements; S4, grouping all elements in a flow state to form a new fluid network topology, and updating the parameters of elements in a no-flow state; and S5, using the new fluid network topology for iterative calculation, directly assigning values to elements in a no-flow and closed state, and transmitting and updating the parameters of static nodes through adjacent nodes. The application can reduce the solution scale of the fluid network, accelerate the solution rate and stability.

Description

Technical Field

[0001] This invention relates to the field of aero-engine simulation design, and in particular to a method for self-identification of fluid network topology based on depth-first search. Background Technology

[0002] Fluid network methods are a commonly used analytical and computational method in fluid system design verification. The basic idea is to simplify the actual fluid system into a one-dimensional network composed of components and nodes. Based on a high-precision component model, the unknown parameters of the network nodes and components are solved to obtain predicted results for parameters such as pressure, flow rate, and temperature of the fluid system that meet engineering accuracy requirements. Before using fluid networks for solving, the component parameters, fluid network topology, and boundary node parameters must be determined. Traditional fluid network dynamic simulation calculations use a fixed fluid network topology and simulate the changes in engine operating conditions by given boundary parameters over time to obtain dynamic simulation results for the air system.

[0003] During actual engine operation, under the combined effects of thermal load and centrifugal force, the radial deformation rates of different components will vary. Specifically, this manifests as changes in the geometric dimensions of components at the transition / static interface; for example, the radial clearance of the air grates may change significantly or even rub against each other. Once rubbing occurs between the air grates, the flow path that originally passed through the gaps between the grates is interrupted, the fluid network topology changes, and the function of some flow paths in the air system fails under non-design conditions. This may lead to abnormal temperatures in some engine components, thereby affecting engine performance and service life. Therefore, accurately predicting the aero-thermal parameter response of the air system under variable topology flow paths has significant engineering application value.

[0004] Currently, scholars in related fields have explored various methods for topology identification of dynamic fluid networks. Wei Lianjiang, based on graph theory topology, explained the phenomenon of reverse branches in ventilation network solution and implemented automatic generation of topological relationships. Inspired by the depth-first search algorithm, he implemented path search between any two points in the ventilation network. Zhou Yan defined component types using an object-oriented approach, represented network topology relationships based on an automatically generated adjacency matrix, and analyzed and searched the network, realizing dynamic modification and topology analysis functions. Research and practice have shown that traditional planar topology network analysis methods struggle to model complex scenarios. Therefore, Wang Hai proposed an improved three-dimensional network modeling method, which overcomes some drawbacks of traditional graph theory-based methods and has significant advantages in solving three-dimensional modeling and simulation scenarios. Chen Huidong, addressing the simulation problem of complex pipe network structures and their flow and heat transfer characteristics, proposed a general pipe network simulation method based on topology decomposition and a two-layer method. This method enables automatic identification and decomposition of the pipe network, and the combination of the two-layer method achieves good convergence and solution speed.

[0005] The aforementioned studies did not consider the topology identification and evolution of fluid networks with changing topologies during dynamic processes, thus failing to achieve dynamic simulation of fluid networks with changing topologies. Traditional fluid network dynamic simulation methods clearly lack the ability to identify and process changes in network topology before and after these changes. In actual engine operation, changes in flow path topology can easily lead to abnormal component temperatures, thereby affecting engine performance, service life, and safety. Therefore, it is necessary to develop a self-identification method for the topology of dynamic fluid networks. Summary of the Invention

[0006] Purpose of the invention: The purpose of this invention is to provide a depth-first search-based self-identification method for fluid network topology that enables the identification and reconstruction of fluid network topology during dynamic simulation of air systems, obtaining a fluid network topology that matches the actual topology, thereby realizing the full-process dynamic simulation analysis and calculation of variable topology air systems.

[0007] Technical solution: The fluid network topology self-identification method of the present invention includes the following steps:

[0008] S1. Based on the geometric and flow parameters of the components in the fluid network, classify the states of the components and nodes, establish the fluid network topology, and initialize the nodes and components. Among them, flow resistance components are divided into closed and open states according to their key geometric parameters, and open states are divided into flow-through and non-flow-through states according to flow parameters. Chamber components are divided into flow-through and non-flow-through states. Based on the state of the components adjacent to the node, the node is divided into active and stationary states.

