A method and apparatus for distributed compressor margin calculation

Abstract

The application provides a distributed compressor margin calculation method and device, and belongs to the field of compressor aerodynamic optimization. The method provided by the application comprises the following steps: a distributed parallel computing architecture is built; a flow field convergence judgment rule based on efficiency residual is constructed; a surge boundary self-searching strategy with dynamic back pressure adjustment is adopted, multi-round full three-dimensional flow field parallel calculation is performed relying on the distributed parallel computing architecture, the calculation results of each round are verified in combination with the flow field convergence judgment rule, the outlet back pressure is iteratively adjusted to approach the near surge point of the compressor, until the near surge point of the convergence and divergence critical state is determined and the performance parameters are obtained; the total pressure ratio and flow performance parameters of the compressor design point are extracted, the near surge point corresponding performance parameters are combined, and the compressor margin is calculated. The distributed compressor margin calculation method and device provided by the application are used to realize the dual improvement of the margin calculation precision and the calculation efficiency, and the calculation resources are maximized.

Description

Technical Field

[0001] This application relates to the field of compressor aerodynamic optimization technology, and in particular to a method and apparatus for calculating the margin of a distributed compressor. Background Technology

[0002] As a core characteristic parameter for compressor aerodynamic optimization, margin directly defines the safe operating boundary of the compressor. The accuracy of its calculation results is crucial to the scientific nature and safety of compressor aerodynamic optimization design. As an important research object of aerodynamic optimization, distributed compressors require margin calculation as a necessary prerequisite for controlling equipment operating stability and optimizing aerodynamic performance. It is also a key link in realizing high-precision aerodynamic optimization design of compressors.

[0003] Currently, there are two main approaches to margin calculation for distributed compressors: one is the exact calculation method, which searches for surge boundaries by continuously changing boundary conditions and gradually increasing back pressure until the flow field calculation diverges to determine the near-surge point, thus completing the margin calculation. This process requires solving multiple full three-dimensional fluid fields. The other is the pre-estimation method, used in a few margin optimization examples, which simplifies the calculation process and reduces the amount of calculation for the full three-dimensional fluid field by predicting the near-surge point. However, both existing calculation methods have significant problems and shortcomings, making it difficult to meet the actual needs of high-precision aerodynamic optimization design for distributed compressors. While the near-surge point prediction method can reduce the amount of calculation, it suffers from extremely low calculation accuracy, failing to provide reliable data support for high-precision optimization design. On the other hand, the exact calculation method requires solving the full three-dimensional fluid field multiple times, resulting in extremely long calculation times. This also means that margins in compressor aerodynamic optimization design are often only used as verification parameters, rather than core optimization parameters. Meanwhile, due to the uncertainty of convergence during margin calculation, the traditional multi-task calculation mode also suffers from serious resource waste. This mode adopts a task pre-allocation strategy to distribute the calculation tasks evenly to each node, but ignores the differences in the computing capabilities of each node. Not only does it require tedious adjustments to the task allocation and information synchronization between nodes in advance, but it is also very easy for nodes to be idle. It is also difficult to handle abnormal tasks that occur during the calculation process in a centralized and efficient manner.

[0004] Therefore, there is an urgent need for a method to address the pain points of insufficient accuracy, computation time, and inefficient resource utilization in existing distributed compressor margin calculations, so as to achieve a dual improvement in margin calculation accuracy and computational efficiency, while maximizing the utilization of computing resources. Summary of the Invention

[0005] In view of this, this application provides a distributed compressor margin calculation method and apparatus to solve the pain points of insufficient accuracy, calculation time and inefficient resource utilization in existing distributed compressor margin calculations, thereby achieving a dual improvement in margin calculation accuracy and calculation efficiency, while maximizing the utilization of computing resources.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] A first aspect of this application provides a method for calculating the margin of a distributed compressor, the method comprising:

[0008] Determine the objective function for the compressor surge boundary search;

[0009] The surge boundary search objective function is solved based on the Pre-MOEA algorithm. In each solution cycle, the working condition nodes from the previous cycle are evolved to determine the number of evolved working condition nodes to be evaluated. A matching number of computational nodes are activated based on this number to complete the evolution evaluation of all working condition nodes to be evaluated. The efficiency residual of each evolved working condition node is calculated, and convergence of each working condition node is determined based on the efficiency residual. Converged and effective working condition nodes are selected. From these effective working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computational node completes the evolution evaluation task of at least one working condition node to be evaluated.

[0010] Connect the surge critical points of all solution cycles and calculate the surge boundary;

[0011] The compressor margin is calculated based on the surge boundary.

[0012] A second aspect of this application provides a distributed compressor margin calculation device, the device comprising a determination module, a solution module, and a calculation module;

[0013] The determining module is used to determine the objective function for the surge boundary search of the compressor;

[0014] The solution module is used to solve the surge boundary search objective function based on the Pre-MOEA algorithm. In each solution cycle, the working condition nodes from the previous cycle are evolved to determine the number of evolved working condition nodes to be evaluated. A matching number of computing nodes are activated based on this number to complete the evolution evaluation of all the working condition nodes to be evaluated. The efficiency residual of each evolved working condition node to be evaluated is calculated, and the convergence of each working condition node is determined based on the efficiency residual. Converged and effective working condition nodes are selected. From the effective working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computing node completes the evolution evaluation task of at least one working condition node to be evaluated.

[0015] The calculation module is used to connect the surge critical points of all solution cycles and calculate the surge boundary;

[0016] The calculation module is also used to calculate the compressor margin based on the surge boundary.

