Cooling fan performance test method and system

Abstract

The invention discloses a cooling fan performance test method and system, and the method comprises the steps: obtaining a multi-working condition test association diagram structure of a performance evaluation system, the structure comprising a fan entity, a working condition entity and a first test response association reflecting a technology coupling effect; generating a feature vector based on a sector coupling topology corresponding to the first cooling fan, the topology including a first fan entity and an associated working condition entity to quantify fan performance characterization data; and determining an adaptive performance reference value by calculating the matching degree of the feature vector and an alternative performance reference value feature vector. According to the invention, through integration of a multi-working-condition coupling relation, accurate characterization of the performance of the fan under a complex working condition is realized, the adaptability of a test result and an actual application scene is improved, and an efficient solution is provided for model selection and performance optimization of the cooling fan.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial control, in particular to a heat dissipation fan performance test method and system. BACKGROUND

[0002] As a key component of industrial equipment, such as mold cooling system, precision instrument heat dissipation module, the performance of heat dissipation fan directly affects the stability and service life of the equipment. The traditional heat dissipation fan performance test method is usually based on single or fixed working condition, such as air volume and pressure test under rated speed, but in actual application, the fan needs to work in a dynamic change environment with multiple working conditions, and there is a complex process coupling effect between different working conditions. The existing test method does not fully consider the coupling relationship of multiple working conditions, resulting in a large deviation between the test results and the actual running performance, which makes it difficult to accurately evaluate the adaptability of the fan in complex scenarios, and the problems of improper selection or low test efficiency are prone to occur. SUMMARY

[0003] The purpose of the present application is to provide a heat dissipation fan performance test method and system.

[0004] In a first aspect, the present application provides a heat dissipation fan performance test method, which comprises:

[0005] Obtain the multi-working condition test association graph structure required by the performance evaluation system; wherein the performance evaluation system is used to provide test working condition adaptation service for heat dissipation fan, and the multi-working condition test association graph structure comprises: at least one fan entity corresponding to each heat dissipation fan, at least one working condition entity corresponding to each test working condition, and a first test response association between the fan entity and the working condition entity, the first test response association is used to reflect the process coupling effect between the heat dissipation fan corresponding to the fan entity and the test working condition corresponding to the working condition entity;

[0006] According to the sector coupling topology corresponding to the first heat dissipation fan in the multi-working condition test association graph structure, generate the feature vector corresponding to the first heat dissipation fan; wherein the sector coupling topology corresponding to the first heat dissipation fan comprises: the first fan entity corresponding to the first heat dissipation fan, and at least one working condition entity having a first test response association with the first fan entity; the feature vector corresponding to the first heat dissipation fan is used to reflect the performance characterization data of the first heat dissipation fan;

[0007] According to the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to each of the at least one alternative performance reference value, determine at least one performance reference value matched with the first heat dissipation fan; wherein the feature vector corresponding to the performance reference value is used to reflect the performance characterization result of the performance reference value.

[0008] In a possible implementation, the generating the feature vector corresponding to the first cooling fan according to the sector coupling topology corresponding to the first cooling fan in the multi-working condition test correlation graph structure comprises:

[0009] For each of the at least one working condition entity having the first test response correlation with the first fan entity, determining a time sequence position of the working condition entity in the working condition time sequence dynamic flow according to a data acquisition entity recorded in a data record of the first test response correlation between the working condition entity and the first fan entity;

[0010] sequencing the at least one working condition entity having the first test response correlation with the first fan entity according to the respective time sequence positions, to generate the working condition time sequence dynamic flow;

[0011] determining the respective parameter acquisition data of the at least one working condition entity included in the working condition time sequence dynamic flow; wherein the parameter acquisition data of the working condition entity is used to indicate the test working condition corresponding to the working condition entity;

[0012] inputting the respective parameter acquisition data of the at least one working condition entity into a cooling fan side feature extraction module; wherein the cooling fan side feature extraction module is a feature analysis model used to generate the feature vector corresponding to the first cooling fan;

[0013] performing feature extraction processing on the respective parameter acquisition data of the at least one working condition entity by a working condition coding component in the cooling fan side feature extraction module, to obtain a dynamic working condition spectrum of the working condition time sequence dynamic flow;

[0014] performing feature extraction processing on the dynamic working condition spectrum of the working condition time sequence dynamic flow by a feature transformation component in the cooling fan side feature extraction module, to obtain the feature vector corresponding to the working condition time sequence dynamic flow; wherein the feature vector corresponding to the working condition time sequence dynamic flow is used to reflect the performance characterization data of at least one test working condition of the process coupling effect of the first cooling fan;

[0015] generating the feature vector corresponding to the first cooling fan according to the feature vector corresponding to the working condition time sequence dynamic flow.

[0016] In a possible implementation, the sector-coupling topology further comprises: at least one working condition performance index entity associated with at least one working condition entity in the sector-coupling topology through a second test response association; wherein the at least one working condition entity in the sector-coupling topology refers to a working condition entity associated with the first fan entity through the first test response association, and the second test response association is used to reflect that a test working condition corresponding to the working condition entity has a working condition performance index corresponding to the working condition performance index entity; and the method further comprises:

[0017] For a first working condition performance index type in the at least one working condition performance index type, at least one first working condition performance index entity corresponding to each working condition entity in the working condition time sequence dynamic flow in the first working condition performance index type is determined according to the second test response association in the sector-coupling topology;

[0018] A working condition performance index set corresponding to the first working condition performance index type is generated according to the at least one first working condition performance index entity corresponding to each working condition entity in the working condition time sequence dynamic flow; and the working condition performance index set comprises a working condition performance index entity having a second test response association with at least one working condition entity in the working condition time sequence dynamic flow;

[0019] For a first working condition performance index type in the at least one working condition performance index type, parameter acquisition data corresponding to each working condition performance index entity included in a working condition performance index set corresponding to the first working condition performance index type is determined; and the parameter acquisition data corresponding to each working condition performance index entity is used to indicate a working condition performance index corresponding to the working condition performance index entity;

[0020] The parameter acquisition data corresponding to each working condition performance index entity included in the working condition performance index set corresponding to the first working condition performance index type is subjected to feature extraction processing by a working condition performance index feature mapping component corresponding to the first working condition performance index type in a heat dissipation fan side feature extraction module, to obtain a feature vector corresponding to the first working condition performance index type; and the heat dissipation fan side feature extraction module is used to generate a feature vector corresponding to the first heat dissipation fan;

[0021] The feature vector corresponding to each working condition performance index type is subjected to feature extraction by a working condition performance index feature aggregator in the heat dissipation fan side feature extraction module, to obtain a feature vector corresponding to the working condition performance index;

[0022] The generation of the feature vector corresponding to the first heat dissipation fan according to the feature vector corresponding to the working condition time sequence dynamic flow comprises:

[0023] The feature fusion component of the heat dissipation fan feature extraction module fuses the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the working condition performance index to obtain a first fused feature vector;

[0024] The feature generation component in the heat dissipation fan side feature extraction module extracts features from the first fused feature vector to obtain a feature vector corresponding to the first heat dissipation fan.

[0025] In a possible implementation, the sector-coupled topology further includes at least one heat dissipation fan characteristic entity having a third test response association with the first fan entity; wherein the third test response association is used to reflect that the heat dissipation fan corresponding to the fan entity has the heat dissipation fan characteristic corresponding to the heat dissipation fan characteristic entity; the method further includes:

[0026] According to the sector-coupled topology, at least one heat dissipation fan characteristic entity corresponding to a respective heat dissipation fan characteristic dimension is generated; wherein the heat dissipation fan characteristic set includes at least one working condition performance index entity having a third test response association with the first fan entity;

[0027] According to the set of heat dissipation fan characteristics corresponding to at least one respective heat dissipation fan characteristic dimension, a feature vector corresponding to the heat dissipation fan characteristic is generated;

[0028] The generation of the feature vector corresponding to the first heat dissipation fan according to the feature vector corresponding to the working condition time sequence dynamic flow includes:

[0029] According to the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the heat dissipation fan characteristic, a feature vector corresponding to the first heat dissipation fan is generated.

[0030] In a possible implementation, the generation of the feature vector corresponding to the heat dissipation fan characteristic according to the set of heat dissipation fan characteristics corresponding to at least one respective heat dissipation fan characteristic dimension includes:

[0031] For a first heat dissipation fan characteristic dimension in the at least one heat dissipation fan characteristic dimension, the set of heat dissipation fan characteristics corresponding to the first heat dissipation fan characteristic dimension includes respective parameter collection data of at least one heat dissipation fan characteristic entity; the respective parameter collection data of the heat dissipation fan characteristic entity is used to indicate the heat dissipation fan characteristic corresponding to the heat dissipation fan characteristic entity;

[0032] The identity of the at least one heat dissipation fan characteristic entity is subjected to feature extraction processing by a heat dissipation fan characteristic feature mapping component corresponding to the first heat dissipation fan characteristic dimension in the heat dissipation fan side feature extraction module, to obtain an embedding vector of the first heat dissipation fan characteristic dimension; wherein the heat dissipation fan side feature extraction module is configured to generate a feature vector corresponding to the first heat dissipation fan characteristic dimension.

[0033] The feature vectors corresponding to the at least one heat dissipation fan characteristic dimension are subjected to feature extraction by a heat dissipation fan characteristic feature aggregator in the heat dissipation fan side feature extraction module, to obtain a feature vector corresponding to the heat dissipation fan characteristic.

[0034] In a possible implementation, the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to at least one alternative performance benchmark value is used to determine at least one performance benchmark value matched by the first heat dissipation fan, including:

[0035] determining the feature vector corresponding to at least one alternative performance benchmark value;

[0036] for a first performance benchmark value in the at least one alternative performance benchmark value, calculating a feature space approximation degree between the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value, and determining a performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value;

[0037] from the at least one alternative performance benchmark value, selecting an alternative performance benchmark value with a performance matching degree not lower than a performance matching degree threshold as a performance benchmark value matched by the first heat dissipation fan.

[0038] In a possible implementation, the determination of the feature vector corresponding to at least one alternative performance benchmark value includes:

[0039] for a first performance benchmark value in the at least one alternative performance benchmark value, obtaining at least one reference working condition performance indicator corresponding to the first performance benchmark value; wherein the reference working condition performance indicator has performance equivalence with the first performance benchmark value;

[0040] determining the feature vector corresponding to the first performance benchmark value according to the first performance benchmark value and the at least one reference working condition performance indicator corresponding to the first performance benchmark value.

[0041] In a possible implementation, the determination of the feature vector corresponding to the first performance benchmark value according to the first performance benchmark value and the at least one reference working condition performance indicator corresponding to the first performance benchmark value includes:

[0042] The performance benchmark value feature mapping component of the working condition side feature extraction module extracts features from the parameter acquisition data corresponding to each of the at least one word in the first performance benchmark value, to obtain parameter representation corresponding to the first performance benchmark value; wherein the working condition side feature extraction module is configured to determine a feature vector corresponding to the first performance benchmark value;

[0043] The working condition parameter mapping unit of the working condition side feature extraction module extracts features from the parameter acquisition data corresponding to each of the at least one word included in the at least one working condition performance indicator, to obtain parameter representation corresponding to the working condition performance indicator;

[0044] The feature fusion component of the working condition side feature extraction module fuses the parameter representation corresponding to the first performance benchmark value and the parameter representation corresponding to the working condition performance indicator, to obtain performance fusion representation;

[0045] The feature generation component of the working condition side feature extraction module extracts features from the performance fusion representation, to determine a feature vector corresponding to the first performance benchmark value.

[0046] In a possible implementation, after determining the at least one performance benchmark value matched by the first heat dissipation fan according to the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to each of the at least one alternative performance benchmark value, the method further includes:

[0047] According to the at least one performance benchmark value matched by the first heat dissipation fan, determining at least one adaptive test working condition from a pool of test working conditions corresponding to the performance evaluation system;

[0048] Sending the adaptive test working condition to a monitoring system corresponding to the first heat dissipation fan.

[0049] In a second aspect, an embodiment of the present application provides a server system, including a server configured to execute the method of the first aspect.

