CN116955933B

CN116955933B - Variable flow microchannel heat exchanger performance detection method, device and equipment

Info

Publication number: CN116955933B
Application number: CN202311220732.7A
Authority: CN
Inventors: 林明钦; 贺前程; 罗晓平
Original assignee: Shenzhen Friend Heat Sink Technology Co ltd
Current assignee: Shenzhen Friend Heat Sink Technology Co ltd
Priority date: 2023-09-21
Filing date: 2023-09-21
Publication date: 2024-01-09
Anticipated expiration: 2043-09-21
Also published as: CN116955933A

Abstract

The invention provides a variable flow microchannel heat exchanger performance detection method, a device and equipment, which are characterized in that a data set description vector of a heat exchange data set of a heat exchanger to be detected is obtained and is determined to be a first data set description vector, the first data set description vector is regulated to realize data cleaning, each round of regulated cleaning is at least combined with the first data set description vector, and a second data set description vector obtained after data cleaning can keep an initial heat exchange performance state in the heat exchange data set of the heat exchanger to be detected, so that deviation of performance identification caused by changing the heat exchange performance state during data cleaning is avoided. When the adjustment is carried out for a plurality of times, the data cleaning results of each adjustment are combined in the later adjustment, so that the data cleaning quality can be improved, and the reliability of identifying the heat exchange performance state of the target description vector is further improved. And classifying the heat exchange performance states by adjusting the description vector of the cleaned second data set to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

Description

Variable flow microchannel heat exchanger performance detection method, device and equipment

Technical Field

The disclosure relates to the technical field of data processing and artificial intelligence, and in particular relates to a method, a device and equipment for detecting performance of a variable flow microchannel heat exchanger.

Background

The variable flow microchannel heat exchanger (Variable Flow Microchannel Heat Exchanger) is a device for heat transfer that employs a microchannel structure to achieve efficient heat exchange. It consists of a plurality of tiny channels through which heat is transferred from one fluid to another, thereby effecting the transfer of thermal energy. The variable flow microchannel heat exchanger is characterized in that the flow rate of the fluid can be adjusted according to the requirement. By varying the use of the different channels, a non-uniform flow distribution can be achieved, thereby optimizing the heat transfer effect. The flexible design enables the heat exchanger to be adjusted according to actual application requirements, and improves heat exchange efficiency and performance. In many application fields, the variable flow microchannel heat exchanger has potential application prospects, such as electronic equipment heat dissipation, heat exchange in chemical process, renewable energy systems and the like. Detecting the heat exchange performance of the variable flow microchannel heat exchanger is an important link for ensuring the heat exchange effect, in the related art, the detection of the heat exchange performance is mainly performed by analyzing the heat exchange data of the variable flow microchannel heat exchanger, such as the fluid inlet temperature, the fluid outlet temperature, the fluid flow, the pressure, the wall surface temperature and other data, by adopting an artificial intelligent auxiliary tool, so as to improve the detection efficiency and the accuracy, and then how to improve the accuracy of the detection of the heat exchange performance is a technical problem which needs to be continuously improved.

Disclosure of Invention

Accordingly, embodiments of the present disclosure provide at least a method, an apparatus, and a device for detecting performance of a variable flow microchannel heat exchanger.

The technical scheme of the embodiment of the disclosure is realized as follows:

in one aspect, an embodiment of the present disclosure provides a method for detecting performance of a variable flow microchannel heat exchanger, applied to a computer device, the method comprising:

acquiring a data set description vector of a heat exchange data set of a heat exchanger to be detected, and taking the data set description vector as a first data set description vector;

performing a plurality of rounds of adjustment on the first dataset description vector, one round of adjustment comprising: determining a description vector to be cleaned based on the description vector of the first data set, obtaining a cleaning intermediate state corresponding to the current round of adjustment according to the historical description vector to be cleaned obtained from the heat exchange data set of the historical heat exchanger to be detected, and cleaning the description vector to be cleaned according to the cleaning intermediate state to obtain an intermediate description vector corresponding to the current round of adjustment;

in the first round of adjustment, the description vector to be cleaned is the description vector of the first data set, in the X-th round of adjustment, the description vector to be cleaned is obtained by aggregating the description vector of the first data set and the obtained intermediate description vector, and X is a positive integer;

Determining an intermediate description vector obtained by final wheel adjustment into a target description vector, and cleaning the first data set description vector according to target data cleaning parameters obtained by mining the target description vector to obtain a second data set description vector;

and classifying the heat exchange performance states through the description vector of the second data set to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