[0017] The distributed compressor margin calculation method and apparatus provided in this application solve the surge boundary search objective function using the Pre-MOEA algorithm. Under a distributed computing architecture, it ensures that each computing node completes the evolutionary evaluation task of at least one operating condition node to be evaluated. Simultaneously, it combines an efficiency residual convergence judgment mechanism and an adaptive evolutionary wake-up rule. These three elements work together to achieve accurate, stable, and efficient acquisition of the surge critical point within each solution cycle. On one hand, leveraging the multi-objective optimization and preference guidance capabilities of the Pre-MOEA algorithm, the operating condition nodes are directed to search towards the critical stability region with low flow rate and high pressure ratio. Combined with distributed parallel computing, this achieves precise matching of computing power and tasks, avoiding resource idleness or overload and improving the overall solution speed. On the other hand, it relies on efficiency residuals to perform quantitative convergence judgment on the operating condition nodes, making node generation and critical state identification highly adaptive. This ensures that the iteration process is stable and does not diverge or produce invalid calculations, while accurately capturing the critical state between convergence and divergence, significantly improving the accuracy and reliability of surge critical point identification. Furthermore, it uses efficiency residuals to perform flow field convergence judgment. The system can use the standard deviation of continuously calculated data as a stability evaluation index to truly reflect the stability of the flow field at the end of the evaluation period. This avoids convergence misjudgment caused by instantaneous fluctuations, local anomalies, or computational instability, ensuring that all effective operating condition nodes entering the optimal screening have a stable, reliable, and credible flow field state. This improves the accuracy of the surge critical point from the source, thereby ensuring that the surge boundary obtained by subsequent fitting truly fits the actual stable operating limit of the compressor. Ultimately, it achieves the effects of high margin calculation accuracy, fast iteration speed, strong robustness, and high distributed resource utilization, providing stable, efficient, and automated margin calculation support for compressor aerodynamic design and optimization. Attached Figure Description

[0018] Figure 1 A flowchart of the distributed compressor margin calculation method provided in Embodiment 1 of this application;

[0019] Figure 2 This is a schematic diagram of the structure of the distributed compressor margin calculation device provided in Embodiment 2 of this application. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0021] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0023] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0024] Figure 1 This is a flowchart illustrating the distributed compressor margin calculation method provided in Embodiment 1 of this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0025] S101. Determine the objective function for the compressor surge boundary search.

[0026] Specifically, the surge boundary search objective function is a target function specifically constructed for the Pre-MOEA multi-objective optimization algorithm for automatically searching the compressor surge boundary. Its role is to guide the evolution of the operating node in each solution cycle. Through efficiency residual convergence judgment, back pressure adjustment and divergence judgment, the operating node gradually approaches the surge critical point of "convergence and divergence critical state", providing a unified optimization criterion and solution basis for connecting all critical points, constructing the surge boundary and calculating the margin.

[0027] In specific implementation, the objective function for the compressor surge boundary search is determined, including: determining the optimization objective of the surge boundary search objective function; the optimization objective includes locating the compressor surge critical point; determining the input variables of the surge boundary search objective function; the input variables include the back pressure, total pressure ratio, and flow rate of the compressor before and after the evolution of the operating condition node in each solution cycle; setting the optimization constraints of the surge boundary search objective function; the optimization constraints include the convergence criterion, the back pressure adjustment step size range, and the physical quantity conservation requirements during the evolution of the operating condition node; defining the output variables of the surge boundary search objective function; the output variables are the surge critical points in each solution cycle; and constructing the surge boundary search objective function based on the optimization objective, input variables, optimization constraints, and output variables.

[0028] Specifically, firstly, the optimization objective of the surge boundary search objective function is clearly defined, taking the surge critical point where the compressor is in a state of convergence and divergence as the core optimization objective, providing a clear direction for the construction of the objective function; secondly, the input variables of the surge boundary search objective function are determined, selecting the compressor back pressure, total pressure ratio, and flow parameters corresponding to the evolution of the operating node before and after each solution cycle as input variables, providing the basic data required for the calculation of the objective function; subsequently, the optimization constraints of the surge boundary search objective function are set, including the convergence criterion based on efficiency residuals and back pressure adjustment during the evolution of the operating node. The minimum and maximum step sizes, as well as the physical quantity conservation requirements during the flow calculation process, are all incorporated into the constraints to standardize the solution range and judgment rules of the objective function. Next, the output variables of the surge boundary search objective function are defined, and the surge critical points that meet the critical state and are selected in each solution cycle are set as output variables to clarify the final output result of the objective function. Finally, guided by the optimization objective determined above, based on the input variables as the data foundation, the optimization constraints as the limiting rules, and the output variables as the result direction, the four are integrated and matched to complete the construction of the surge boundary search objective function.

[0029] S102. Solve the surge boundary search objective function based on the Pre-MOEA algorithm.

[0030] In each solution cycle, the working condition nodes of the previous cycle are evolved to determine the number of working condition nodes to be evaluated after evolution. Based on this number, a matching number of computing nodes are activated to complete the evolution evaluation of all the working condition nodes to be evaluated. The efficiency residual of each evolved working condition node to be evaluated is calculated, and based on the efficiency residual, it is determined whether each working condition node to be evaluated has converged. Converged and effective working condition nodes are selected. From the effective working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computing node completes the evolution evaluation task of at least one working condition node to be evaluated.

[0031] Specifically, the Pre-MOEA algorithm is a preference-guided multi-objective evolutionary optimization algorithm used to handle multi-objective, constrained optimization search problems. Through population evolution, non-dominated sorting, and preference direction guidance, it automatically guides the search point towards the target region to determine critical boundaries and optimal critical points. In this application, the Pre-MOEA algorithm is used to automatically search for compressor surge boundaries.