[0009] S2 takes the fluid network topology as the object and gives the element e whose state changes. Starting from the inlet node i of element e, the upstream flow path is traversed by the depth-first search algorithm. The state of all the elements that have been traversed is switched to the no-flow state.

[0010] S3, starting from the exit node j of component e, performs a downstream traversal search on the downstream flow path using a depth-first search algorithm, and switches the state of all traversed components to a no-flow state.

[0011] S4: Form a new fluid network topology by combining all the components in the flow state, and update the parameters of the components in the non-flow state.

[0012] S5 applies a new fluid network topology to iterative calculations of air systems, using a direct assignment method to process components in non-flowing and closed states, and passing and updating parameters of stationary nodes through adjacent nodes.

[0013] Furthermore, in step S1, the nodes are divided into source nodes, sink nodes, branching nodes, merging nodes, and general internal nodes: the source node is the system's inlet boundary node, and the sink node is the system's outlet boundary node; the out-degree d out ≥2 and in-degree d in Internal nodes with a degree greater than 0 are branching nodes; out-degree d out >0 and in-degree d in Internal nodes with a value of ≥2 are merging nodes; the remaining nodes are general internal nodes, which are only connected to the two elements.

[0014] When initializing nodes and components, it is necessary to determine the connection relationship and direction between components and nodes; and set the initial state of each component to be in a flowing state and the node to be in an active state.

[0015] Furthermore, in step S2, the "backtracking" condition for the upstream traversal search is as follows: if the search reaches an already visited node, a branch node, or a source node, a "backtracking" operation is performed; otherwise, the upstream traversal continues, and the state of all traversed elements is switched to a no-flow state; the upstream traversal ends when the inlet node i is returned.

[0016] Furthermore, in step S3, the "backtracking" condition for downstream traversal search is as follows: if the search reaches an already visited node, a merging node, or a sink node, a "backtracking" operation is performed; otherwise, downstream traversal continues, and the state of all traversed elements is switched to a no-flow state; the downstream traversal ends when the exit node j is returned.

[0017] Furthermore, the implementation steps of the downstream traversal search are as follows:

[0018] A1, create empty stacks for components and nodes respectively to store component and node information during the traversal process;

[0019] A2, Input the current node number, push the node onto the stack;

[0020] A3. Check whether the node m stored at the top of the stack is a boundary node or a merging node; if node m is a boundary node or a merging node, mark the node as "stationary" and proceed to step A4; otherwise, skip step A4 and proceed to step A5.

[0021] A4. Pop the top element from the stack and change its state. If the stack is empty, the process ends; otherwise, continue to step A3.

[0022] A5: Starting from the current top node of the stack, traverse the downstream nodes and components, push the corresponding nodes and components onto the stack, and execute step A3.

[0023] Furthermore, in step S4, the parameters of the element in the no-flow state are updated as follows: the flow rate m through the no-flow element. e =0, the pressure at the inlet and outlet nodes on both sides of the flow-free element is in equilibrium.

[0024] Compared with the prior art, the significant advantages of this invention are as follows:

[0025] 1. Compared with traditional topology identification algorithms, this invention provides definitions and judgment methods for component and node states, offering a basis for identifying and reconstructing fluid network topology. This invention proposes a method for traversing upstream and downstream node components after any component state switch, accurately reconstructing the fluid network topology after changes in component flow states, providing a reference for general fluid network methods. Based on the fundamental principle of depth-first traversal, this invention improves the "backtracking" condition during the traversal process, making it applicable to the judgment and identification of fluid network topology. Finally, by automatically identifying and generating new fluid networks through the algorithm, the efficiency of modeling and simulation can be significantly improved, reducing the probability of human error caused by manual modeling.

[0026] 2. This invention directly assigns values ​​to the parameters of local non-flowing state elements and stationary state nodes in the fluid network, which can further reduce the solution scale of the fluid network and accelerate the solution rate and stability. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the geometry of the air system;

[0028] Figure 2(a) is a schematic diagram of the air system flow path structure.

[0029] Figure 2(b) is an enlarged view of C in Figure 2(a);

[0030] Figure 3 This is a schematic diagram of the fluid network topology of an air system.