In some embodiments, for one round of adjustment:

when the first round of adjustment is performed, determining the description vector of the first data set into a description vector to be cleaned, obtaining a corresponding cleaning intermediate state according to a corresponding historical description vector to be cleaned, which is obtained from a heat exchange data set of a historical heat exchanger to be detected, and cleaning the description vector to be cleaned according to the cleaning intermediate state to obtain an intermediate description vector corresponding to the first round of adjustment;

when the second round of adjustment is performed, the description vector of the first data set and an intermediate description vector obtained by the previous round of adjustment of the first round of adjustment are polymerized into a description vector to be cleaned, a corresponding cleaning intermediate state is obtained according to the corresponding historical description vector to be cleaned obtained for the historical heat exchange data set to be detected of the heat exchanger, and the description vector to be cleaned is cleaned according to the cleaning intermediate state, so that the intermediate description vector corresponding to the second round of adjustment is obtained;

And when the adjustment is performed for the third round, the first data set description vector and the two intermediate description vectors obtained by the last two times of adjustment of the first round of adjustment are polymerized into a description vector to be cleaned, a corresponding cleaning intermediate state is obtained according to the corresponding historical description vector to be cleaned obtained for the heat exchange data set of the historical heat exchanger to be detected, and the description vector to be cleaned is cleaned according to the cleaning intermediate state, so that the intermediate description vector corresponding to the third round of adjustment is obtained.

In some embodiments, the aggregation process of the description vectors to be cleaned includes:

vector splicing is carried out on each obtained intermediate description vector and the first data set description vector through a preset splicing order to obtain a description vector to be cleaned; or alternatively; carrying out weight balance combination on each intermediate description vector and the first data set description vector through the aggregation adjustment parameters corresponding to each obtained intermediate description vector and the aggregation adjustment parameters corresponding to the first data set description vector, so as to obtain a description vector to be cleaned; wherein, each aggregation adjustment parameter is used for describing the importance of the corresponding intermediate description vector or the first data set description vector to the description vector to be cleaned;

The target data cleaning parameters obtained by mining the target description vector clean the first data set description vector to obtain a second data set description vector, including:

performing dimension reduction operation on the target description vector to obtain a temporary description vector, wherein the number of vector elements of the temporary description vector corresponds to that of the first data set description vector;

carrying out numerical interval normalization on the temporary description vector to obtain a normalization vector, and determining the normalization vector as a target data cleaning parameter;

and calculating an inner product of the target data cleaning parameter and the first data set description vector to obtain a second data set description vector.

In some embodiments, the classifying the heat exchange performance states according to the description vector of the second dataset to obtain the heat exchange performance states of the heat exchange dataset of the heat exchanger to be detected includes:

carrying out data set description vector reasoning on the second data set description vector to obtain a classification support coefficient containing a heat exchange performance state in the second data set description vector;

and if the classification support coefficient is larger than a preset heat exchange performance state classification reference support coefficient, determining that the heat exchange data set of the heat exchanger to be detected contains the corresponding heat exchange performance state.

In some embodiments, the method is implemented according to a trained heat exchange performance state detection algorithm that is trained by:

obtaining a set of heat exchanger data samples, each heat exchanger data sample in the set of heat exchanger data samples comprising: a first heat exchange performance sample description vector extracted from a heat exchange performance data sample to be cleaned and a corresponding sample annotation, wherein the sample annotation comprises a sample heat exchange performance state mark, and the sample heat exchange performance state mark is used for describing whether the heat exchange performance state corresponds to the corresponding heat exchange performance data sample to be cleaned;

determining a heat exchanger data sample in the heat exchanger data sample set, and inputting a corresponding first heat exchange performance sample description vector into the heat exchange performance state detection algorithm to obtain an inferred heat exchange performance state mark for classifying the heat exchange performance state;

and correcting the algorithm learnable variable in the heat exchange performance state detection algorithm through the loss between the inferred heat exchange performance state mark and the corresponding sample heat exchange performance state mark.

In some embodiments, the algorithmically learnable variables include a first learnable variable for purging and a second learnable variable for supporting coefficient determination; the obtaining the inferred heat exchange performance state label for classifying the heat exchange performance state comprises the following steps:

Cleaning the first heat exchange performance sample description vector through the first learnable variable to obtain a corresponding second heat exchange performance sample description vector;

carrying out data set description vector reasoning on the second heat exchange performance sample description vector through the second learnable variable to obtain a classification support coefficient supporting the heat exchange performance state, and comparing the classification support coefficient with a preset heat exchange performance state classification reference support coefficient to obtain a reasoning heat exchange performance state mark;

the sample annotation also comprises pure heat exchange performance sample description vectors extracted from the corresponding data samples of the heat exchanger to be cleaned, wherein the pure heat exchange performance sample description vectors are data set description vectors which do not need cleaning, and the algorithm learnable variables comprise a first learnable variable used for cleaning and a second learnable variable used for supporting coefficient determination; the correcting the algorithm learning variable in the heat exchange performance state detection algorithm through the loss between the inferred heat exchange performance state mark and the corresponding sample heat exchange performance state mark comprises the following steps:

optimizing the first learnable variable by a first loss between the second heat exchange performance sample description vector and the corresponding clean heat exchange performance sample description vector;