[0032] The solution cycle refers to the complete process unit of the Pre-MOEA algorithm in iteratively solving the surge boundary search objective function, which involves "evolution of the load case node → generation of the node to be evaluated → parallel evaluation → convergence judgment → selection of the optimal node". The Pre-MOEA algorithm gradually approaches the surge critical point by repeatedly executing this process unit, and each complete cycle constitutes one solution cycle. The efficiency residual is a quantitative indicator used to determine whether the load case node to be evaluated has converged. It is obtained by statistically processing the historical efficiency calculation data of the evolved load case node and is used to characterize the stability of the flow state of the load case node.

[0033] Furthermore, the "previous cycle's working condition nodes" refer to the top N working condition nodes with the optimal objective function that were selected and retained after the previous solution cycle ended, serving as the base population for evolutionary operations in the current solution cycle; the evolved working condition nodes to be evaluated refer to the new working condition nodes generated after performing evolutionary operations on the working condition nodes of the previous cycle, which require performance evaluation and convergence judgment, and are the objects that need to be calculated and verified in the current solution cycle; the computing nodes refer to the computing power nodes in the distributed architecture used to perform evolutionary evaluation tasks, including local nodes and remote nodes, which are awakened according to the number of working condition nodes to be evaluated, and are used to complete the computing tasks of all working condition nodes to be evaluated in parallel; the converged effective working condition nodes refer to the working condition nodes to be evaluated after evolutionary evaluation that meet the convergence conditions and have a stable and effective flow state after efficiency residual judgment, and are the basis for subsequent selection of optimal nodes; the top N evolved working condition nodes with the optimal objective function refer to the top N nodes selected from the converged effective working condition nodes based on the surge boundary search objective function, which will serve as the "previous cycle's working condition nodes" for the next solution cycle and continue to participate in iterative evolution.

[0034] It should be noted that the Pre-MOEA algorithm transforms the compressor surge boundary search into a multi-objective optimization problem with stability constraints. Using total pressure ratio and flow rate as optimization objectives, it performs a directional search towards the critical stability region with low flow rate and high pressure ratio under preference guidance. During the solution process, the algorithm uses operating condition nodes as an evolutionary population. In each solution cycle, it first evolves the previous generation of operating condition nodes to generate a new generation of operating condition nodes to be evaluated. Then, based on the number of nodes to be evaluated, it wakes up a corresponding number of computing nodes, using a distributed parallel approach to complete the evolutionary evaluation of all nodes. Finally, it uses the computational efficiency residual to determine whether the nodes have converged, selecting stable nodes. The algorithm first selects the effective operating nodes; then, using non-dominated sorting, it selects the top N nodes with the optimal objective function from the effective nodes as the base population for the next generation of evolution. Through a cyclical process of multi-cycle iterative evolution, parallel evaluation, convergence screening, and optimal node retention, the algorithm gradually brings the operating nodes closer to the surge critical state, automatically capturing the surge critical points that satisfy "convergence but close to divergence" at each speed. After all cycles are completed, the algorithm outputs all surge critical points for subsequent fitting to form a complete surge boundary. Finally, based on the design point and combined with the critical operating conditions on the surge boundary, the compressor margin is automatically calculated, realizing the integrated and automated solution of the surge boundary and margin. N is set according to actual needs; in this embodiment, it is not limited to this.

[0035] In specific implementation, during each solution cycle, the operating condition nodes of the previous cycle are evolved to determine the number of operating condition nodes to be evaluated after evolution. This includes: evolving the operating condition nodes based on the compressor back pressure and performance parameters corresponding to the effective operating condition nodes that converged in the previous cycle; if the current operating condition node is determined to be converged after evolution evaluation, the compressor back pressure is increased according to a preset back pressure change rate to generate the next operating condition node to be evaluated; if the current operating condition node is determined to be divergent after evolution evaluation, the current compressor back pressure is reduced and the back pressure adjustment step size is decreased to regenerate the next operating condition node to be evaluated; the steps of operating condition node evolution, evolution evaluation, convergence judgment, divergence judgment, and back pressure adjustment are executed cyclically until a critical operating condition node is obtained. The critical operating condition node is determined as the surge critical point, and the number of all operating condition nodes to be evaluated in this solution cycle is counted; if the convergence amplitude of the critical operating condition node is less than a preset value, the back pressure of the critical operating condition node is increased by the minimum step size, and the critical operating condition node diverges.

[0036] Specifically, upon entering a single solution cycle, the first step is to retrieve the converged valid operating condition nodes selected from the previous solution cycle. The compressor back pressure parameters, total pressure ratio, flow rate, and other performance parameters corresponding to these nodes are extracted. Based on this, an evolutionary operation is performed on the operating condition nodes. After initial evolution, the evaluation and iteration phase begins. Evolutionary evaluation is conducted on each evolved operating condition node. After evaluation, convergence and divergence judgments are performed simultaneously. Based on the judgment results, corresponding node generation operations are executed: if the current operating condition node is determined to be in a convergent state after evolutionary evaluation, the compressor back pressure corresponding to that operating condition node is gradually increased according to a pre-set back pressure change rate, and the next operating condition node to be evaluated is generated based on the adjusted back pressure parameters; if the current operating condition node is determined to be in a divergent state after evolutionary evaluation, the current compressor back pressure value of that operating condition node is decreased, and the back pressure adjustment step size is reduced. The next operating condition node to be evaluated is regenerated based on the adjusted back pressure parameters and step size. Following the above rules, the entire process of evolving operating condition nodes, evaluating evolution, determining convergence, determining divergence, and adjusting compressor back pressure is continuously and iteratively executed until a critical operating condition node that meets the requirements is selected. This critical operating condition node must satisfy the following conditions: its convergence amplitude must be less than a preset convergence threshold, and after increasing its corresponding compressor back pressure by the minimum executable step size, the node immediately exhibits a divergent state. This critical operating condition node is then determined as the surge critical point for this solution cycle. After determining the surge critical point, all operating condition nodes to be evaluated generated within this solution cycle are statistically analyzed one by one, and the total number of evolved operating condition nodes to be evaluated within this cycle is obtained.