[0031] Figure 4(a) is a schematic diagram of the flow state of the toothed element;

[0032] Figure 4(b) is a schematic diagram of the closed state of the ferrule element;

[0033] Figure 4(c) is a schematic diagram of the flow-free state of the comb element;

[0034] Figure 5 This is a curve showing the variation of the gap between the comb teeth;

[0035] Figure 6 Here is a flowchart of the dynamic fluid network identification algorithm;

[0036] Figure 7 Flowchart of the dynamic fluid network topology reconstruction algorithm;

[0037] Figure 8 A schematic diagram of the fluid network topology after the CDP filaments (e2) have been worn down;

[0038] Figure 9 A schematic diagram of the fluid network topology after the sealing cavity teeth (e12) have been worn down. Detailed Implementation

[0039] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] This invention addresses the need for dynamic simulation calculations of fluid networks with varying topologies. It establishes a depth-first search-based self-identification method for fluid network topologies using graph theory algorithms and dynamic simulation methods for fluid networks. This method enables the identification and reconstruction of fluid network topologies during dynamic simulations of air systems, yielding fluid network topologies that match reality. This allows for full-process dynamic simulation analysis and calculation of air systems with varying topologies. To achieve the above objectives, the implementation steps of this invention are as follows:

[0041] Step one: Classify the states of components and nodes in the fluid network based on the changes in their geometric and flow parameters. Flow resistance components such as grates and orifices may experience internal flow blockage; based on their key geometric parameters, their states are classified as closed and open. The open state can be further divided into flowing and non-flowing states based on flow parameters. Chamber components generally do not experience blockage, therefore they are classified as flowing and non-flowing states. Nodes are abstract virtual objects; based on the state of the components adjacent to the node, nodes are classified as active and stationary states.

[0042] Establish the fluid network topology and initialize the nodes and components, determine the connection relationships and directions between the components and nodes, and use this as the initial topology of the air system. In the initial state, each component is in a flow state and the nodes are in an active state.

[0043] By defining the flow state of components, the component states are distinguished into flowing, closed, and non-flowing states. The flow state of a component is mainly determined by its characteristic geometric dimensions. Taking a common flow loss component as an example, the distance between the bushing and the tooth tip of a grate is called the clearance value. Generally, the clearance value is used as the characteristic geometric dimension for judging the flow state of the grate, and the orifice diameter is used as the characteristic geometric dimension for judging the flow state of the orifice component. Based on the node outlet degree d... out and in-degree d in Different sizes define four typical node types: source, sink, branching node, and merging node. Source and sink nodes are the system's inlet and outlet boundary nodes. A branching node refers to a node with a specific degree d. out ≥2 and in-degree d in Internal nodes with a degree greater than 0, merging nodes indicate degrees d. out >0 and in-degree d in Internal nodes with ≥2 are general internal nodes. Nodes not included in the above four types in the system are only connected to two elements.

[0044] Step 2: Using the network topology created in Step 1 as the object, given the component number e whose state has changed, the state of component e changes from flowing state to closed state, and the flow path that originally passed through component e is broken. Starting from the inlet node i of component e, perform upstream traversal search. During the traversal, update the state of neighboring branch component nodes. The upstream search method is inspired by the traditional depth-first search algorithm, and the "backtracking" condition in the search process is improved to better adapt to the traversal search of fluid networks. The "backtracking" condition of the upstream traversal search can be summarized as follows: if the search reaches a visited node, branch node, or source node, perform the "backtracking" operation; otherwise, continue the upstream traversal, and switch the state of all traversed components to a non-flowing state; until the starting node i is returned, the upstream traversal ends.

[0045] Based on the traditional depth-first search (DFS) method, this embodiment improves the conditions for performing the "backtracking" operation. On the basis of the traditional DFS method's "backtracking" judgment conditions, the splitting node and merging node are included in the judgment conditions for performing the "backtracking" operation. Taking upstream search as an example, the "backtracking" operation is performed when the search reaches the visited node, boundary node, or splitting node.

[0046] "Backtracking" refers to the process of a component node reverting to the position of the previously visited node in the traversal order. In programming, this is mainly achieved through recursion or stack data structures. This invention primarily utilizes the stack data structure to implement the traversal and "backtracking" of component nodes.

[0047] Step 3, similar to the upstream traversal principle in Step 2, starts from the exit node j of component e and searches the downstream flow path. The "backtracking" condition for the downstream traversal search can be summarized as follows: if the search reaches an already visited node, a merging node, or a sink node, a "backtracking" operation is performed; otherwise, the downstream traversal continues, and the state of all traversed components is switched to a no-flow state; the downstream traversal ends when the starting node j is returned.