[0050] Compared with the prior art, the application provides the following beneficial effects: the performance test method and system of the heat dissipation fan disclosed by the application are adopted, a multi-working condition test correlation graph structure of a performance evaluation system is obtained, the structure includes a fan entity, a working condition entity and a first test response correlation reflecting process coupling effect, a feature vector is generated based on a sector coupling topology corresponding to the first heat dissipation fan, the topology includes the first fan entity and the associated working condition entity, so as to quantize fan performance characterization data, and the matching degree of the feature vector and an alternative performance benchmark value feature vector is calculated to determine the adaptive performance benchmark value. The application realizes accurate characterization of the performance of the fan under complex working conditions by integrating the multi-working condition coupling relationship, improves the adaptability of the test result to the actual application scene, and provides an efficient solution for the selection and performance optimization of the heat dissipation fan. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0052] Figure 1 The flowchart of the heat dissipation fan performance test method provided by the embodiment of the application is shown in the figure.

[0053] Figure 2 The schematic diagram of the computer device provided by the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the application clearer, the following will combine the drawings in the embodiments of the application to clearly and completely describe the technical solutions in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, not all the embodiments. The components of the embodiments of the application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0055] The specific embodiments of the application will be described in detail below with reference to the drawings. Figure 1 The flowchart of the heat dissipation fan performance test method provided by the embodiment of the application is shown in the figure.

[0056] In step S201, a multi-working condition test association graph structure required to be called by a performance evaluation system is acquired; wherein the performance evaluation system is used to provide a test working condition adaptation service for a cooling fan, and the multi-working condition test association graph structure comprises: at least one fan entity corresponding to each cooling fan, at least one working condition entity corresponding to each test working condition, and a first test response association between the fan entity and the working condition entity, wherein the first test response association is used to reflect a process coupling effect between the cooling fan corresponding to the fan entity and the test working condition corresponding to the working condition entity;

[0057] In step S202, a feature vector corresponding to a first cooling fan is generated according to a sector coupling topology of the first cooling fan in the multi-working condition test association graph structure; wherein the sector coupling topology of the first cooling fan comprises: a first fan entity corresponding to the first cooling fan, and at least one working condition entity having a first test response association with the first fan entity; and the feature vector corresponding to the first cooling fan is used to reflect performance characterization data of the first cooling fan.

[0058] In step S203, at least one performance reference value matched by the first cooling fan is determined according to a performance matching degree between the feature vector corresponding to the first cooling fan and a feature vector corresponding to each of at least one alternative performance reference value; wherein the feature vector corresponding to the performance reference value is used to reflect a performance characterization result of the performance reference value.

[0059] In the embodiment of the application, in the industrial mold production scene, the performance of the heat dissipation module fan is directly related to the mold cooling efficiency and product forming quality. For example, the continuous high temperature of a large injection mold may cause the deformation of the plastic part, and the local overheating of a precision stamping die may aggravate the die wear. To ensure that the heat dissipation fan adapts to the cooling needs of different molds, the server as the execution subject needs to carry out test work systematically through the performance evaluation system, and the core process starts from obtaining a multi-condition test association graph structure. The structure is the basis for representing the process coupling relationship between the fan and the test condition, and includes a fan entity, a condition entity, and a first test response association for reflecting the process coupling effect of the two. Taking the MF-800A type heat dissipation fan (hereinafter referred to as the “first heat dissipation fan”) used in the production line as an example, which is mainly adapted to large injection molds, the server retrieves the corresponding multi-condition test association graph structure of the first heat dissipation fan from the graph database of the performance evaluation system: the fan entity is explicitly MF-800A type, and the condition entity covers the key test parameters of the mold cooling system, including environmental temperature conditions (35°C normal temperature, 45°C medium temperature, 55°C high temperature), air flow rate conditions (1.2 m / s low flow rate, 1.5 m / s medium flow rate, 1.8 m / s high flow rate), load power conditions (80W low load, 100W medium load, 120W high load), and mold cavity pressure conditions (5MPa, 8MPa, 10MPa, used to reflect the degree of heat accumulation inside the mold). The first test response association records the performance change rule of the fan under different conditions through specific data, for example, the heat dissipation efficiency of MF-800A under 35°C normal temperature condition decreases from 92% to 88% after 2 hours of continuous operation, the noise increases from 58dB to 62dB, and the power consumption fluctuates from 95W to 98W; while under the 55°C high temperature condition, the heat dissipation efficiency decreases more significantly from 85% to 78%, and the noise increases from 63dB to 68dB. These data together constitute a quantitative description of the process coupling effect between the fan and the condition, and the server obtains the above complete association relationship by calling the query interface of the graph database to form the multi-condition test association graph structure.

[0060] After obtaining the multi-condition test correlation graph structure, the server needs to generate a feature vector for the first cooling fan to reflect its performance characterization data. This process is based on the fan's corresponding sector coupling topology. The sector coupling topology specifically includes the MF-800A fan entity (i.e., the first fan entity) and all condition entity connected with the entity through the first test response association. Considering that the test of the cooling fan in industrial mold production needs to simulate the heat accumulation process of the mold from startup to full load operation, the server first needs to sort the condition entity according to the process test logic sequence to form a condition time sequence dynamic flow. For example, the test flow of MF-800A is: first test 1.2 m / s and 1.5 m / s air flow rate conditions at 35℃ constant temperature, then gradually increase the environment temperature to 45℃ and superimpose 100W load power condition, and finally test 120W load power and 8MPa cavity pressure conditions at 55℃ high temperature, thus the determined condition time sequence dynamic flow is "35℃ constant temperature + 1.2 m / s flow rate → 35℃ constant temperature + 1.5 m / s flow rate → 45℃ medium temperature + 100W load → 55℃ high temperature + 120W load + 8MPa cavity pressure". Then, the server obtains the parameter collection data of each condition entity by calling the data interface of the test equipment, including the measured values of environment temperature, air flow rate, load power, cavity pressure, etc. For example, at 35℃ constant temperature + 1.2 m / s flow rate condition, the environment temperature is measured as 35.2℃, the air flow rate is 1.18 m / s, the test duration is 120 minutes, and the initial cooling efficiency is 92.3%; at 55℃ high temperature + 120W load + 8MPa cavity pressure condition, the environment temperature is 54.9℃, the load power is 120.3W, the cavity pressure is 7.98MPa, the test duration is 180 minutes, and the initial cooling efficiency is 78.2%.

[0061] The server inputs these parameter collection data into the pre-trained cooling fan side feature extraction module to generate the feature vector of MF-800A. The module first processes the parameter collection data through the condition coding component: first, all parameters are normalized to the [0, 1] interval through the formula x_norm = (x-x_min) / (x_max-x_min) (e.g., environment temperature 35℃ corresponds to 0.35, 55℃ corresponds to 0.55, x_min / x_max is taken from historical data extreme values); then, the LSTM layer captures the dynamic change law in the condition sequence (e.g., the attenuation trend of cooling efficiency when the environment temperature rises from 35℃ to 55℃, the fluctuation feature of power consumption when the load power increases, etc.), outputting a time sequence feature matrix with a dimension of T×64 (T is the length of the condition time sequence dynamic flow); then, spatial feature fusion is performed through a 1×1 convolution kernel to obtain a dynamic condition spectrum.

[0062] The model architecture and training details are as follows: 1) Model architecture: LSTM-CNN hybrid architecture is adopted, including: input layer (receiving normalized operating conditions / performance parameters), operating condition encoding component (2-layer LSTM, hidden unit number 64, activation function tanh, processing time-dependent), dynamic operating condition spectrum generation layer (1x1 convolution kernel, number 32, step 1, padding 0, activation function ReLU, fusion spatial features), feature transformation component (PCA layer, cumulative variance contribution rate 95%, dimension reduction redundant features), output layer (fully connected layer, output 64-dimensional feature vector);

[0063] 2) Training data source: historical heat dissipation fan test database (containing 1200+ fan models with multi-condition time series data: 10 types of operating condition parameters such as ambient temperature and air flow rate, 8 types of performance data such as heat dissipation efficiency and noise), labeled as "performance meets standards / does not meet standards" (based on GB / T14880-2021 and other industrial standards);

[0064] 3) Training process: supervised learning, loss function is cross-entropy loss, optimizer is Adam (learning rate 1e-4), training rounds are 50 rounds, validation set proportion is 20%, early stopping strategy is adopted (validation set loss does not decrease for 5 consecutive rounds, then stop training);. Then, the feature transformation component reduces the dimension of the dynamic operating condition spectrum and extracts key features, through global analysis of time series data, multi-dimensional operating condition information is condensed into a fixed length vector, for example, by analyzing the heat dissipation efficiency decay rate, noise change value, power consumption fluctuation value and other core indicators under different operating conditions, a 32-length vector is generated, which contains the performance representation data of MF-800A fan under multi-condition, that is, the feature vector corresponding to the first heat dissipation fan.

[0065] After generating the feature vector of the first cooling fan, the server needs to determine the adaptive performance benchmark value by matching with the feature vector of the alternative performance benchmark value. The alternative performance benchmark value is derived from the standard performance indicators of industrial mold cooling systems, such as large injection mold requiring cooling fan to have a cooling efficiency of not less than 75% and a noise of not more than 70dB under the condition of 55℃ high temperature and 120W load, precision stamping mold requiring cooling efficiency of not less than 85% and noise of not more than 65dB under the condition of 45℃ medium temperature and 100W load, etc. Each benchmark value corresponds to a feature vector, which is generated by the working condition side feature extraction module from the associated reference working condition performance indicators (such as the above-mentioned cooling efficiency, noise threshold), and contains the performance requirements of the benchmark value for different working conditions. The server filters out the benchmark value with a matching degree not less than a set threshold by calculating the performance matching degree between the feature vector of MF-800A and the feature vector of each alternative benchmark value (such as by approximating the degree of the vector in the feature space, the higher the numerical value, the better the matching). For example, the cooling efficiency of MF-800A under the condition of 55℃ high temperature is 78.2% (higher than the threshold of 75%), and the noise is 68dB (lower than the threshold of 70dB), and the matching degree between its feature vector and the performance benchmark value vector of large injection mold reaches 0.92 (the threshold is set to 0.85), so it is determined that this benchmark value is the adaptive benchmark. Finally, the server selects the corresponding adaptive test working condition (such as 55℃ high temperature, 120W load, 8MPa cavity pressure, etc.) from the test working condition pool of the performance evaluation system according to the matched performance benchmark value, and sends these working conditions to the monitoring system corresponding to MF-800A fan, guiding the subsequent performance test and optimization work, and ensuring that the fan meets the mold cooling demand in actual production.

[0066] In the embodiment of the application, the generation of the feature vector corresponding to the first cooling fan according to the sector coupling topology of the first cooling fan in the multi-working condition test association graph structure can be implemented by the following examples.

[0067] According to at least one working condition entity associated with the first fan entity in the sector coupling topology and having a first test response, a working condition time sequence dynamic flow is generated;

[0068] A feature vector corresponding to the working condition time sequence dynamic flow is generated; wherein the feature vector corresponding to the working condition time sequence dynamic flow is used to reflect the performance characterization data of at least one test working condition associated with the process coupling effect of the first cooling fan;

[0069] According to the feature vector corresponding to the working condition time sequence dynamic flow, the feature vector corresponding to the first cooling fan is generated.

[0070] In the embodiment of the present application, in the industrial mold production scene, the server performs the feature vector generation step for the first cooling fan (MF-800A model, suitable for large injection molds). First, based on the sector coupling topology, the server determines the working condition time sequence dynamic flow. The sector coupling topology includes the MF-800A fan entity and the working condition condition entity associated through the first test response, specifically: 35°C normal temperature + 1.2m / s flow rate working condition, 35°C normal temperature + 1.5m / s flow rate working condition, 45°C medium temperature + 100W load working condition, 55°C high temperature + 120W load + 8MPa cavity pressure working condition. The server sorts these working condition condition entities according to the process logic of the mold from startup to full load operation as “35°C normal temperature + 1.2m / s flow rate→35°C normal temperature + 1.5m / s flow rate→45°C medium temperature + 100W load→55°C high temperature + 120W load + 8MPa cavity pressure”, forming the working condition time sequence dynamic flow.