[0037] The method provided in this embodiment can evolve based on the backpressure and performance parameters of the effective working condition nodes in the previous cycle within each solution cycle. It adaptively adjusts the backpressure increase rate or backslides the step size according to the convergence or divergence state, accurately locating critical working condition nodes and surge thresholds during iterative iterations, while simultaneously counting the number of nodes to be evaluated. On one hand, this highly couples the generation of working condition nodes with the search process for surge thresholds, ensuring that the number of nodes to be evaluated matches the actual search requirements and avoiding the waste of computational resources caused by the generation of invalid nodes. On the other hand, it can adaptively adjust the node generation strategy based on convergence and divergence results, making the search process more stable, reliable, and rapidly approaching the surge boundary. Combined with the distributed parallel mechanism that wakes up corresponding computing nodes based on the number of nodes, it can achieve precise matching of computing resources and evaluation tasks, ensuring that each computing node undertakes at least one evaluation task. This improves parallel computing efficiency, shortens the overall solution cycle, and guarantees the accuracy and stability of surge threshold location, providing a high-quality data foundation for subsequent construction of surge boundaries and accurate computational margins. Simultaneously, it achieves adaptive search strategy, efficient resource scheduling, and guaranteed computational accuracy.

[0038] In specific implementation, the number of computing nodes matching the quantity is activated to complete the evolutionary evaluation of all the nodes under evaluation, including: using the compressor design point as the initial benchmark, the compressor back pressure adjustment task corresponding to the node under evaluation is encapsulated as a task to be calculated and pushed to the shared queue task pool; local computing nodes and remote computing nodes autonomously obtain the task to be calculated from the shared queue task pool, and execute the calculation task using the Threading parallel process mode. A single thread executes the evolutionary evaluation task of one node under evaluation and calls multi-core parallel computation to complete the calculation; after each computing node completes the evolutionary evaluation, the calculation results are sent back to the shared node for unified storage, and the number of evolutionary evaluation tasks completed in this solution cycle is counted and determined.

[0039] Specifically, firstly, the compressor design point is used as the initial benchmark for the entire evolutionary evaluation process. The compressor backpressure adjustment tasks corresponding to all evaluation nodes within the current solution cycle are uniformly encapsulated into standardized tasks to be calculated. All tasks to be calculated are then pushed sequentially to a shared queue task pool in the distributed architecture for unified management. Next, local and remote computing nodes are started. Each node, based on its idle status, autonomously retrieves a task from the shared queue task pool. After retrieving a task, the node executes the calculation using a Threading parallel process mode, strictly maintaining that each thread independently executes the evolutionary evaluation task for one evaluation node. Simultaneously, multi-core parallel computing capabilities are utilized to complete the evolutionary evaluation and related parameter calculations for that node. Once all local and remote computing nodes have completed their assigned evolutionary evaluation tasks, each node sends the generated calculation results back to the shared node, which centrally receives, organizes, and stores all results. Finally, all completed evolutionary evaluation tasks within this solution cycle are statistically analyzed to confirm the total number of completed tasks.

[0040] In specific implementation, the efficiency residual of each evolved working condition node to be evaluated is calculated, and the convergence of each working condition node is determined based on the efficiency residual. Converged and valid working condition nodes are then selected. This includes: reading the computational history data corresponding to each evolved working condition node to be evaluated; extracting the efficiency computational history data of the last preset number of steps in the evolutionary evaluation process of the target working condition node to be evaluated, calculating the standard deviation of the efficiency computational history data, and using the standard deviation as the efficiency residual of the target working condition node to be evaluated; comparing the efficiency residual with a preset convergence judgment threshold; if the efficiency residual is less than the convergence judgment threshold, the target working condition node to be evaluated is determined to have converged; if the efficiency residual is not less than the convergence judgment threshold, the target working condition node to be evaluated is determined to have not converged, and idle computing nodes are awakened to re-complete the evolutionary evaluation of the target working condition node to be evaluated.

[0041] Specifically, firstly, for each completed evolutionary node to be evaluated, all computational history data generated and stored during its evolution process is read. After reading, for each target node to be evaluated, the efficiency calculation history data of the last preset number of steps in the evolution process is extracted from its corresponding computational history data (the preset number of steps is set according to actual needs, and this embodiment does not limit it; for example, in one embodiment, the preset number of steps is 200). Mathematical operations are performed based on this continuous efficiency calculation history data to obtain the standard deviation of the data, and this standard deviation is directly defined as the efficiency residual of the current target node to be evaluated. After completing the efficiency residual calculation, the efficiency residual corresponding to each target node to be evaluated is compared with a pre-set convergence judgment threshold (e.g., 0.0001). The process involves comparing the values ​​one by one. When the efficiency residual is less than the convergence threshold, the target node to be evaluated is determined to be in a convergent state and included in the valid node sequence. When the efficiency residual is greater than or equal to the convergence threshold, the target node to be evaluated is determined to be in a non-converged state. The currently idle computing nodes are then scheduled and awakened, and the complete evolutionary evaluation process is re-executed for the non-converged target node to be evaluated. Following the same processing method, the operations of data reading, efficiency data extraction, efficiency residual calculation, convergence determination, and re-evaluation of non-converged nodes are performed sequentially for all evolved nodes to be evaluated. Finally, the screening process of all nodes is completed, and all converged valid nodes in this solution cycle are obtained.