[0048] Step four: After steps two and three, the updated components and nodes are obtained. By traversing all component and node objects, all components in the flow state are grouped into a new fluid network for iterative solution of the air system. Simultaneously, the parameters of components in the non-flow state are updated, and the flow rate m through the non-flow components is determined. e =0, the pressure at the inlet and outlet nodes on both sides of the flow-free element is in equilibrium, i.e., p i =p j For a closed element, the flow rate m e =0, meaning there is no correlation between the pressure at the inlet and outlet nodes.

[0049] Different processing methods were applied to fluid network elements and nodes in different states. For elements in the flow state and nodes in the active state, a new fluid network topology was generated for iterative solution. For elements in the non-flow state and closed state, a direct assignment method was used to process them. The parameters of stationary nodes were passed and updated through adjacent nodes.

[0050] Step 5: Update the fluid network topology regenerated in Step 4 to the air system iterative solution program. This completes the implementation of the variable topology fluid network topology self-identification function.

[0051] This invention uses a typical engine high-pressure turbine air system as an example to illustrate a method for self-identification of fluid network topology based on depth-first search. Figure 1 As shown, the air system model first draws air from the compressor outlet, which then passes through the CDP (Compressor Discharge Pressure) grates and drum cavity before merging with the pre-swirling cold air into the gas collection chamber. On the other side, air flows from the air intake port of the outer casing of the combustion chamber through a pre-designed flow channel and through the pre-swirling hole into the gas collection chamber. Part of the cooling gas in the gas collection chamber flows into the chamber between the high-pressure turbine disk and the sealing disk through the receiving hole and finally flows into the blade root to cool the first-stage moving blade. Another part of the cold air in the gas collection chamber merges into the mainstream combustion gas through the sealing cavity grates and the rim seal.

[0052] Based on the air system model, the three-dimensional model is simplified to a one-dimensional mathematical model, retaining the main flow resistance and heat exchange elements, while ignoring the geometric features of the three-dimensional circumferential distribution. The simplified flow path structure is as follows: Figure 2(a) , 2(b) As shown in Figure 2(a), nodes 1 and 6 represent the source (inlet node), nodes 14 and 16 represent the sink (outlet node), node 5 is the merging node, node 11 is the branching node, and other single-inlet / single-outlet nodes are intermediate nodes. The arrows indicate the direction of gas flow, and the naming convention for the components is "component type [component number]". Figure 2(a) is divided into four branch flow paths:

[0053] FN1 branch flow path: indicates the flow path from node 1 to node 5, which includes elements such as cavity [1], grate [2], cavity [3], hole [4], and grate [5];

[0054] FN2 branch flow path: indicates the flow path from node 6 to node 5, which includes elements such as hole [6], cavity [7], hole [8], cavity [9], and pre-rotated hole

[10] , as shown in Figure 2(b);

[0055] FN3 branch flow path: indicates the flow path from node 11 to node 14, which includes components such as grating teeth

[12] , cavity

[13] , and rim seal

[14] ;

[0056] FN4 branch flow path: indicates the flow path from node 11 to node 16, which includes elements such as holes

[15] and cavities

[16] .

[0057] This section of the air system indicates that bleed air from the compressor end and the outside of the combustion chamber cools the high-pressure turbine, and finally flows into the main flow path through the first-stage turbine's dynamic and static rim seals and the blade's internal cooling passages. In this embodiment, the internal air system is primarily considered; therefore, the blade's internal cooling flow path is not considered, and the blade root bleed air inlet is considered as the system's outlet boundary node. Based on the flow path structure, the following is obtained: Figure 3 The equivalent fluid network topology shown is as follows. Figure 3 All components are in a state of flux.

[0058] During actual engine operation, due to the combined effects of thermal load and centrifugal force, the radial deformation rates of different components will vary. Specifically, this manifests as changes in the geometric dimensions of components at the static / rotational interface. For example, the radial clearance of the comb teeth may change significantly or even rub against each other. Figure 4(a) shows the flow state of the comb teeth, where the clearance is greater than 0 and there is a pressure difference between the fluids on both sides of the comb teeth, resulting in flow. Figure 4(b) shows the closed state of the comb teeth, where the clearance is 0, the fluids on both sides are isolated, and no fluid flows through the comb teeth. Figure 4(c) shows the non-flow state of the comb teeth, where the clearance is greater than 0, the pressure difference between the fluids on both sides of the comb teeth is 0, and there is no flow. The process of the comb tooth clearance decreasing to 0 is simulated by using a ramp function with a given relative clearance size. The ramp function of the relative clearance of the comb teeth is as follows: Figure 5 As shown, t0 represents the moment when the gap begins to change, t r The ramp time is expressed in seconds. The ramp function represents the time elapsed from time t0 to t1. r In seconds, the gap between the teeth decreases linearly from its initial value to 0. r The magnitude of t determines the slope of the ramp function, i.e., t r The smaller the gap between the teeth, the faster the rate of change.