[0071] Then, the server obtains the parameter collection data of each working condition through the test equipment interface: 35°C normal temperature + 1.2m / s flow rate working condition, the measured environmental temperature is 35.2°C, the flow rate is 1.18m / s, the test duration is 120min, and the initial cooling efficiency is 92.3%; 35°C normal temperature + 1.5m / s flow rate working condition, the measured temperature is 34.8°C, the flow rate is 1.49m / s, the duration is 120min, and the efficiency is 93.1%; 45°C medium temperature + 100W load working condition, the measured temperature is 45.1°C, the load is 99.7W, the duration is 150min, and the efficiency is 85.6%; 55°C high temperature + 120W load + 8MPa cavity pressure working condition, the measured temperature is 54.9°C, the load is 120.3W, the pressure is 7.98MPa, the duration is 180min, and the efficiency is 78.2%. The server inputs these data into the cooling fan side feature extraction module, which first normalizes the parameters (such as temperature 35°C→0.35, 55°C→0.55), captures the working condition sequence dynamic law through the time sequence feature extraction (such as the decay trend of the cooling efficiency from 92.3% to 78.2% when the temperature rises), and generates a dynamic working condition spectrum; then, through the feature transformation component, the multi-dimensional working condition information is condensed into a vector of length 32, i.e. the feature vector corresponding to the working condition time sequence dynamic flow, which includes the performance characterization data such as the cooling efficiency decay rate of MF-800A under multiple working conditions and the noise change amount.

[0072] Finally, the server directly uses the feature vector corresponding to the working condition time sequence dynamic flow as the feature vector of the first cooling fan (MF-800A), because it has completely reflected the process coupling effect of the fan and each test working condition, and can be used for subsequent matching with the performance benchmark value.

[0073] In the embodiments of the present application, the generation of the working condition time sequence dynamic flow according to at least one working condition entity associated with the first fan entity in the sector coupling topology with the first test response can be implemented by the following examples.

[0074] For each of the at least one working condition entity associated with the first fan entity with the first test response, the data acquisition entity stored in the data record of the first test response association between the working condition entity and the first fan entity is determined to determine the time sequence position of the working condition entity in the working condition time sequence dynamic flow.

[0075] The at least one working condition entity associated with the first fan entity with the first test response is sorted according to the corresponding time sequence position to generate the working condition time sequence dynamic flow.

[0076] In the embodiments of the present application, for example, in the industrial mold production scene, when the server generates the working condition time sequence dynamic flow for the first cooling fan (MF-800A model, suitable for large injection mold), the time sequence position of each working condition entity needs to be determined first. The working condition entities associated with the first fan entity (MF-800A) through the first test response association in the sector coupling topology of the fan include: "35℃ normal temperature + 1.2m / s air flow rate", "35℃ normal temperature + 1.5m / s air flow rate", "45℃ medium temperature + 100W load power", and "55℃ high temperature + 120W load power + 8MPa cavity pressure". In the data record of the first test response association between each working condition entity and MF-800A, the data acquisition entity stores the test execution process order identifier (such as "test stage code") for reflecting the logical flow of the mold from preheating to full load operation. For example, the test stage code in the data acquisition entity of the "35℃ normal temperature + 1.2m / s air flow rate" working condition is "ST01" (corresponding to the basic cooling working condition of the mold start-up preheating stage); the code of "35℃ normal temperature + 1.5m / s air flow rate" is "ST02" (the flow rate optimization working condition of the preheating stage); the code of "45℃ medium temperature + 100W load power" is "ST03" (the typical working condition of the mold half-load production stage); and the code of "55℃ high temperature + 120W load power + 8MPa cavity pressure" is "ST04" (the limit working condition of the mold full-load continuous production stage). The server determines the time sequence positions of the working condition entities in the working condition time sequence dynamic flow as 1, 2, 3, and 4 in turn by reading the test stage codes in the data acquisition entities of the working condition entities.

[0077] Subsequently, the server sorts the above working condition entities in the order of time sequence positions 1 to 4, generates a working condition time sequence dynamic flow: "35℃ normal temperature + 1.2m / s air flow rate→ 35℃ normal temperature + 1.5m / s air flow rate→ 45℃ medium temperature + 100W load power→ 55℃ high temperature + 120W load power + 8MPa cavity pressure". The dynamic flow completely reflects the working condition change sequence that the MF-800A fan needs to experience in the actual production process of the industrial mold, and provides a time-sequenced working condition data basis for subsequent feature vector generation.

[0078] In the embodiments of the present application, the generation of the feature vector corresponding to the working condition time sequence dynamic flow can be implemented by the following examples.

[0079] Determine the parameter acquisition data corresponding to each of the at least one working condition entity included in the working condition time sequence dynamic flow; wherein the parameter acquisition data corresponding to the working condition entity is used to indicate the test working condition corresponding to the working condition entity;

[0080] Input the parameter acquisition data corresponding to each of the at least one working condition entity into the heat dissipation fan side feature extraction module; wherein the heat dissipation fan side feature extraction module is a feature analysis model for generating the feature vector corresponding to the first heat dissipation fan;

[0081] Through the working condition coding component of the heat dissipation fan side feature extraction module, the parameter acquisition data corresponding to each of the at least one working condition entity is subjected to feature extraction processing to obtain a dynamic working condition spectrum of the working condition time sequence dynamic flow;

[0082] Through the feature transformation component in the heat dissipation fan side feature extraction module, the dynamic working condition spectrum of the working condition time sequence dynamic flow is subjected to feature extraction processing to obtain the feature vector corresponding to the working condition time sequence dynamic flow.

[0083] In an embodiment of the present application, in an industrial mold production scenario, the server generates a feature vector corresponding to the working condition time sequence dynamic flow of the first cooling fan (MF-800A model, suitable for large injection molds), first determines the parameter collection data of each working condition entity in the dynamic flow. The dynamic flow contains four working conditions: "35°C normal temperature + 1.2m / s air flow rate" "35°C normal temperature + 1.5m / s air flow rate" "45°C medium temperature + 100W load power" "55°C high temperature + 120W load power + 8MPa cavity pressure". The server obtains the measured parameters of each working condition by calling the industrial data acquisition interface of the test equipment (such as a real-time database based on the OPCUA protocol): in the "35°C normal temperature + 1.2m / s air flow rate" working condition, the environment temperature is measured as 35.2°C (set value 35°C), the air flow rate is 1.18m / s (set value 1.2m / s), the test duration is 120 minutes, and the initial cooling efficiency is 92.3%; in the "35°C normal temperature + 1.5m / s air flow rate" working condition, the environment temperature is 34.8°C, the flow rate is 1.49m / s, the duration is 120 minutes, and the efficiency is 93.1%; in the "45°C medium temperature + 100W load power" working condition, the environment temperature is 45.1°C, the load power is 99.7W (set value 100W), the duration is 150 minutes, and the efficiency is 85.6%; in the "55°C high temperature + 120W load power + 8MPa cavity pressure" working condition, the environment temperature is 54.9°C, the load power is 120.3W, the cavity pressure is 7.98MPa (set value 8MPa), the duration is 180 minutes, and the efficiency is 78.2%. These parameter collection data completely record the test conditions and initial performance of each working condition, which are used to accurately indicate the actual running state of MF-800A in different mold production stages.

[0084] Subsequently, the server inputs the above parameter collection data into the pre-trained "cooling fan side feature extraction module", which is a feature analysis model optimized based on industrial time sequence data, and is specially used for processing fan performance feature extraction under multi-working condition coupling. The module first processes the parameter collection data through a "working condition coding component": first, all parameters are normalized to a unified interval (for example, environment temperature 35°C corresponds to 0.35, 55°C corresponds to 0.55, flow rate 1.2m / s corresponds to 0.4, 1.8m / s corresponds to 0.6), to eliminate dimensional differences; then, through time sequence feature extraction technology, the dynamic change law in the working condition sequence is captured, for example, in the process of environment temperature rising from 35°C to 55°C, the cooling efficiency gradually decreases from 92.3% to 78.2%, and when the load power increases from 100W to 120W, the power consumption fluctuates from 99.7W to 120.3W. After this processing, the working condition coding component outputs an intermediate feature matrix reflecting the time sequence dynamic law, i.e. a dynamic working condition spectrum;

[0085] 1) Definition: Dynamic working condition spectrum is a two-dimensional matrix representing the dynamic coupling characteristics of the working condition parameters and performance parameters of a cooling fan under time-series working conditions, with dimensions of TxF (T is the number of time-series steps, corresponding to the length of the working condition time-series flow; F is the feature dimension, including normalized working condition parameters, performance parameters, and coupling characteristics);

[0086] 2) Calculation steps: a. Parameter normalization: normalize the working condition parameters such as ambient temperature and air flow rate, and the performance parameters such as cooling efficiency and noise to the [0, 1] interval through x_norm = (x - x_min) / (x_max - x_min);

[0087] b. Time-series window division: divide the time-series data with a length of N into T overlapping windows (window size 10, step size 5, ensuring time-series continuity);

[0088] c. Local feature extraction: calculate 16-dimensional statistical features (including parameter changes and parameter coupling relationships) such as mean, variance, slope, and extreme value difference for each window;

[0089] d. Matrix construction: arrange the 16-dimensional features of the T windows in time-series order to form a dynamic working condition spectrum matrix of T x 16;

[0090] In this embodiment, the dynamic working condition spectrum has dimensions of 64 x 4 (T = 64 time-series windows, F = 4 core features: temperature-efficiency coupling, load-power consumption coupling, flow rate-noise coupling, and pressure-steady state time coupling), and the matrix elements include key dynamic features such as temperature-efficiency attenuation coefficient and load-power consumption sensitivity.

[0091] Next, the "feature transformation component" of the module performs dimensionality reduction and key feature extraction on the dynamic working condition spectrum: through a 1 x 1 convolution kernel (number 32, step size 1, padding 0, activation function ReLU), the 64 x 4 dynamic working condition spectrum is spatially fused, reducing the number of feature channels from 64 to 32 while retaining the core dynamic information; then, the global average pooling operation is performed on the time-series features of each channel to obtain a vector with a length of 32. This vector contains key performance characterization data of the MF-800A in the working condition time-series flow, such as the average attenuation rate of the cooling efficiency when the ambient temperature increases by 10°C (7.7% attenuation from 35°C to 45°C, 9.0% attenuation from 45°C to 55°C), the noise change when the load power increases by 20W (63dB at 100W, 68dB at 120W, with an increase of 5dB), and the cooling efficiency improvement at different flow rates (92.3% at 1.2m / s, 93.1% at 1.5m / s, with an increase of 0.8%). Finally, the 32-dimensional vector output by the feature transformation component is the feature vector corresponding to the working condition time-series flow, which completely quantifies the process coupling effect between the MF-800A and the multiple test working conditions.

[0092] In the embodiment of the present application, the sector coupling topology further comprises at least one working condition performance index entity connected with at least one working condition entity in the sector coupling topology through a second test response association; wherein the at least one working condition entity in the sector coupling topology refers to a working condition entity connected with the first fan entity through the first test response association, and the second test response association is used to reflect that the test working condition corresponding to the working condition entity has the working condition performance index corresponding to the working condition performance index entity; the embodiment of the present application further provides the following implementation manners.

[0093] According to the sector coupling topology, a working condition performance index set corresponding to each of at least one working condition performance index type is generated; wherein the working condition performance index set comprises a working condition performance index entity having a second test response association with at least one working condition entity in the working condition time sequence dynamic flow;

[0094] According to the working condition performance index set corresponding to each of the at least one working condition performance index type, a feature vector corresponding to the working condition performance index is generated;

[0095] The generating of the feature vector corresponding to the first cooling fan according to the feature vector corresponding to the working condition time sequence dynamic flow comprises:

[0096] The generating of the feature vector corresponding to the first cooling fan according to the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the working condition performance index.