[0042] The method provided in this embodiment reads the computational history data of the node to be evaluated, extracts the efficiency calculation data of the last preset number of steps, calculates the standard deviation as the efficiency residual, and then compares the efficiency residual with the convergence judgment threshold to determine convergence. Simultaneously, it wakes up idle computing nodes to re-evolutionarily evaluate non-converged nodes. This method can accurately identify stable flow nodes with stable flow states using stable and quantifiable standards, avoiding misjudgments caused by data fluctuations or computational instability, and ensuring the reliability and consistency of effective node conditions. By using the efficiency data of the last preset number of steps to calculate the standard deviation, it can fully reflect the stability of the node in the later stages of evolutionary evaluation. This approach allows convergence judgments to better align with actual critical state determinations. Furthermore, it automatically triggers a re-evaluation mechanism for non-converged nodes, effectively eliminating invalid data and filling in qualified nodes. This ensures that all operating condition nodes entering the subsequent optimal node selection process meet stability requirements, improving the accuracy of surge critical point location. Simultaneously, it utilizes distributed parallel computing resources for automatic recalculation without manual intervention, enhancing the robustness, continuity, and solution efficiency of the entire algorithm iteration. This provides a solid and reliable data foundation for the subsequent Pre-MOEA algorithm to select the optimal top N nodes, accurately search for surge boundaries, and precisely calculate compressor margins.

[0043] S103. Connect the surge critical points of all solution cycles and calculate the surge boundary.

[0044] Specifically, the surge critical point refers to the operating condition node that meets the convergence condition and is at the critical state of convergence and divergence during the iterative solution process of the Pre-MOEA algorithm. This node diverges after its compressor back pressure is increased by the minimum step size, and is the surge critical operating point determined in each solution cycle. The surge boundary refers to the continuous boundary curve formed by connecting and fitting the surge critical points under different operating conditions obtained from all solution cycles, representing the critical limit for stable operation of the compressor under different operating conditions.

[0045] In practice, the surge critical points of all solution cycles are connected to calculate the surge boundary, including: recording the surge critical points determined in each solution cycle, extracting the compressor back pressure and performance parameters corresponding to each surge critical point to form a surge critical point dataset; filtering the surge critical point dataset, removing abnormal critical points, and retaining all convergent surge critical points that meet the critical state; and using linear fitting or interpolation methods to connect the surge critical points of all solution cycles in sequence to form a continuous compressor surge boundary.

[0046] Specifically, firstly, the surge critical points determined in each solution cycle are fully recorded, and the compressor back pressure, total pressure ratio, flow rate, and other performance parameters corresponding to each surge critical point are extracted one by one. All extracted data are integrated to form a surge critical point dataset. Next, the surge critical point dataset is filtered to identify and remove abnormal critical points, retaining only all valid surge critical points that meet the convergence requirements and conform to the surge critical state. Finally, linear fitting or interpolation is used to connect the filtered valid surge critical points from all solution cycles in the order of operating conditions, and a continuous compressor surge boundary is generated through data fitting.

[0047] It should be noted that after the Pre-MOEA algorithm completes all iterative calculations for all solution cycles and obtains the surge critical points at each speed, and before connecting and fitting the surge critical points to form the surge boundary, there is a post-processing optimization process. The specific implementation process is as follows: based on the speed, the rate of change of the surge critical point is estimated, and the calculation interval is divided according to the rate of change. The surge critical points at both ends of each interval are calculated separately, and the middle point is obtained by fitting the change trajectory of the previous interval. If the actual calculated value at the endpoint of the interval deviates too much from the fitted predicted value, it is determined that an abnormal calculation result has occurred. At this time, a middle point is added in the interval and the calculation is recalculated to identify and recalculate abnormal calculation results, ensuring that the surge boundary formed by subsequent fitting is accurate and smooth.

[0048] S104. Calculate the compressor margin based on the surge boundary.

[0049] The compressor margin includes overall margin and pressure ratio margin.

[0050] Compressor margin is a key aerodynamic indicator used to measure the safe range within which a compressor can still operate stably without surge after deviating from its design point, and it directly determines the compressor's safe operating boundary. The compressor margin calculated in this application includes comprehensive margin and pressure ratio margin. The comprehensive margin is based on flow rate and reflects the compressor's flow rate safety margin between the near-surge point and the design point; the pressure ratio margin is based on the total pressure ratio and reflects the compressor's pressure boosting capability safety margin between the near-surge point and the design point.

[0051] In specific implementation, the design point of the compressor is determined, and the design total pressure ratio and design flow rate of the design point are extracted; based on the surge boundary, the near-surge point corresponding to the design point is determined on the isobaric ratio or isospeed line using the Pre-MOEA algorithm, and the total pressure ratio and flow rate corresponding to the near-surge point are extracted; the ratio of the total pressure ratio to the flow rate corresponding to the near-surge point is calculated as the first quotient; the ratio of the design total pressure ratio to the design flow rate is calculated as the second quotient; the ratio of the first quotient to the second quotient is calculated, and the difference between the ratio and 1 is converted into a percentage to obtain the overall margin.