[0059] First, taking the CDP grating [2] (denoted as e2) as the object of state switching, when the gap of the e2 grating is less than the minimum gap value, the upstream and downstream traversal search is performed on the inlet and outlet nodes of e2 respectively. Figure 6 The flowchart of the fluid network topology self-identification method is presented, with an expanded explanation of the downstream traversal. The specific steps are as follows:

[0060] Step A1: Create empty stacks for components and nodes respectively to store component and node information during the traversal process;

[0061] Step A2: Input the current node number and push the node onto the stack.

[0062] Step A3: Check whether the node m stored in the top element of the stack is a boundary node or a merging node; if node m is a boundary node or a merging node, mark the node as "static" and proceed to step A4; otherwise, skip step A4 and proceed to step A5.

[0063] Step A4: Pop the top element from the stack and change its state. If the stack is empty, the process ends; otherwise, continue to step A3.

[0064] Step A5: Starting from the current top node of the stack, traverse the downstream nodes and components, push the corresponding nodes and components onto the stack, and then execute step A3.

[0065] Taking the downstream traversal search starting from the exit node 3 of component e2 as an example, first push node 3 onto the stack, and search for components and nodes downstream. If it is an unvisited node or component, push it onto the stack. If the downstream node has been visited, or is a sink or a confluence node, pop the component and node from the stack. Repeat steps A1-A5 until the traversal ends when the starting node is returned. The specific traversal process is shown in Table 1, where cavity [1] is denoted as e1, cavity [3] as e3, hole [4] as e4, and tooth [5] as e5.

[0066] Table 1 shows the downstream traversal search process starting from node 3 at the e2 component's exit node.

[0067] 1 3 3 - - 2 3，4 3，4 e3 e3 3 3，4，5 3，4，5 e3, e4 e3, e4 4 3，4 3，4，5 e3 e3, e4 5 3，4，5 3，4，5 e3, e5 e3, e4, e5 6 3，4 3，4，5 e3 e3, e4, e5 7 3 3，4，5 - e3, e4, e5

[0068] The upstream traversal search process starting from import node 2 of component e2 is shown in Table 2.

[0069] Table 2 shows the upstream traversal search process starting from import node 2 of component e2.

[0070] 1 2 2 - - 2 2，1 2，1 e1 e1 3 2 2，1 - e1

[0071] After the above traversal process, we obtain the following: Figure 8 The fluid network topology shown is such that the CDP grating [2] (e2) technology is switched to the closed state, and the e1, e3, e4, and e5 technology is switched to the non-flow state.

[0072] Next, the sealing cavity grating

[12] (denoted as e12) is used as the object of state switching. When the gap of e12 grating is less than the minimum gap value, the upstream and downstream traversal search is performed with the inlet and outlet nodes of e12 respectively. The upstream traversal search process is performed with the outlet node 12 of e12 component as the starting point, as shown in Table 3.

[0073] Table 3 shows the upstream traversal search process starting from the e12 component's exit node 12.

[0074] 1 12 12 - - 2 12，13 12，13 e13 e13 3 12，13，14 12，13，14 e13, e14 e13, e14 4 12，13 12，13，14 e13 e13, e14 5 12 12，13，14 - e13, e14

[0075] The upstream traversal search process starting from import node 11 of component e12 is shown in Figure 4.

[0076] Table 4 shows the upstream traversal search process starting from import node 11 of component e12.

[0077] 1 11 11 - -

[0078] During the upstream traversal, it is found that node 11 is a branch node, so no further traversal is performed, and the top element of the node stack is the starting node 11, thus ending the upstream traversal. After the above traversal process, the following is obtained: Figure 9The fluid network topology shown has the sealing cavity grate (e12) technology switched to the closed state, and the e13 and e14 technologies switched to the non-flow state.