[0097] In the embodiment of the application, in the industrial mold production scenario, the sector-coupling topology of the first cooling fan (MF-800A model, suitable for large injection molds) also contains a working condition performance index entity associated with each working condition entity through a second test response association, where the second test response association is used to reflect the specific performance index that the test working condition has, for example, the "35℃ normal temperature + 1.2m / s flow rate" working condition through the association points to the index entity recording the heat dissipation efficiency, noise, and power consumption data. The association relationship also contains a data acquisition timestamp (such as the performance sampling points of the 30th minute, the 60th minute, and the 120th minute). The server first determines the working condition performance index type to be heat dissipation efficiency type, noise type, and power consumption type according to the core evaluation dimensions of the industrial mold cooling system, and then extracts the corresponding working condition performance index set of each type from the sector-coupling topology through the second test response association: the heat dissipation efficiency type set contains the heat dissipation efficiency index entities of each working condition, such as the "35℃ normal temperature + 1.2m / s flow rate" working condition recording the efficiency values of 92.3% at the 30th minute, 91.8% at the 60th minute, and 88.0% at the 120th minute, and the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition recording the efficiency values of 78.2% at the 30th minute, 76.5% at the 60th minute, and 72.0% at the 180th minute; the noise type set contains the noise index entities of each working condition, such as the "35℃ normal temperature + 1.2m / s flow rate" working condition recording the noise values of 58dB at the 30th minute, 60dB at the 60th minute, and 62dB at the 120th minute, and the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition recording the noise values of 68dB at the 30th minute, 70dB at the 60th minute, and 72dB at the 180th minute; and the power consumption type set contains the power consumption index entities of each working condition, such as the "35℃ normal temperature + 1.2m / s flow rate" working condition recording the power consumption values of 95W at the 30th minute, 96W at the 60th minute, and 98W at the 120th minute, and the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition recording the power consumption values of 115W at the 30th minute, 118W at the 60th minute, and 122W at the 180th minute.

[0098] Then, the server generates a feature vector corresponding to the performance index of the working condition through the heat dissipation fan side feature extraction module: for the heat dissipation efficiency type set, the "heat dissipation efficiency feature mapping component" in the module processes the parameter collection data of each index entity, extracts the efficiency decay slope (such as the decay slope of the "35℃ normal temperature + 1.2m / s flow rate" working condition is -0.036% / min), the steady-state efficiency value (88.0%), the fluctuation amplitude (±0.5%) and the like, and generates a 16-dimensional heat dissipation efficiency feature vector; the noise type set extracts the noise amplitude slope (the amplitude of the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition is 0.028dB / min), the peak noise value (72dB) and the like through the "noise feature mapping component", and generates a 16-dimensional noise feature vector; the power consumption type set extracts the power consumption growth rate (the growth rate of the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition is 0.033W / min), the average power consumption (118W) and the like through the "power consumption feature mapping component", and generates a 16-dimensional power consumption feature vector. Subsequently, the "working condition performance index feature aggregator" of the module performs weighted fusion on the three 16-dimensional vectors based on the attention mechanism, and gives a weight of 30% to the full load working condition of the mold (i.e. the "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition), and finally generates a 48-dimensional feature vector corresponding to the performance index of the working condition, which contains the dynamic characteristics of each performance dimension, such as the overall trend of heat dissipation efficiency decay, the nonlinear growth law of noise with load, etc.

[0099] Finally, the server fuses the previously generated working condition time sequence dynamic flow feature vector (32-dimensional, reflecting the dynamic coupling effect of the working condition) and the working condition performance index feature vector (48-dimensional, reflecting the specific performance): first, the 32-dimensional and 48-dimensional vectors are spliced into an 80-dimensional vector, and then the principal component analysis (PCA, cumulative variance contribution rate ≥95%, the first 64 principal components are reserved in this embodiment) is used to retain key information and reduce the dimension to 64-dimensional. The 64-dimensional vector is the feature vector corresponding to the first heat dissipation fan (MF-800A), which contains not only the time sequence dynamic features of the working condition (such as temperature gradient, load change), but also the dynamic laws of core performance indexes such as heat dissipation efficiency, noise and power consumption, and can comprehensively quantify the performance characterization data of the fan in the whole process of industrial mold production.

[0100] In the embodiment of the application, the generation of at least one working condition performance index set corresponding to each of the working condition performance index types according to the sector coupling topology can be implemented by the following examples.

[0101] For the first working condition performance index type in the at least one working condition performance index type, at least one first working condition performance index entity corresponding to each working condition entity in the working condition time sequence dynamic flow in the first working condition performance index type is determined according to the second test response association in the sector coupling topology.

[0102] According to the at least one first working condition performance index entity corresponding to each working condition condition entity in the working condition time sequence dynamic flow, a working condition performance index set corresponding to the first working condition performance index type is generated.

[0103] In the embodiment of the application, in the industrial mold production scene, for the first cooling fan (MF-800A model, suitable for large injection mold), the server first determines at least one working condition performance index type, wherein the first working condition performance index type is the heat dissipation efficiency type (measuring the core heat dissipation capacity of the fan under different mold production working conditions). Then, according to the second test response association in the sector coupling topology, each working condition condition entity in the working condition time sequence dynamic flow is traversed to determine the corresponding first working condition performance index entity (i.e., the heat dissipation efficiency index entity) in the first working condition performance index type (the heat dissipation efficiency type).

[0104] The working condition time sequence dynamic flow contains four working condition condition entities, which are “35℃ normal temperature + 1.2m / s flow rate”, “35℃ normal temperature + 1.5m / s flow rate”, “45℃ medium temperature + 100W load” and “55℃ high temperature + 120W load + 8MPa cavity pressure” in sequence. The server finds that each working condition condition entity is directed to at least one heat dissipation efficiency index entity through the second test response association by querying the graph database of the sector coupling topology, and the association relationship contains a data acquisition identifier (such as “performance sampling stage”) for positioning the heat dissipation efficiency data in different test stages. For example, the second test response association record of the “35℃ normal temperature + 1.2m / s flow rate” working condition condition entity shows that it corresponds to three heat dissipation efficiency index entities: “initial stage efficiency entity” (sampling at the 30th minute of the test, recording the heat dissipation efficiency of 92.3%), “mid-stage efficiency entity” (sampling at the 60th minute of the test, recording 91.8%) and “steady-state efficiency entity” (sampling at the 120th minute of the test, recording 88.0%), and the “performance sampling stage” in the association attribute is marked as “initial stage”, “mid-stage” and “final stage” respectively, corresponding to different heat dissipation states of the mold preheating stage.

[0105] For the "35℃ normal temperature + 1.5m / s flow rate" working condition entity, the second test response associated heat dissipation efficiency indicator entity includes: the "initial efficiency entity" at the 30th minute of the test (93.1%), the "mid-term efficiency entity" at the 60th minute (92.5%), and the "steady-state efficiency entity" at the 120th minute (89.2%). The associated data also records the gain amplitude of the efficiency due to the flow rate increase (0.8% compared to the 1.2m / s flow rate). The "45℃ medium temperature + 100W load" working condition entity is associated with the second test response to the heat dissipation efficiency indicator entity: the "half-load initial efficiency entity" at the 30th minute of the test (85.6%), the "half-load mid-term efficiency entity" at the 60th minute (83.2%), and the "half-load steady-state efficiency entity" at the 150th minute (78.0%). The associated data records the attenuation effect of the medium temperature environment on the efficiency (about 11% lower than the normal temperature working condition). The "55℃ high temperature + 120W load + 8MPa cavity pressure" working condition entity (full load working condition of the mold) corresponds to the heat dissipation efficiency indicator entity, which includes: the "full load initial efficiency entity" at the 30th minute of the test (78.2%), the "full load mid-term efficiency entity" at the 60th minute (76.5%), and the "full load steady-state efficiency entity" at the 180th minute (72.0%). The associated data particularly records the efficiency attenuation slope under the coupling of high temperature and high pressure (0.034% per minute).

[0106] After determining the first working condition performance indicator entity (heat dissipation efficiency indicator entity) corresponding to each working condition condition entity, the server aggregates these entities to generate a working condition performance indicator set corresponding to the first working condition performance indicator type (heat dissipation efficiency type). The set includes all the heat dissipation efficiency indicator entities described above, covering the heat dissipation efficiency dynamic data of the mold in each working condition in the whole process from preheating to full load production, providing a basis for subsequent extraction of the heat dissipation efficiency dimension feature vector.

[0107] In the embodiments of the present application, the generation of the feature vector corresponding to the working condition performance indicator according to the working condition performance indicator set corresponding to each of the at least one working condition performance indicator type can be implemented by the following examples.

[0108] For the first working condition performance indicator type in the at least one working condition performance indicator type, the determination of the parameter acquisition data corresponding to each of the at least one working condition performance indicator entity included in the working condition performance indicator set corresponding to the first working condition performance indicator type; wherein the parameter acquisition data corresponding to each of the working condition performance indicator entities is used to indicate the working condition performance indicator corresponding to the working condition performance indicator entity;

[0109] The first working condition performance index type corresponding to the working condition performance index feature mapping component in the heat dissipation fan side feature extraction module is used for performing feature extraction processing on the parameter acquisition data corresponding to each of the at least one working condition performance index entity included in the working condition performance index set corresponding to the first working condition performance index type, to obtain a feature vector corresponding to the first working condition performance index type; wherein the heat dissipation fan side feature extraction module is used to generate a feature vector corresponding to the first heat dissipation fan;

[0110] The working condition performance index feature aggregator in the heat dissipation fan side feature extraction module is used to perform feature extraction on the feature vectors corresponding to each of the at least one working condition performance index type, to obtain a feature vector corresponding to the working condition performance index.

[0111] In the embodiment of the application, in the industrial mold production scene, for the first cooling fan (MF-800A model, suitable for large injection mold), the server first determines at least one working condition performance index type, wherein the first working condition performance index type is the heat dissipation efficiency type (core evaluation dimension, directly related to the mold cooling effect). The server extracts the parameter collection data of all working condition performance index entities in the corresponding working condition performance index set of this type by accessing the time sequence database of the performance evaluation system. This set contains the heat dissipation efficiency index entity associated with the four working condition condition entities in the working condition time sequence dynamic flow (“35°C normal temperature + 1.2m / s flow rate” “35°C normal temperature + 1.5m / s flow rate” “45°C medium temperature + 100W load” “55°C high temperature + 120W load + 8MPa cavity pressure”). The parameter collection data of each entity is uploaded in real time through the industrial bus interface (such as Profinet) of the test equipment, including the fields of “test timestamp” “efficiency measured value” “environmental interference coefficient” (used to correct the influence of temperature fluctuation on efficiency). For example, in the heat dissipation efficiency index entity of the “35°C normal temperature + 1.2m / s flow rate” working condition, the parameter collection data is as follows: the efficiency measured value of the 30th minute (timestamp “T00:30:00”) is 92.3%, the environmental interference coefficient is 0.98 (the corrected efficiency is 90.45%); the measured value of the 60th minute (“T01:00:00”) is 91.8%, the interference coefficient is 0.97 (after correction, 90.05%); the measured value of the 120th minute (“T02:00:00”) is 88.0%, the interference coefficient is 0.96 (after correction, 84.48%). The heat dissipation efficiency index entity data of the “55°C high temperature + 120W load + 8MPa cavity pressure” working condition (mold full load production stage) is as follows: the measured value of the 30th minute (“T00:30:00”) is 78.2%, the interference coefficient is 0.92 (after correction, 72.0%); the measured value of the 60th minute (“T01:00:00”) is 76.5%, the interference coefficient is 0.91 (after correction, 69.6%); the measured value of the 180th minute (“T03:00:00”) is 72.0%, the interference coefficient is 0.90 (after correction, 64.8%). These data completely record the dynamic changes of heat dissipation efficiency with time and the environmental interference correction results under different working conditions, which are used to accurately indicate the working condition performance index corresponding to the heat dissipation efficiency index entity.

[0112] Subsequently, the server inputs the parameter collection data into the "heat dissipation efficiency feature mapping component" corresponding to the first working condition performance index type (heat dissipation efficiency type) of the heat dissipation fan side feature extraction module. This component is designed for industrial time series performance data, with built-in time series trend analysis and key indicator extraction algorithms: first, the corrected efficiency value is normalized (map 0-100% efficiency to the [0,1] interval, such as 84.48% corresponding to 0.8448), eliminating the influence of environmental interference coefficients; then linear fitting is performed on the time series data by least squares method to extract the heat dissipation efficiency decay slope (such as the decay slope of the "35°C constant temperature + 1.2m / s flow rate" working condition is -0.0005 / minute, that is, the efficiency decreases by 0.05% per minute); at the same time, the steady-state efficiency value (take the average value of the last 30 minutes of testing, such as the steady-state efficiency of this working condition is 84.48%), the efficiency fluctuation amplitude (calculate the standard deviation of all sampling points, ±0.32%) and the initial-steady-state efficiency difference (90.45%-84.48%=5.97%) are calculated. Through these key indicators, the component compresses the multi-dimensional time series data into a 16-dimensional heat dissipation efficiency feature vector, and each dimension in the vector corresponds to a core characteristic (such as dimension 1 for decay slope, dimension 2 for steady-state efficiency, dimension 3 for fluctuation amplitude, etc.).