[0052] Specifically, the design point of a compressor refers to its rated operating point under design conditions. It is the standard operating state of the compressor, predetermined during its research and development, that meets indicators such as rated flow rate, pressure ratio, and efficiency. In this application, the design total pressure ratio and design flow rate corresponding to the compressor's design point are the benchmark parameters for the compressor's stable, safe, and efficient operation, and also the comparison benchmark for subsequent calculation margins.

[0053] Specifically, firstly, the compressor design point is determined, and the corresponding design total pressure ratio and design flow rate parameters are retrieved and extracted. Next, based on the fitted surge boundary, the Pre-MOEA algorithm is used to locate and determine the near-surge point corresponding to the compressor design point on the isobaric ratio line or isospeed line, and the corresponding total pressure ratio and flow rate parameters are extracted. Then, the ratio of the near-surge point total pressure ratio to the near-surge point flow rate is calculated, and this result is recorded as the first quotient. Simultaneously, the ratio of the design total pressure ratio to the design flow rate at the design point is calculated, and this result is recorded as the second quotient. Next, the ratio of the first quotient to the second quotient is calculated, and 1 is subtracted from this ratio. The difference is then converted into a percentage to obtain the compressor's overall margin.

[0054] For example, in one embodiment, the calculation process of the comprehensive margin can be expressed as follows:

[0055] ;

[0056] in, For overall margin; This represents the total pressure ratio corresponding to the near-asthma point; To calculate the flow rate corresponding to the near-surging point; To design the overall pressure ratio; For designing traffic flow.

[0057] Optionally, the compressor margin is calculated based on the surge boundary, including: calculating the ratio of the total pressure ratio corresponding to the near surge point to the design total pressure ratio; subtracting the ratio from 1 and converting it to a percentage to obtain the pressure ratio margin.

[0058] Specifically, first, the design point of the compressor is determined, and the design total pressure ratio corresponding to the design point is extracted; based on the obtained surge boundary, the near-surge point corresponding to the design point is determined on the isobaric ratio line or isospeed line, and the total pressure ratio corresponding to the near-surge point is extracted; then, the ratio between the total pressure ratio of the near-surge point and the design total pressure ratio of the design point is calculated; the difference between this ratio and 1 is calculated, and the difference is converted into a percentage form to finally obtain the compressor pressure ratio margin.

[0059] For example, in one embodiment, the calculation process for the pressure margin can be expressed as follows:

[0060] ;

[0061] in, For pressure ratio margin; This represents the total pressure ratio corresponding to the near-asthma point; To design the overall pressure ratio.

[0062] Optionally, after calculating the compressor margin based on the surge boundary, the method further includes: encapsulating the compressor margin value into a fitness value format that can be recognized by the aerodynamic optimization algorithm, and substituting the margin value into preset optimization constraints to generate constraint satisfaction parameters; inputting the encapsulated fitness value and constraint satisfaction parameters into the objective function and constraint condition module of the aerodynamic optimization algorithm, and having the aerodynamic optimization algorithm automatically generate next-generation compressor geometry parameters based on the margin calculation results, and using the next-generation compressor geometry parameters as a new task to be calculated; the margin calculation results include the compressor margin value and the corresponding compressor geometry parameters.

[0063] Specifically, after calculating the compressor margin based on the surge boundary, subsequent optimization steps are executed. First, the calculated compressor margin value is formatted and encapsulated to a fitness value format that the aerodynamic optimization algorithm can directly recognize and call. Simultaneously, this margin value is substituted into pre-set optimization constraints to generate corresponding constraint satisfaction parameters. Then, the encapsulated fitness value and constraint satisfaction parameters are input into the objective function and constraint module of the aerodynamic optimization algorithm. The margin calculation results used for this input include the compressor margin value and the corresponding compressor geometry parameters. Upon receiving the relevant parameters, the aerodynamic optimization algorithm automatically performs iterative optimization calculations based on the margin calculation results to generate next-generation compressor geometry parameters. Finally, the generated next-generation compressor geometry parameters are used as a new task to be calculated, providing input data for the next round of surge boundary search and margin calculation, completing the entire closed-loop optimization process.

[0064] The method provided in this embodiment fully considers the automation requirements of the compressor aerodynamic optimization process and the core objective of converting margin from verification parameters to optimization parameters. By reserving optimization constraint interfaces and completing data format adaptation, it directly connects the technical link between margin calculation and aerodynamic optimization algorithm. This solves the inefficiency problem of manually organizing and inputting margin calculation results into optimization software in the traditional mode. At the same time, it solves the industry pain point that margin values ​​cannot be directly used as input for optimization algorithms and can only be verified afterward. It realizes automated closed-loop iteration of margin calculation and aerodynamic optimization, which greatly improves the efficiency and accuracy of compressor aerodynamic optimization design.