[0079] Finally, adopting such Figure 7 The dynamic fluid network topology reconstruction algorithm shown obtains the fluid network topology structure composed of components and nodes in the flow state, which participates in the fluid network solution calculation; while the components in the no-flow state, closed state, and stationary state are updated by direct assignment. The specific assignment method is as follows: for components in the no-flow state, the pressure of their inlet and outlet nodes is equal, and the flow rate through the no-flow components is 0; for components in the closed state, there is no relative relationship between their inlet and outlet pressures, and the flow rate through the closed components is 0.

[0080] The fluid network topology self-identification method proposed in this invention has good applicability to general fluid networks, can identify and reconstruct new and correct fluid network topologies, and has good practicality for dynamic simulation research methods in engineering.

[0081] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for self-identification of fluid network topology based on depth-first search, characterized in that, Includes the following steps: S1. Based on the geometric parameters and flow rate of the components in the fluid network, classify the states of the components and nodes, establish the fluid network topology, and initialize the nodes and components. Among them, flow resistance components are divided into closed and open states according to their key geometric parameters. The open state is divided into flow state and no flow state according to the flow rate. The chamber components are divided into flow state and no flow state. According to the state of the components adjacent to the node, the node is divided into active state and stationary state. S2 takes the fluid network topology as the object and gives the element e whose state changes. Starting from the inlet node i of element e, the upstream flow path is traversed by the depth-first search algorithm. The state of all the elements that have been traversed is switched to the no-flow state. S3, starting from the exit node j of component e, performs a downstream traversal search on the downstream flow path using a depth-first search algorithm, and switches the state of all traversed components to a no-flow state. S4: Form a new fluid network topology by combining all the components in the flow state, and update the parameters of the components in the non-flow state. S5 applies a new fluid network topology to iterative calculations of air systems, using a direct assignment method to process components in non-flowing and closed states, and passing and updating parameters of stationary nodes through adjacent nodes.

2. The fluid network topology self-identification method based on depth-first search according to claim 1, characterized in that, In step S1, nodes are divided into source nodes, sink nodes, branching nodes, merging nodes, and general internal nodes: the source node is the system's inlet boundary node, and the sink node is the system's outlet boundary node; the out-degree... And in degree Internal nodes are branching nodes; out-degree And in degree The internal nodes are merging nodes; the remaining nodes are general internal nodes, which are only connected to the two elements. When initializing nodes and components, it is necessary to determine the connection relationship and direction between components and nodes; and set the initial state of each component to be in a flowing state and the node to be in an active state.

3. The self-identification method for fluid network topology based on depth-first search according to claim 1, characterized in that, In step S2, the "backtracking" condition for the upstream traversal search is as follows: if the search reaches an already visited node, a branch node, or a source node, a "backtracking" operation is performed; otherwise, the upstream traversal continues, and the state of all traversed elements is switched to a no-flow state; the upstream traversal ends when the inlet node i is returned.

4. The fluid network topology self-identification method based on depth-first search according to claim 1, characterized in that, In step S3, the "backtracking" condition for downstream traversal search is as follows: if the search reaches an already visited node, a merging node, or a sink node, a "backtracking" operation is performed; otherwise, downstream traversal continues, and the state of all traversed elements is switched to a no-flow state; the downstream traversal ends when the exit node j is returned.

5. The fluid network topology self-identification method based on depth-first search according to claim 4, characterized in that, The implementation steps of the downstream traversal search are as follows: A1, create empty stacks for components and nodes respectively to store component and node information during the traversal process; A2, Input the current node number, push the node onto the stack; A3. Check whether the node m stored at the top of the stack is a boundary node or a merging node; if node m is a boundary node or a merging node, mark the node as "stationary" and proceed to step A4; otherwise, skip step A4 and proceed to step A5. A4. Pop the top element from the stack and change its state. If the stack is empty, the process ends. If the stack is not empty, continue to step A3; A5: Starting from the current top node of the stack, traverse the downstream nodes and components, push the corresponding nodes and components onto the stack, and execute step A3.

6. The self-identification method for fluid network topology based on depth-first search according to claim 1, characterized in that, In step S4, the parameters of the components in the no-flow state are updated as follows: the flow rate through the no-flow component. The pressure at the inlet and outlet nodes on both sides of the flow-free element is in a balanced state.

Images