[0113] For other types (noise type, power consumption type) in at least one working condition performance index type, the server performs the same process: the parameter collection data of the noise type working condition performance index set includes the "time stamp" "noise decibel value" "acoustic frequency spectrum distribution" (such as 1000Hz frequency band noise proportion) of each working condition noise indicator entity, after inputting into the "noise feature mapping component", extracting noise amplitude slope (0.028dB / minute for the "55°C high temperature + 120W load + 8MPa cavity pressure" working condition), peak noise (72dB), spectral concentration (0.85, the higher the value, the more concentrated the noise frequency band), etc., to generate a 16-dimensional noise feature vector; the parameter collection data of the power consumption type set includes "time stamp" "power consumption wattage" "power factor", after inputting into the "power consumption feature mapping component", extracting power consumption growth rate (0.033W / minute), average power consumption (118W), power factor fluctuation (±0.02), etc., to generate a 16-dimensional power consumption feature vector.

[0114] Finally, the server aggregates the above three 16-dimensional feature vectors (heat dissipation efficiency type, noise type, and power consumption type) through a "performance index feature aggregator" in the heat dissipation fan side feature extraction module. Based on the risk weight rules of industrial mold production (the highest risk in full load working condition), the aggregator gives different weights to the performance features of each working condition: the feature weight of the mold full load working condition ("55°C high temperature + 120W load + 8MPa cavity pressure") is set to 30%, the half load working condition ("45°C medium temperature + 100W load") is set to 25%, and the preheating stage working condition ("35°C normal temperature") is set to 22.5% respectively. Through weighted summation and feature dimension reduction (retaining 95% information entropy), the aggregator finally outputs a 48-dimensional feature vector corresponding to the performance index, which integrates the attenuation trend of heat dissipation efficiency, the nonlinear growth law of noise, the coupling sensitivity of power consumption and load, and other core dynamic characteristics, providing key input of performance dimension for the generation of subsequent fan feature vectors.

[0115] In the embodiments of the present application, the generation of the feature vector corresponding to the first heat dissipation fan according to the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the working condition performance index can be implemented by the following examples.

[0116] The feature fusion component of the heat dissipation fan feature extraction module fuses the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the working condition performance index to obtain a first fused feature vector;

[0117] The feature generation component in the heat dissipation fan side feature extraction module extracts features from the first fused feature vector to obtain the feature vector corresponding to the first heat dissipation fan.

[0118] In the embodiments of the present application, for example, in the industrial mold production scene, for the first heat dissipation fan (MF-800A type, suitable for large injection mold), the server calls the "feature fusion component" of the heat dissipation fan side feature extraction module to fuse the previously generated working condition time sequence dynamic flow feature vector (32-dimensional, reflecting the dynamic coupling effect of working condition conditions such as temperature gradient, load change and fan process correlation) and working condition performance index feature vector (48-dimensional, reflecting the dynamic characteristics of heat dissipation efficiency, noise, and power consumption, such as efficiency decay slope and noise amplitude law). The fusion adopts a "feature splicing" strategy, which directly splices the 32-dimensional and 48-dimensional vectors in order to form an 80-dimensional first fused feature vector that contains both dynamic correlation information of working condition conditions and core characteristics of performance indicators, for example, the temperature coupling effect of the "55°C high temperature + 120W load" working condition (dimension 12 in the 32-dimensional vector) and the corresponding heat dissipation efficiency decay slope (dimension 5 in the 48-dimensional vector) form a continuous feature sequence after splicing, retaining the correlation between the two.

[0119] Subsequently, the server processes the 80-dimensional first fusion feature vector through a "feature generation component" in the module. The component is built-in with a principal component analysis (PCA) algorithm optimized for industrial data, extracts principal components with the largest variance contribution by calculating the covariance matrix, retains 95% of the feature information to avoid overfitting, and reduces the 80-dimensional vector to 64 dimensions. During the dimension reduction process, key features of the mold under full load conditions ("55℃ high temperature + 120W load + 8MPa cavity pressure") are mainly retained, such as the comprehensive attenuation rate (integrated heat dissipation efficiency, noise, and power consumption attenuation trend, corresponding to vector dimension 8), working condition coupling sensitivity (reflecting the nonlinear correlation of temperature-load-performance, corresponding to dimension 23), and steady-state performance deviation (difference between measured value and design value, corresponding to dimension 45), etc. Finally, the 64-dimensional vector output by the feature generation component is the feature vector corresponding to the first cooling fan (MF-800A), which comprehensively quantifies the performance characterization data of the fan in the entire production process of the industrial mold and can be directly used for subsequent matching with the performance benchmark value.

[0120] In the embodiment of the application, the sector-coupled topology further comprises: at least one cooling fan characteristic entity having a third test response association with the first fan entity; wherein the third test response association is used to reflect that the cooling fan corresponding to the fan entity has the cooling fan characteristic corresponding to the cooling fan characteristic entity; and the method further comprises:

[0121] According to the sector-coupled topology, at least one cooling fan characteristic entity corresponding to each of the cooling fan characteristic dimensions is generated; wherein the cooling fan characteristic set comprises at least one working condition performance index entity having a third test response association with the first fan entity;

[0122] According to the cooling fan characteristic set corresponding to each of the at least one cooling fan characteristic dimension, a feature vector corresponding to the cooling fan characteristic is generated;

[0123] The generating of the feature vector corresponding to the first cooling fan according to the feature vector corresponding to the working condition time sequence dynamic flow comprises:

[0124] The generating of the feature vector corresponding to the first cooling fan according to the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the cooling fan characteristic.

[0125] In the embodiment of the application, in the industrial mold production scene, the sector-coupling topology of the first cooling fan (MF-800A model, suitable for large injection molds) also contains a cooling fan characteristic entity associated with the first fan entity (MF-800A) through a third test response, which is used to reflect the physical and performance characteristics inherent to the fan itself, such as structural parameters, electrical parameters, mechanical reliability, etc., which directly affect the adaptability of the fan under different mold working conditions. The server discovers through querying the graph database of the performance evaluation system that the MF-800A fan entity is associated with three types of cooling fan characteristic entities through the third test response: structural characteristic entities (such as size, material), electrical characteristic entities (such as rated voltage, power), and mechanical characteristic entities (such as speed, bearing type). The association relationship contains a characteristic category identifier (such as "characteristic code") for distinguishing inherent attributes of different dimensions.

[0126] The server first determines the cooling fan characteristic dimensions according to the industrial fan design specification: structural characteristic type (affecting installation adaptability and basic cooling potential), electrical characteristic type (affecting compatibility with the mold power supply system), and mechanical characteristic type (affecting long-term operation reliability). For each dimension, through the third test response association in the sector-coupling topology, extract the cooling fan characteristic entities associated with the MF-800A fan entity, and generate a characteristic entity set corresponding to the dimension. For example, the structural characteristic type set contains "size characteristic entities" (recorded fan diameter 120mm, thickness 38mm, suitable for mold cooling hole size), "material characteristic entities" (aluminum alloy fan blades, thermal conductivity coefficient 202W / (m·K), plastic frame, heat resistance temperature 120℃); the electrical characteristic type set contains "rated voltage entities" (DC 24V, fluctuation range ±5%), "rated power entities" (120W, starting power 150W); the mechanical characteristic type set contains "speed characteristic entities" (maximum speed 2800rpm, speed regulation range 1500-2800rpm), "bearing characteristic entities" (double ball bearing, service life 50000 hours, temperature resistance -20℃~70℃).

[0127] Next, the server extracts parameter data from each characteristic entity. This data is stored in the fan's factory testing database, including design and measured values ​​(e.g., in the dimensional characteristic entity, the diameter design value is 120mm, the measured value is 119.8mm, and the deviation is -0.2mm). The server inputs this data into the "characteristic feature mapping component" corresponding to each characteristic dimension in the cooling fan side feature extraction module: structural characteristic data is input into the "structural feature mapping component" to extract dimensional deviation rate (-0.17%), material thermal conductivity (202W / (m·K)), etc., generating a 12-dimensional structural feature vector; electrical characteristic data is input into the "electrical feature mapping component" to extract voltage fluctuation tolerance (±5%), power deviation (measured 119.7W vs design 120W, deviation -0.25%), etc., generating a 12-dimensional electrical feature vector; mechanical characteristic data is input into the "mechanical feature mapping component" to extract speed adjustment accuracy (±50rpm), bearing life (50,000 hours), etc., generating a 12-dimensional mechanical feature vector. Subsequently, the module's "feature aggregator" performs weighted aggregation on these three 12-dimensional vectors (40% for structural characteristics, 35% for mechanical characteristics, and 25% for electrical characteristics, as the mold scenario focuses more on installation compatibility and long-term reliability), generating a 36-dimensional feature vector corresponding to the cooling fan characteristics. This vector contains a quantitative description of the fan's inherent attributes, such as structural compatibility score (0.98 / 1.0, calculated based on dimensional deviation) and mechanical reliability index (0.95 / 1.0, based on bearing life and speed stability).

[0128] Finally, the server fuses the previously generated operating condition time-series dynamic flow feature vector (32-dimensional, reflecting the dynamic coupling effect between operating conditions and the fan) with the cooling fan characteristic feature vector (36-dimensional, reflecting the inherent properties of the fan). The fusion is completed by the "feature fusion component" of the cooling fan-side feature extraction module, employing a "weighted concatenation" strategy: first, the operating condition time-series dynamic flow feature vector is assigned a 60% weight (because the dynamic performance of the operating condition better reflects actual operating performance), and the characteristic feature vector is assigned a 40% weight, then concatenated in dimensional order to form a 68-dimensional fusion vector. Subsequently, the module's "feature generation component" uses principal component analysis (PCA) to retain 95% of the information entropy, reducing the 68-dimensional vector to 48 dimensions, focusing on retaining key characteristics (such as the interaction term between structural adaptability score and high-temperature operating conditions, and the correlation between mechanical reliability index and long-term operating efficiency decay). The final 48-dimensional vector is the feature vector corresponding to the first cooling fan (MF-800A). This vector contains both the dynamic performance of the fan under mold conditions and its inherent characteristics, providing comprehensive data support for accurate matching with the performance benchmark value in the future.

[0129] In this embodiment of the invention, generating a feature vector corresponding to a cooling fan characteristic based on a cooling fan characteristic set corresponding to at least one cooling fan characteristic dimension includes:

[0130] For a first heat dissipation fan characteristic dimension in the at least one heat dissipation fan characteristic dimension, determining that the first heat dissipation fan characteristic dimension corresponds to parameter collection data of each heat dissipation fan characteristic entity included in a heat dissipation fan characteristic set; the parameter collection data of each heat dissipation fan characteristic entity is used to indicate a heat dissipation fan characteristic corresponding to the heat dissipation fan characteristic entity;

[0131] Through a heat dissipation fan characteristic feature mapping component corresponding to the first heat dissipation fan characteristic dimension in a heat dissipation fan side feature extraction module, performing feature extraction processing on an identity identifier of the at least one heat dissipation fan characteristic entity to obtain an embedding vector of the first heat dissipation fan characteristic dimension; wherein the heat dissipation fan side feature extraction module is used to generate a feature vector corresponding to the first heat dissipation fan characteristic dimension;

[0132] Through a heat dissipation fan characteristic feature aggregator in the heat dissipation fan side feature extraction module, performing feature extraction on the feature vector corresponding to each of the at least one heat dissipation fan characteristic dimension to obtain a feature vector corresponding to the heat dissipation fan characteristic.