[0065] The method provided in this embodiment solves the surge boundary search objective function using the Pre-MOEA algorithm. Under a distributed computing architecture, it ensures that each computing node completes the evolutionary evaluation task of at least one evaluation node. Simultaneously, it combines an efficiency residual convergence judgment mechanism and an adaptive evolutionary wake-up rule. These three elements work together to achieve accurate, stable, and efficient acquisition of surge critical points within each solution cycle. On one hand, leveraging the multi-objective optimization and preference guidance capabilities of the Pre-MOEA algorithm, it directs the search for critical stability regions with low flow rates and high pressure ratios. Combined with distributed parallel computing, it achieves precise matching of computing power and tasks, avoiding resource idleness or overload and improving the overall solution speed. On the other hand, it relies on efficiency residuals to perform quantitative convergence judgment on the nodes, making node generation and critical state identification highly adaptive. This ensures that the iteration process is stable and does not diverge or produce invalid calculations, while accurately capturing the critical state between convergence and divergence, significantly improving the accuracy and reliability of surge critical point identification. Furthermore, it uses efficiency residuals to perform flow field convergence judgment. The system can use the standard deviation of continuously calculated data as a stability evaluation index to truly reflect the stability of the flow field at the end of the evaluation period. This avoids convergence misjudgment caused by instantaneous fluctuations, local anomalies, or computational instability, ensuring that all effective operating condition nodes entering the optimal screening have a stable, reliable, and credible flow field state. This improves the accuracy of the surge critical point from the source, thereby ensuring that the surge boundary obtained by subsequent fitting truly fits the actual stable operating limit of the compressor. Ultimately, it achieves the effects of high margin calculation accuracy, fast iteration speed, strong robustness, and high distributed resource utilization, providing stable, efficient, and automated margin calculation support for compressor aerodynamic design and optimization.

[0066] Corresponding to the aforementioned embodiment of a distributed compressor margin calculation method, this application also provides an embodiment of a distributed compressor margin calculation device.

[0067] Figure 2 This is a schematic diagram of the distributed compressor margin calculation device provided in Embodiment 2 of this application. Please refer to... Figure 2 The apparatus provided in this embodiment includes a determining module 210, a solving module 220, and a calculation module 230;

[0068] The determining module 210 is used to determine the search objective function for the surge boundary of the compressor;

[0069] The solution module 220 is used to solve the surge boundary search objective function based on the Pre-MOEA algorithm. In each solution cycle, the working condition nodes from the previous cycle are evolved to determine the number of evolved working condition nodes to be evaluated. A matching number of computing nodes are activated based on this number to complete the evolution evaluation of all the working condition nodes to be evaluated. The efficiency residual of each evolved working condition node to be evaluated is calculated, and the convergence of each working condition node is determined based on the efficiency residual. Converged valid working condition nodes are selected. From the valid working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computing node completes the evolution evaluation task of at least one working condition node to be evaluated.

[0070] The calculation module 230 is used to connect the surge critical points of all solution cycles and calculate the surge boundary;

[0071] The calculation module 230 is also used to calculate the compressor margin based on the surge boundary.

[0072] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0073] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0074] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0075] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for calculating the margin of a distributed compressor, characterized in that, The method includes: Determine the objective function for the compressor surge boundary search; The surge boundary search objective function is solved based on the Pre-MOEA algorithm. In each solution cycle, the working condition nodes from the previous cycle are evolved to determine the number of evolved working condition nodes to be evaluated. A matching number of computational nodes are activated based on this number to complete the evolution evaluation of all working condition nodes to be evaluated. The efficiency residual of each evolved working condition node is calculated, and convergence of each working condition node is determined based on the efficiency residual. Converged and effective working condition nodes are selected. From these effective working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computational node completes the evolution evaluation task of at least one working condition node to be evaluated. Connect the surge critical points of all solution cycles and calculate the surge boundary; Calculate the compressor margin based on the surge boundary; Calculate the efficiency residual for each node in the evaluation condition after evolution, including: Read the historical computational data corresponding to each post-evolutionary node to be evaluated. Extract the historical efficiency calculation data of the last preset number of steps in the evolutionary evaluation process of the target working condition node to be evaluated, calculate the standard deviation of the historical efficiency calculation data, and use the standard deviation as the efficiency residual of the target working condition node to be evaluated. In each solution cycle, the load case nodes from the previous cycle are evolved to determine the number of load case nodes to be evaluated after the evolution, including: Based on the compressor back pressure and performance parameters corresponding to the effective operating condition nodes that converged in the previous cycle, the operating condition nodes are evolved. If the current operating condition node is determined to be converged after evolution evaluation, the compressor back pressure is increased according to the preset back pressure change rate to generate the next operating condition node to be evaluated. If the current operating condition node is determined to be divergent after evolution evaluation, the current compressor back pressure is reduced and the back pressure adjustment step size is decreased, and the next operating condition node to be evaluated is regenerated. The process involves iteratively executing steps such as operating condition node evolution, evolution evaluation, convergence judgment, divergence judgment, and back pressure adjustment until a critical operating condition node is obtained. This critical operating condition node is then identified as the surge critical point, and the number of all operating condition nodes to be evaluated within this solution cycle is counted. If the convergence amplitude of the critical operating condition node is less than a preset value, the back pressure of the critical operating condition node is increased by the minimum step size, causing the critical operating condition node to diverge.

2. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, Calculate the efficiency residual for each node to be evaluated after evolution, determine whether each node has converged based on the efficiency residual, and select converged valid nodes, including: The efficiency residual is compared with a preset convergence threshold. If the efficiency residual is less than the convergence threshold, the target evaluation node is determined to have converged. If the efficiency residual is not less than the convergence threshold, the target evaluation node is determined to have not converged, and an idle computing node is awakened to re-complete the evolution evaluation of the target evaluation node.

3. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, Awaken the corresponding number of computing nodes based on the stated quantity, and complete the evolutionary evaluation of all nodes in the evaluated operating conditions, including: Using the compressor design point as the initial benchmark, the compressor back pressure adjustment task corresponding to the operating condition node to be evaluated is encapsulated as a task to be calculated and pushed to the shared queue task pool. Local computing nodes and remote computing nodes autonomously obtain tasks to be computed from the shared queue task pool, and execute computing tasks using the Threading parallel process mode. A single thread executes the evolution evaluation task of a node under evaluation and calls multi-core parallelism to complete the computation. After each computing node completes the evolutionary evaluation, it sends the calculation results back to the shared node for unified storage, and counts and determines the number of evolutionary evaluation tasks completed within this solution cycle.

4. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, The compressor margin includes a comprehensive margin and a pressure ratio margin. The compressor margin is calculated based on the surge boundary, including: Determine the design point of the compressor, and extract the design total pressure ratio and design flow rate of the design point; Based on the surge boundary, the pre-surge point corresponding to the design point is determined on the isobaric ratio or isospeed line using the Pre-MOEA algorithm, and the total pressure ratio and flow rate corresponding to the pre-surge point are extracted; Calculate the ratio of total pressure to flow rate corresponding to the near-surge point, and use it as the first quotient; Calculate the ratio of the total design pressure ratio to the design flow rate, and use it as the second quotient. Calculate the ratio of the first quotient to the second quotient, subtract 1 from the ratio, and convert it to a percentage to obtain the overall margin.

5. The method for calculating the margin of a distributed compressor according to claim 4, characterized in that, The compressor margin is calculated based on the surge boundary, including: Calculate the ratio of the total pressure ratio corresponding to the near-breathing point to the design total pressure ratio; The pressure margin is obtained by subtracting the ratio from 1 and converting it to a percentage.

6. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, Connect the surge critical points of all solution cycles and calculate the surge boundary, including: Record the surge critical points determined in each solution cycle, extract the compressor back pressure and performance parameters corresponding to each surge critical point, and form a surge critical point dataset; The surge critical point dataset is filtered to remove abnormal critical points and retain all convergent surge critical points that meet the critical state. By using linear fitting or interpolation methods, the surge critical points of all the solution cycles after screening are connected sequentially to form a continuous compressor surge boundary.

7. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, Determine the objective function for the compressor surge boundary search, including: The optimization objective of the surge boundary search objective function is determined; the optimization objective includes locating the compressor surge critical point; Determine the input variables for the surge boundary search objective function; the input variables include the compressor's back pressure, total pressure ratio, and flow rate before and after the evolution of the operating condition node in each solution cycle; Optimization constraints are set for the objective function of surge boundary search; the optimization constraints include convergence criteria, back pressure adjustment step size range, and physical quantity conservation requirements during the evolution process of operating nodes. Define the output variable of the surge boundary search objective function; the output variable is the surge critical point in each solution cycle. Based on the optimization objective, input variables, optimization constraints, and output variables, a surge boundary search objective function is constructed.

8. The method for calculating the margin of a distributed compressor according to claim 1, characterized in that, After calculating the compressor margin based on the surge boundary, the method further includes: The compressor margin value is encapsulated into a fitness value format that can be recognized by the aerodynamic optimization algorithm, and the margin value is substituted into the preset optimization constraints to generate constraint satisfaction parameters. The encapsulated fitness value and constraint satisfaction parameters are input into the objective function and constraint condition module of the aerodynamic optimization algorithm. The aerodynamic optimization algorithm automatically generates the next-generation compressor geometry parameters based on the margin calculation results, and uses the next-generation compressor geometry parameters as a new task to be calculated. The margin calculation results include the compressor margin value and the corresponding compressor geometry parameters.

9. A distributed compressor margin calculation device, characterized in that, The device includes a determination module, a solution module, and a calculation module; The determining module is used to determine the objective function for the surge boundary search of the compressor; The solution module is used to solve the surge boundary search objective function based on the Pre-MOEA algorithm. In each solution cycle, the working condition nodes from the previous cycle are evolved to determine the number of evolved working condition nodes to be evaluated. A matching number of computing nodes are activated based on this number to complete the evolution evaluation of all the working condition nodes to be evaluated. The efficiency residual of each evolved working condition node to be evaluated is calculated, and the convergence of each working condition node is determined based on the efficiency residual. Converged and effective working condition nodes are selected. From the effective working condition nodes, the top N evolved working condition nodes with the optimal objective function are selected, and each computing node completes the evolution evaluation task of at least one working condition node to be evaluated. The calculation module is used to connect the surge critical points of all solution cycles and calculate the surge boundary; The calculation module is also used to calculate the compressor margin based on the surge boundary; Calculate the efficiency residual for each node in the evaluation condition after evolution, including: Read the historical computational data corresponding to each post-evolutionary node to be evaluated. Extract the historical efficiency calculation data of the last preset number of steps in the evolutionary evaluation process of the target working condition node to be evaluated, calculate the standard deviation of the historical efficiency calculation data, and use the standard deviation as the efficiency residual of the target working condition node to be evaluated. In each solution cycle, the load case nodes from the previous cycle are evolved to determine the number of load case nodes to be evaluated after the evolution, including: Based on the compressor back pressure and performance parameters corresponding to the effective operating condition nodes that converged in the previous cycle, the operating condition nodes are evolved. If the current operating condition node is determined to be converged after evolution evaluation, the compressor back pressure is increased according to the preset back pressure change rate to generate the next operating condition node to be evaluated. If the current operating condition node is determined to be divergent after evolution evaluation, the current compressor back pressure is reduced and the back pressure adjustment step size is decreased, and the next operating condition node to be evaluated is regenerated. The process involves iteratively executing steps such as operating condition node evolution, evolution evaluation, convergence judgment, divergence judgment, and back pressure adjustment until a critical operating condition node is obtained. This critical operating condition node is then identified as the surge critical point, and the number of all operating condition nodes to be evaluated within this solution cycle is counted. If the convergence amplitude of the critical operating condition node is less than a preset value, the back pressure of the critical operating condition node is increased by the minimum step size, causing the critical operating condition node to diverge.

Images