[0133] In an embodiment of the present application, exemplary, in the industrial mold production scenario, for the first cooling fan (MF-800A model, suitable for large injection mold), the server first determines at least one cooling fan characteristic dimension, wherein the first cooling fan characteristic dimension is a structural characteristic type (directly affects the installation adaptability of the fan and the mold cooling hole and the basic cooling potential, is the core inherent characteristic of the industrial mold scenario). The server extracts the parameter collection data of all characteristic entities in the cooling fan characteristic set corresponding to this dimension by accessing the fan factory detection database. This set contains the "size characteristic entity", "material characteristic entity" and "mounting interface characteristic entity" associated with the MF-800A fan entity through the third test response, and the parameter collection data of each entity records the key indicators of the fan structure design, including "characteristic identification ID" (such as "STRUCT-001" corresponding to the size characteristic), "design value", "actual measurement value", "deviation rate" and "industrial standard threshold value" (used to judge whether it meets the mold installation specification). For example, the parameter collection data of the "size characteristic entity" is: characteristic identification ID "STRUCT-001", design diameter 120mm, actual measurement diameter 119.8mm (deviation rate -0.17%), design thickness 38mm, actual measurement 37.9mm (deviation rate -0.26%), industrial standard threshold value ±0.5% (deviation is within the qualified range); The parameter collection data of the "material characteristic entity" is: characteristic identification ID "STRUCT-002", fan blade material aluminum alloy (code "AL-6061"), thermal conductivity 202W / (m·K) (design value 200W / (m·K), actual measurement deviation +1%), frame material ABS plastic (code "ABS-H100"), heat resistance temperature 120℃ (design value 110℃, actual measurement deviation +9.09%); The parameter collection data of the "mounting interface characteristic entity" is: characteristic identification ID "STRUCT-003", screw hole spacing 80mm (design value 80mm, actual measurement 79.9mm, deviation rate -0.125%), interface type "4-pin PWM" (complies with the mold cooling system communication protocol).

[0134] The server inputs the above parameter collection data into the "structure characteristic feature mapping component" corresponding to the first heat dissipation fan characteristic dimension (structure characteristic type) of the heat dissipation fan side feature extraction module. This component is designed for industrial structure parameters, with built-in characteristic coding and numerical conversion logic: first, the "characteristic identifier ID" of the characteristic entity is coded (for example, "STRUCT-001" is mapped to [1, 0, 0], and "STRUCT-002" is mapped to [0, 1, 0]), then the parameter data is normalized (for example, the deviation rate of -0.17% is normalized to the [-1, 1] interval as -0.34, and the thermal conductivity of 202 W / (m·K) is normalized to [0, 1] as 0.91), and then the key structure characteristics are extracted through multi-layer perception (MLP), such as size comprehensive deviation rate (integrating diameter and thickness deviation, calculated as -0.215%), material performance redundancy (difference between actual measured value of heat resistance temperature 120°C and maximum environmental temperature of mold 55°C, 65°C after normalization, 0.65), interface adaptation degree (hole spacing deviation -0.125%, protocol matching degree 100%, comprehensive 0.99). Through these key features, the component compresses the multi-dimensional structure parameters into a 12-dimensional structure characteristic embedding vector, and each dimension in the vector corresponds to a core structure index (such as dimension 1 for size comprehensive deviation rate, dimension 2 for material thermal conductivity, dimension 3 for interface adaptation degree, etc.).

[0135] For other dimensions in at least one heat dissipation fan characteristic dimension (electrical characteristic type, mechanical characteristic type), the server performs the same process: the electrical characteristic type dimension is processed by the "electrical characteristic feature mapping component", and the characteristic entity includes "rated voltage entity" (DC 24V, actual measured value 23.8V, deviation -0.83%) and "power entity" (rated 120W, actual measured value 119.7W, deviation -0.25%), and extracts key indicators such as voltage stability (fluctuation ±1.2%) and power deviation rate to generate a 12-dimensional electrical characteristic embedding vector; the mechanical characteristic type dimension is processed by the "mechanical characteristic feature mapping component", and the characteristic entity includes "rotational speed entity" (maximum 2800rpm, actual measured value 2790rpm, deviation -0.36%) and "bearing entity" (double ball bearing, service life 50000 hours, actual measured value 51200 hours, deviation +2.4%), and extracts indicators such as rotational speed adjustment accuracy (±45rpm) and bearing life redundancy to generate a 12-dimensional mechanical characteristic embedding vector.

[0136] Finally, the server aggregates the above three 12-dimensional embedding vectors (structural characteristics, electrical characteristics, mechanical characteristics) through the "heat dissipation fan characteristic feature aggregator" in the heat dissipation fan side feature extraction module. The aggregator fuses the three 12-dimensional vectors into a 36-dimensional vector by weighted summation based on the importance weight of the characteristics of the industrial mold scene (the structural characteristics affect the installation adaptation, giving a weight of 40%; the mechanical characteristics affect the long-term reliability, giving a weight of 35%; the electrical characteristics affect the power supply compatibility, giving a weight of 25%), and then enhances the nonlinear feature expression (such as highlighting the coupling relationship between structural deviation and mechanical life) through the ReLU activation function. Finally, the 36-dimensional vector output by the aggregator is the feature vector corresponding to the characteristics of the heat dissipation fan, which contains the quantitative representation of the inherent structure, electrical, and mechanical characteristics of the fan, such as structural comprehensive adaptation degree (integration size, interface, material characteristics, corresponding dimension 5), mechanical reliability index (based on speed stability and bearing life, corresponding dimension 18), and electrical compatibility score (based on voltage fluctuation and power deviation, corresponding dimension 30), etc., providing accurate input of fan inherent properties for subsequent fusion with dynamic characteristics of working conditions.

[0137] In the embodiment of the application, the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to each of the at least one alternative performance benchmark value is determined, and the at least one performance benchmark value matched by the first heat dissipation fan can be implemented by the following examples.

[0138] Determine the feature vector corresponding to each of the at least one alternative performance benchmark value.

[0139] For a first performance benchmark value in the at least one alternative performance benchmark value, calculate the feature space approximation degree between the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value, and determine the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value.

[0140] From the at least one alternative performance benchmark value, select an alternative performance benchmark value whose performance matching degree is not lower than a performance matching degree threshold as a performance benchmark value matched by the first heat dissipation fan.

[0141] In an embodiment of the application, in an industrial mold production scenario, for a first cooling fan (MF-800A model, suitable for large injection molds), the server first determines at least one alternative performance benchmark value, which is derived from a standard performance index library of industrial mold cooling systems, covering the cooling needs of different mold types, including "large injection mold benchmark value" (suitable for high temperature and high load working conditions), "precision stamping mold benchmark value" (suitable for medium temperature and medium load working conditions), and "small die-casting mold benchmark value" (suitable for normal temperature and low load working conditions). The server generates a feature vector corresponding to each alternative benchmark value by calling a working condition side feature extraction module (similar in architecture to the fan side module, designed specifically for benchmark value characterization). Taking the "large injection mold benchmark value" (first performance benchmark value) as an example, its reference working condition performance indicators include: cooling efficiency ≥ 75% under 55°C high temperature working condition, noise ≤ 70dB, power consumption ≤ 125W, and steady-state running time ≥ 180 minutes under 8MPa cavity pressure. The server generates the feature vector of this benchmark value through the working condition side feature extraction module, with the following steps:

[0142] 1) Text word feature extraction: Use an industrial field fine-tuned BERT-base model (12-layer Transformer, 768-dimensional hidden layer) to embed keywords such as "large injection mold", "full load working condition", and "high temperature cooling", resulting in a 768-dimensional word semantic vector; then reduce it to 32 dimensions through a fully connected layer (activation function ReLU) to retain the core scene semantics;

[0143] 2) Threshold requirement conversion (take "cooling efficiency ≥ 75%" as an example):

[0144] a. Numerical normalization: normalize the threshold value 75% to 0.75 by th_norm = th / 100 (since the maximum cooling efficiency is 100%);

[0145] b. Threshold type coding: "≥" (lower limit requirement) is coded as [1, 0], and "≤" (upper limit requirement) is coded as [0, 1];

[0146] c. Feature fusion: concatenate "32-dimensional word semantic vector + 0.75 normalized threshold + [1, 0] type code" to get a 35-dimensional vector; then reduce it to 16 dimensions through a fully connected layer to ensure uniformity with other indicator dimensions;

[0147] ③Multi-index fusion: The 16-dimensional vector of the four indexes of "heat dissipation efficiency, noise, power consumption, and steady-state running time" is spliced into 64 dimensions, and the dimensionality is confirmed to be reasonable through PCA verification (cumulative variance contribution rate 95%) to finally generate a 64-dimensional benchmark value feature vector. Dimension 8 in the vector corresponds to "high-temperature heat dissipation efficiency threshold (0.75)", dimension 23 corresponds to "noise upper limit (70dB→0.70)", and dimension 45 corresponds to "power consumption upper limit (125W→0.83, formula 125 / 150, 150W is the maximum power upper limit of industrial fans)".

[0148] Then, the server calculates the performance matching degree between the MF-800A fan feature vector (64 dimensions) and the benchmark value feature vector (64 dimensions) for the first performance benchmark value (large injection mold benchmark value). The matching degree is calculated by the "feature space weighted approximation degree" algorithm, which gives different working condition dimensions different weights based on the characteristics of the industrial scene: the feature dimensions related to the full load working condition of the mold (such as dimension 8 "comprehensive attenuation rate" and dimension 23 "working condition coupling sensitivity" in the MF-800A feature vector) are given a weight of 40%, the half load working condition dimensions are given a weight of 30%, and the preheating working condition dimensions are given a weight of 30%. In the specific calculation, the cosine similarity of the corresponding dimensions of the two vectors is calculated first (for example, the high-temperature heat dissipation efficiency feature value of MF-800A is 0.782, the benchmark value is 0.75, and the similarity is 0.98), and then the weighted sum is calculated according to the weight. After calculation, the weighted approximation degree of MF-800A and the large injection mold benchmark value is 0.92 (i.e. performance matching degree 0.92); the matching degree with the precision stamping die benchmark value (requiring heat dissipation efficiency ≥85% at 45°C medium temperature) is 0.76 (because the efficiency of MF-800A at 45°C working condition is 85.6% close to the threshold, but the weight of high-temperature working condition is low, resulting in insufficient overall matching degree); the matching degree with the small die-casting die benchmark value is 0.68 (because the small die requires high efficiency at low load, and the low load efficiency advantage of MF-800A is not obvious).

[0149] Finally, the server sets the performance matching degree threshold to 0.85 according to the industrial mold test specification (i.e. matching degree ≥0.85 is considered to be suitable), and selects the results that meet the conditions from the alternative benchmark values. Comparison of calculation results: the matching degree of the large injection mold benchmark value is 0.92≥0.85, the matching degree of the precision stamping die benchmark value is 0.76<0.85, and the matching degree of the small die-casting die benchmark value is 0.68<0.85. Therefore, the server determines that the performance benchmark value matched by the first heat dissipation fan (MF-800A) is the "large injection mold benchmark value", which will be used as the basis for subsequent test working condition adaptation to ensure that the fan meets the heat dissipation requirements of large injection molds under high temperature and high load.

[0150] In the embodiments of the present application, the determination of the feature vector corresponding to each of the at least one alternative performance benchmark value can be implemented by the following examples.

[0151] For a first performance benchmark value in the at least one alternative performance benchmark value, at least one reference working condition performance indicator corresponding to the first performance benchmark value is obtained; wherein the reference working condition performance indicator has performance equivalence with the first performance benchmark value;

[0152] According to the first performance benchmark value and the at least one reference working condition performance indicator corresponding to the first performance benchmark value, a feature vector corresponding to the first performance benchmark value is determined.

[0153] In the embodiments of the present application, for example, in the industrial mold production scene, for the first cooling fan (MF-800A model, suitable for large injection molds), the server determines the first performance benchmark value in the at least one alternative performance benchmark value as "large injection mold cooling system standard performance benchmark value" (hereinafter referred to as "injection benchmark value"), which is based on GB / T14880-2021 "Industrial Ventilator Performance Test Standard" and mold industry cooling specification. The core requirement is "to ensure that the mold can continuously cool under the full load working condition of 55℃ high temperature, 120W load and 8MPa cavity pressure, and the key performance indicators do not exceed the industrial safety threshold". The server accesses the benchmark value database of the performance evaluation system to obtain at least one reference working condition performance indicator corresponding to the injection benchmark value. These indicators have performance equivalence with the benchmark value, that is, the performance boundary of the benchmark value is directly defined by quantitative indicators, which specifically include: high-temperature cooling efficiency indicator (cooling efficiency ≥75% under 55℃ working condition, which is the core requirement of mold cooling effect), noise control indicator (noise ≤70dB under full load working condition, which meets the workshop acoustic environment standard), power consumption limit indicator (continuous operation power consumption ≤125W, which avoids mold power supply system overload), and steady-state running time indicator (continuous running ≥180 minutes under 8MPa cavity pressure, which avoids performance decay exceeding limit, which corresponds to mold continuous production demand). Each reference indicator includes "working condition condition label" (such as "high temperature full load"), "threshold type" (lower limit / upper limit) and "industrial standard code" (such as "EFF-55-075" corresponding to 55℃ cooling efficiency threshold), which ensures one-to-one correspondence with the performance requirements of the injection benchmark value.

[0154] Subsequently, the server generates a corresponding feature vector through a working condition side feature extraction module according to the injection benchmark value and its reference working condition performance indicators. The module first converts the threshold values of the reference indicators into standardized data through a "benchmark value parameter analysis component": the high-temperature heat dissipation efficiency indicator "> 75%" is converted into a normalized value of 0.75 (0-1 interval, 1 corresponds to 100%), the noise control indicator "<= 70dB" is converted into 0.70 (0 corresponds to 0dB, 1 corresponds to 100dB), the power consumption limit indicator "<= 125W" is converted into 0.83 (125W / 150W, 150W is the maximum power upper limit of industrial fans), and the steady-state running time indicator "> 180 minutes" is converted into 0.90 (180 minutes / 200 minutes, 200 minutes is the longest continuous production cycle of the mold). Then, the "benchmark value feature mapping component" performs feature coding on these standardized data, assigns a fixed dimension to each indicator (such as high-temperature heat dissipation efficiency corresponding to dimension 8, noise corresponding to dimension 23, power consumption corresponding to dimension 45, and steady-state time corresponding to dimension 56), and fills them into a 64-dimensional initial vector. Finally, the "feature generation component" retains 95% information entropy by principal component analysis (PCA) and removes redundant dimensions (such as low-temperature working condition indicator dimensions irrelevant to the injection mold), generates a 64-dimensional injection benchmark value feature vector, and the dimension 8 in the vector is 0.75 (high-temperature heat dissipation efficiency threshold), the dimension 23 is 0.70 (noise upper limit), the dimension 45 is 0.83 (power consumption threshold), and the dimension 56 is 0.90 (steady-state time threshold), which fully reflects the performance requirements of the injection benchmark value on the large injection mold heat dissipation fan.

[0155] In the embodiments of the present application, the determination of the feature vector corresponding to the first performance benchmark value according to the first performance benchmark value and at least one reference working condition performance indicator corresponding to the first performance benchmark value can be implemented by the following examples.

[0156] The performance benchmark value feature mapping component of the working condition side feature extraction module performs feature extraction processing on the parameter acquisition data corresponding to each of the at least one word in the first performance benchmark value, to obtain a parameter representation corresponding to the first performance benchmark value;

[0157] The working condition parameter mapping unit of the working condition side feature extraction module performs feature extraction processing on the parameter acquisition data corresponding to each of the at least one word included in the at least one working condition performance indicator, to obtain a parameter representation corresponding to the working condition performance indicator;

[0158] The feature fusion component of the working condition side feature extraction module performs fusion processing on the parameter representation corresponding to the first performance benchmark value and the parameter representation corresponding to the working condition performance indicator, to obtain a performance fusion representation;

[0159] The performance fusion representation is subjected to feature extraction processing by a feature generation component of the working condition side feature extraction module, to determine a feature vector corresponding to the first performance benchmark value.

[0160] In an embodiment of the present application, in the industrial mold production scenario, for the first performance benchmark value "large injection mold cooling system standard performance benchmark value" (hereinafter referred to as "injection benchmark value"), the server generates a corresponding feature vector through the working condition side feature extraction module, and the specific process is as follows:

[0161] The server first calls the "performance benchmark value feature mapping component" of the module to perform feature extraction on the parameter collection data corresponding to the key "words" in the description text of the injection benchmark value. The text description of the injection benchmark value is "standard of heat dissipation performance of large injection mold under full load condition (GB / T14880-2021)", and the core words include "large injection mold", "full load condition", "heat dissipation performance", and "GB / T14880-2021". The parameter collection data associated with these words is stored in the industrial standard database, including "scene type code" (such as "MOLD-LARGE" corresponding to large injection mold), "working condition level" ("FULL-LOAD" corresponding to full load), "performance dimension label" ("HEAT-DISSIPATION" corresponding to heat dissipation performance), and "standard version number" ("GB2021"). The component converts these words into structured parameters through a pre-trained industrial text embedding model (based on BERT fine-tuning): "large injection mold" is mapped to the scene code vector [1, 0, 0] (distinguished from small and medium-sized molds), "full load condition" is mapped to the working condition level vector [0, 0, 1] (distinguished from preheating, half load, and full load), "heat dissipation performance" is mapped to the performance dimension weight [0.8, 0.1, 0.1] (heat dissipation accounts for 80% of the weight, noise accounts for 10%, and power consumption accounts for 10%), and "GB / T14880-2021" is mapped to the standard compliance code 1 (indicating compliance with the standard). The component integrates these structured parameters into a 32-dimensional "benchmark value parameter representation vector", where dimensions 1-3 are scene codes, 4-6 are working condition levels, 7-9 are performance dimension weights, 10 is standard compliance, and the remaining dimensions are filled with default industrial scene parameters (such as mold material compatibility coefficient 0.95).

[0162] Next, the server processes the parameter data collected from the key words in the reference operating condition performance indicators (high-temperature heat dissipation efficiency, noise, power consumption, and steady-state operating time) of the injection molding baseline value through the module's "operating condition parameter mapping unit". Taking the indicator "55℃ high-temperature heat dissipation efficiency ≥75%" as an example, the words in its text description include "55℃ high temperature", "heat dissipation efficiency" and "≥75%", and the corresponding parameter data collected are: "operating condition value" 55 (unit: ℃), "performance indicator type" "EFF" (heat dissipation efficiency code), "threshold symbol" "≥" (lower threshold), "threshold value" 75 (unit: %), and "industrial standard code" "EFF-55-075". The unit uses an "index structured parsing algorithm" to convert these terms into parameters: "55℃ high temperature" is converted into a temperature normalization value of 0.55 (55℃ / 100℃); "heat dissipation efficiency" is converted into an index type code [1,0,0] (distinguishing between heat dissipation, noise, and power consumption); and "≥75%" is converted into a threshold vector [0.75,1] (0.75 is the normalization threshold, and 1 represents the lower limit). The same processing is applied to the other three reference indices: "noise ≤70dB" is converted into [0.70,0] (0.70 is the normalization value, and 0 represents the upper limit), and "power consumption ≤125W" is converted into [0.83,0].

[0163] (125W / 150W=0.83), "steady-state operating time ≥180 minutes" is converted to [0.90,1](180 minutes / 200 minutes=0.90). The unit integrates the parameters of the four indicators into a 48-dimensional "operating condition indicator parameter representation vector", with each indicator corresponding to 12 dimensions (operating conditions, indicator type, threshold, etc.).

[0164] The server invokes the module's "feature fusion component" to fuse the baseline parameter representation vector (32 dimensions) and the operating condition indicator parameter representation vector (48 dimensions). The fusion is based on the correlation weights of the industrial scenario: the "full load operating condition" label of the injection molding baseline (dimension 5 of the baseline parameter representation) is highly correlated with the "55℃ high temperature + 120W load + 8MPa cavity pressure" operating condition in the reference indicator, and is assigned a 30% weight to this correlation dimension; the "heat dissipation performance" dimension weight (dimension 7 of the baseline parameter representation) is correlated with the high-temperature heat dissipation efficiency indicator (dimensions 1-12 of the operating condition indicator parameter representation), and is assigned a 25% weight; the remaining dimensions are fused according to the default weight (45%). In practice, the component merges the 32-dimensional and 48-dimensional vectors into 80 dimensions through matrix concatenation, and then strengthens the features of high-weight dimensions (such as the operating condition level of the baseline value and the temperature conditions of the indicator) through an attention mechanism, generating an 80-dimensional "performance fusion representation," which includes the scenario attributes of the baseline value and the quantitative requirements of the indicator, such as dimension 23 corresponding to the coupled feature of "full load operating condition + high-temperature heat dissipation efficiency."

[0165] Finally, the server processes the 80-dimensional performance fusion representation through the module's "feature generation component". The component is built-in with industrial standard principal component analysis (PCA) algorithm, calculates the covariance matrix of the fusion representation, extracts the top 64 principal components (retaining 95% information entropy), and eliminates redundant dimensions (such as "small mold adaptability" dimension irrelevant to injection mold). During the dimension reduction process, the key coupling features are preserved: dimension 8 corresponds to "heat dissipation efficiency threshold under full load working condition" (fusion of "full load" label of benchmark value and 75% threshold of the indicator, normalized value 0.75), dimension 23 corresponds to "high temperature noise upper limit" (fusion of 55℃ working condition and 70dB threshold, normalized value 0.70), and dimension 45 corresponds to "steady power consumption control" (fusion of 180 minutes running and 125W threshold, normalized value 0.83). Finally, the component outputs a 64-dimensional injection benchmark value feature vector, which completely quantifies the performance requirements of large injection molds on cooling fans and can be directly used for matching with the MF-800A fan feature vector.

[0166] In the embodiments of the present application, after determining the at least one performance benchmark value matched by the first cooling fan according to the performance matching degree between the feature vector corresponding to the first cooling fan and the feature vector corresponding to each of the at least one alternative performance benchmark value, the following implementation is further provided.

[0167] According to the at least one performance benchmark value matched by the first cooling fan, at least one adaptive test working condition is determined from the pool of test working conditions corresponding to the performance evaluation system.

[0168] The adaptive test working condition is sent to the monitoring system corresponding to the first cooling fan.

[0169] In an embodiment of the application, in an industrial mold production scenario, the server determines that the performance benchmark value matched by the first cooling fan (MF-800A model, suitable for large injection molds) is "large injection mold cooling system standard performance benchmark value", and then needs to select the test working condition that matches from the "test working condition condition pool" corresponding to the performance evaluation system. The working condition pool stores the standardized test working conditions of the industrial mold cooling system, and each working condition is associated with a performance benchmark value label (such as "large injection mold-full load" and "precision stamping mold-half load") for quickly locating the matching scenario. The server queries the metadata database of the working condition pool (stored using MongoDB, with the index field being "benchmark value ID"), matches the working condition cluster corresponding to the "large injection mold cooling system standard performance benchmark value", and the cluster includes three types of core test working conditions: full-load continuous operation working condition (simulating mold 24-hour continuous production), temperature gradient fluctuation working condition (simulating the change of ambient temperature caused by day and night temperature difference), and load sudden change impact working condition (simulating the load mutation when the mold production starts and stops), each working condition includes detailed parameter combinations (ambient temperature, load power, cavity pressure, test duration) and industrial test standard numbers (such as "TEST-001-FULL").

[0170] During specific screening, the server extracts the working conditions with a parameter matching degree ≥90% in the working condition pool based on the reference indicators of the benchmark value (55°C high temperature, 120W load, 8MPa cavity pressure): the full-load continuous operation working condition parameters are "ambient temperature 55°C±2°C, load power 120W±5W, cavity pressure 8MPa±0.2MPa, test duration 180 minutes", which matches the full-load requirement of the benchmark value with a matching degree of 98%; the temperature gradient fluctuation working condition parameters are "ambient temperature 35°C→45°C→55°C (increasing by 10°C per hour), load power 100W→120W (synchronously increasing), cavity pressure 5MPa→8MPa (synchronously increasing), test duration 300 minutes", which covers the transition scenario from preheating to full load of the benchmark value, with a matching degree of 92%; and the load sudden change impact working condition parameters are "load power 80W→120W (sudden change within 10 seconds), ambient temperature 55°C, cavity pressure 8MPa, test duration 60 minutes", which simulates the extreme load after the mold is urgently stopped and restarted, with a matching degree of 95%. The three types of working conditions are all certified by the industrial mold test specification (standard number "GB / T28884-2023"), ensuring that the test results can be directly used for fan performance evaluation.

[0171] Subsequently, the server encapsulates the selected suitable test conditions (including condition name, parameter combination, test duration, data acquisition frequency (1 minute / time), and safety threshold (e.g., automatic shutdown when noise ≥ 75dB)) into JSON format data and sends it via MQTT protocol to the "Mold Heat Dissipation Monitoring System" corresponding to the MF-800A fan (deployed on the edge server of the mold production line, IP address 192.168.1.100, port 1883). After receiving the data, the monitoring system parses the condition parameters and generates a test task sheet, controlling industrial testing equipment (such as constant temperature environmental chamber, programmable load cell, pressure simulator) to execute the above conditions in sequence, synchronously collecting real-time data such as fan heat dissipation efficiency, noise, and power consumption to verify whether the MF-800A meets the actual cooling requirements of large injection molds.

[0172] It is worth noting that this solution does not conduct "full-condition testing." Instead, it uses partial pre-test data of the fan (such as preliminary test data from 3-5 typical operating conditions) to recommend precise and suitable test conditions (such as recommending only the "55℃ high temperature + 120W load" condition applicable to the fan) through feature extraction and benchmark matching. In other words, it uses "small data + feature matching" to achieve "seeing the big picture from a small sample," reducing redundant testing and improving testing efficiency. At the same time, the dynamic operating condition spectrum can capture coupling features that traditional static testing cannot cover, and more accurately reflect the actual industrial performance of the fan.

[0173] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 performs the aforementioned cooling fan performance testing method. Figure 2 As shown, Figure 2 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0174] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. A method of testing performance of a heat dissipating fan, characterized by, The method comprises: obtaining a multi-working condition test association graph structure required to be called by a performance evaluation system; wherein the performance evaluation system is used to provide a test working condition adaptation service for a cooling fan, the multi-working condition test association graph structure comprises: at least one fan entity corresponding to each of the cooling fans, at least one working condition entity corresponding to each of the test working conditions, and a first test response association between the fan entity and the working condition entity, the first test response association is used to reflect a process coupling effect between the cooling fan corresponding to the fan entity and the test working condition corresponding to the working condition entity; generating a feature vector corresponding to a first cooling fan according to a sector coupling topology corresponding to the first cooling fan in the multi-working condition test association graph structure; wherein the sector coupling topology corresponding to the first cooling fan comprises: a first fan entity corresponding to the first cooling fan, and at least one working condition entity having a first test response association with the first fan entity; the feature vector corresponding to the first cooling fan is used to reflect performance characterization data of the first cooling fan; determining at least one performance benchmark value matched by the first cooling fan according to a performance matching degree between the feature vector corresponding to the first cooling fan and a feature vector corresponding to each of at least one alternative performance benchmark value; wherein the feature vector corresponding to the performance benchmark value is used to reflect a performance characterization result of the performance benchmark value.

2. The method of claim 1, wherein, The generating of the feature vector corresponding to the first cooling fan according to the sector coupling topology corresponding to the first cooling fan in the multi-working condition test association graph structure comprises: for each of the at least one working condition entity having the first test response association with the first fan entity, determining a time sequence position of the working condition entity in the working condition time sequence dynamic flow according to a data acquisition entity recorded in data of the first test response association between the working condition entity and the first fan entity; sorting the at least one working condition entity having the first test response association with the first fan entity according to the respective time sequence positions to generate the working condition time sequence dynamic flow; determining parameter acquisition data corresponding to each of the at least one working condition entity included in the working condition time sequence dynamic flow; wherein the parameter acquisition data corresponding to the working condition entity is used to indicate a test working condition corresponding to the working condition entity; inputting the parameter acquisition data corresponding to each of the at least one working condition entity into a cooling fan side feature extraction module; wherein the cooling fan side feature extraction module is a feature analysis model used to generate the feature vector corresponding to the first cooling fan; performing feature extraction processing on the parameter acquisition data corresponding to each of the at least one working condition entity by a working condition coding component of the cooling fan side feature extraction module to obtain a dynamic working condition spectrum of the working condition time sequence dynamic flow. The feature transformation component in the heat dissipation fan side feature extraction module performs feature extraction processing on the dynamic working condition spectrum of the working condition time sequence dynamic flow, to obtain a feature vector corresponding to the working condition time sequence dynamic flow; wherein the feature vector corresponding to the working condition time sequence dynamic flow is used to reflect performance characterization data of at least one test working condition condition of the process coupling effect of the first heat dissipation fan; According to the feature vector corresponding to the working condition time sequence dynamic flow, a feature vector corresponding to the first heat dissipation fan is generated.

3. The method of claim 2, wherein, The sector coupling topology further comprises: at least one working condition performance index entity respectively connected with at least one working condition condition entity in the sector coupling topology through a second test response association; wherein the at least one working condition condition entity in the sector coupling topology refers to a working condition condition entity connected with the first fan entity through the first test response association, and the second test response association is used to reflect that the test working condition condition corresponding to the working condition condition entity has the working condition performance index corresponding to the working condition performance index entity; the method further comprises: For a first working condition performance index type in the at least one working condition performance index type, according to the second test response association in the sector coupling topology, at least one first working condition performance index entity corresponding to each working condition condition entity in the working condition time sequence dynamic flow in the first working condition performance index type is determined; According to the at least one first working condition performance index entity corresponding to each working condition condition entity in the working condition time sequence dynamic flow, a working condition performance index set corresponding to the first working condition performance index type is generated; wherein the working condition performance index set includes a working condition performance index entity having a second test response association with at least one working condition condition entity in the working condition time sequence dynamic flow; For a first working condition performance index type in the at least one working condition performance index type, the parameter acquisition data corresponding to each working condition performance index entity included in the working condition performance index set corresponding to the first working condition performance index type is determined; wherein the parameter acquisition data corresponding to each working condition performance index entity is used to indicate the working condition performance index corresponding to the working condition performance index entity; Through the working condition performance index feature mapping component corresponding to the first working condition performance index type in the heat dissipation fan side feature extraction module, the parameter acquisition data corresponding to each working condition performance index entity included in the working condition performance index set corresponding to the first working condition performance index type is subjected to feature extraction processing, to obtain a feature vector corresponding to the first working condition performance index type; wherein the heat dissipation fan side feature extraction module is used to generate a feature vector corresponding to the first heat dissipation fan; Through the working condition performance index feature aggregator in the heat dissipation fan side feature extraction module, the feature vectors corresponding to the at least one working condition performance index type are subjected to feature extraction, to obtain a feature vector corresponding to the working condition performance index; The feature vector corresponding to the working condition time sequence dynamic flow is generated according to the feature vector corresponding to the working condition time sequence dynamic flow, including: The feature fusion component of the heat dissipation fan feature extraction module fuses the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the working condition performance index, to obtain a first fused feature vector; The feature generation component in the heat dissipation fan side feature extraction module performs feature extraction on the first fused feature vector, to obtain a feature vector corresponding to the first heat dissipation fan.

4. The method of claim 2, wherein, The sector-coupled topology further includes at least one heat dissipation fan characteristic entity having a third test response association with the first fan entity; wherein the third test response association reflects that the heat dissipation fan corresponding to the fan entity has the heat dissipation fan characteristic corresponding to the heat dissipation fan characteristic entity; the method further includes: According to the sector-coupled topology, generating at least one heat dissipation fan characteristic entity corresponding to a respective heat dissipation fan characteristic dimension; wherein the set of heat dissipation fan characteristics includes at least one working condition performance index entity having a third test response association with the first fan entity; According to the set of heat dissipation fan characteristics corresponding to at least one respective heat dissipation fan characteristic dimension, generating a feature vector corresponding to the heat dissipation fan characteristic; The generation of the feature vector corresponding to the first heat dissipation fan according to the feature vector corresponding to the working condition time sequence dynamic flow includes: According to the feature vector corresponding to the working condition time sequence dynamic flow and the feature vector corresponding to the heat dissipation fan characteristic, generating the feature vector corresponding to the first heat dissipation fan.

5. The method of claim 4, wherein, The generation of the feature vector corresponding to the heat dissipation fan characteristic according to the set of heat dissipation fan characteristics corresponding to at least one respective heat dissipation fan characteristic dimension includes: For a first heat dissipation fan characteristic dimension in the at least one heat dissipation fan characteristic dimension, determining parameter acquisition data corresponding to at least one heat dissipation fan characteristic entity included in the set of heat dissipation fan characteristics corresponding to the first heat dissipation fan characteristic dimension; the parameter acquisition data corresponding to the heat dissipation fan characteristic entity is used to indicate the heat dissipation fan characteristic corresponding to the heat dissipation fan characteristic entity; Through the heat dissipation fan characteristic feature mapping component corresponding to the first heat dissipation fan characteristic dimension in the heat dissipation fan side feature extraction module, performing feature extraction processing on the identity of the at least one heat dissipation fan characteristic entity, to obtain an embedding vector of the first heat dissipation fan characteristic dimension; wherein the heat dissipation fan side feature extraction module is used to generate a feature vector corresponding to the first heat dissipation fan characteristic dimension; Through the heat dissipation fan characteristic feature aggregator in the heat dissipation fan side feature extraction module, performing feature extraction on the feature vector corresponding to each of the at least one heat dissipation fan characteristic dimension, to obtain the feature vector corresponding to the heat dissipation fan characteristic.

6. The method of claim 1, wherein, The determination of at least one performance benchmark value matched by the first heat dissipation fan according to the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to each of the at least one alternative performance benchmark value includes: Determining a feature vector corresponding to each of the at least one alternative performance benchmark value; For a first performance benchmark value in the at least one alternative performance benchmark value, calculate a feature space approximation degree between the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value, and determine a performance matching degree of the feature vector corresponding to the first heat dissipation fan and the feature vector of the first performance benchmark value; From the at least one alternative performance benchmark value, select an alternative performance benchmark value with a performance matching degree not lower than a performance matching degree threshold as the first heat dissipation fan matched performance benchmark value.

7. The method of claim 6, wherein, The determining of the feature vector corresponding to each of the at least one alternative performance benchmark value comprises: For a first performance benchmark value in the at least one alternative performance benchmark value, obtain at least one reference working condition performance index corresponding to the first performance benchmark value; wherein the reference working condition performance index has performance equivalence with the first performance benchmark value; According to the first performance benchmark value and the at least one reference working condition performance index corresponding to the first performance benchmark value, determine the feature vector corresponding to the first performance benchmark value.

8. The method of claim 7, wherein, The determining of the feature vector corresponding to the first performance benchmark value according to the first performance benchmark value and the at least one reference working condition performance index corresponding to the first performance benchmark value comprises: Through a performance benchmark value feature mapping component of a working condition side feature extraction module, perform feature extraction processing on parameter collection data corresponding to each of at least one word included in the first performance benchmark value, to obtain parameter representation corresponding to the first performance benchmark value; wherein the working condition side feature extraction module is used to determine the feature vector corresponding to the first performance benchmark value; Through a working condition parameter mapping unit of the working condition side feature extraction module, perform feature extraction processing on parameter collection data corresponding to each of at least one word included in the at least one working condition performance index, to obtain parameter representation corresponding to the working condition performance index; Through a feature fusion component of the working condition side feature extraction module, perform fusion processing on the parameter representation corresponding to the first performance benchmark value and the parameter representation corresponding to the working condition performance index, to obtain performance fusion representation; Through a feature generation component of the working condition side feature extraction module, perform feature extraction processing on the performance fusion representation, to determine the feature vector corresponding to the first performance benchmark value.

9. The method of claim 1, wherein, After the determining of the at least one performance benchmark value matched by the first heat dissipation fan according to the performance matching degree between the feature vector corresponding to the first heat dissipation fan and the feature vector corresponding to each of the at least one alternative performance benchmark value, further comprising: According to the at least one performance benchmark value matched by the first heat dissipation fan, determine at least one adaptive test working condition from a test working condition pool corresponding to the performance evaluation system; Send the adaptive test working condition to a monitoring system corresponding to the first heat dissipation fan.

10. A server system, characterized by The server is configured to execute the method of any one of claims 1-9.

Images