Acquiring a second loss between the inferred heat exchange performance state mark and the corresponding sample heat exchange performance state mark, and fusing the first loss and the second loss to obtain a fused loss;

optimizing the first and second learnable variables by the fusion penalty.

acquiring a heat exchange data set and a heat exchange performance state relation network of the heat exchanger to be detected;

inputting the heat exchange data set of the heat exchanger to be detected into a heat exchange characteristic recognition algorithm to perform heat exchange characteristic recognition, and obtaining data cluster description vectors of a plurality of heat exchange data clusters of the heat exchange data set of the heat exchanger to be detected;

extracting a description vector of the heat exchange performance state relation network through a description vector extraction module to obtain a heat exchange performance description vector of the heat exchange performance state relation network, wherein the heat exchange performance description vector represents a state derivative relation of the heat exchange performance state relation network;

performing description vector aggregation on the heat exchange performance description vector, the data cluster description vector and the second data set description vector to obtain a target aggregation vector;

And inputting the target aggregate vector into a heat exchange performance state classification algorithm to classify the heat exchange performance states, and obtaining the heat exchange performance states corresponding to the heat exchange data set of the heat exchanger to be detected.

In some embodiments, the description vector extraction module includes a plurality of description vector fusion units mapped to a plurality of nodes in the heat exchange performance state relation network, and the description vector extraction module performs description vector extraction on the heat exchange performance state relation network to obtain a heat exchange performance description vector of the heat exchange performance state relation network, including:

performing heat exchange characteristic identification on the heat exchange performance state of each network node in the plurality of network nodes to obtain a network node heat exchange description vector corresponding to each network node;

inputting a net heat exchange description vector corresponding to a terminal net of the heat exchange performance state relation net into a description vector fusion unit corresponding to the terminal net to obtain a net fusion description vector corresponding to the terminal net;

starting from an upper-level network knot of the terminal network knot, browsing the network knots, inputting a network knot fusion description vector corresponding to a lower-level network knot of the current processed network knot and a network knot heat exchange description vector corresponding to the current processed network knot into a description vector fusion unit corresponding to the current processed network knot, and obtaining a network knot fusion description vector corresponding to the current processed network knot;

After the network nodes are browsed, determining a network node fusion description vector corresponding to the network node at the top end in the heat exchange performance state relation network as the heat exchange performance description vector;

the heat exchange characteristic recognition algorithm comprises a heat exchange data clustering module, a heat exchange characteristic recognition module and a heat exchange characteristic fusion module, wherein the heat exchange characteristic recognition algorithm is used for carrying out heat exchange characteristic recognition on the heat exchange data set to be detected, so as to obtain data cluster description vectors of a plurality of heat exchange data clusters of the heat exchange data set to be detected, and the heat exchange characteristic recognition algorithm comprises the following steps:

inputting the heat exchange data set of the heat exchanger to be detected into the heat exchange data clustering module to perform data clustering to obtain a plurality of heat exchange data clusters;

inputting the plurality of heat exchange data clusters into the heat exchange characteristic recognition module for heat exchange characteristic recognition to obtain basic heat exchange characteristic vectors, data cluster position vectors and data cluster interval vectors corresponding to the plurality of heat exchange data clusters;

inputting the basic heat exchange feature vector, the data cluster position vector and the data cluster interval vector into the heat exchange feature fusion module to perform cross-correlation heat exchange feature fusion on the plurality of heat exchange data clusters, so as to obtain the data cluster description vector;

The performing description vector aggregation on the heat exchange performance description vector, the data cluster description vector and the second data set description vector to obtain a target aggregation vector comprises:

inputting the heat exchange performance description vector and the data cluster description vector into a description vector aggregation module to perform description vector aggregation to obtain a basic aggregation vector;

and inputting the basic aggregate vector, the second data set description vector and the heat exchange performance description vector into a vector splicing unit for vector splicing to obtain the target aggregate vector.

In a second aspect, the present disclosure provides a variable flow microchannel heat exchanger performance detection apparatus comprising:

the feature extraction module is used for acquiring a data set description vector of a heat exchange data set of the heat exchanger to be detected, and taking the data set description vector as a first data set description vector;

the characteristic adjustment module is used for carrying out multi-round adjustment on the description vector of the first data set, and one round of adjustment comprises: determining a description vector to be cleaned based on the description vector of the first data set, obtaining a cleaning intermediate state corresponding to the current round of adjustment according to the historical description vector to be cleaned obtained from the heat exchange data set of the historical heat exchanger to be detected, and cleaning the description vector to be cleaned according to the cleaning intermediate state to obtain an intermediate description vector corresponding to the current round of adjustment;

the data cleaning module is used for determining the intermediate description vector obtained by the final wheel adjustment into a target description vector, and cleaning the first data set description vector according to target data cleaning parameters obtained by mining the target description vector to obtain a second data set description vector;

and the performance detection module is used for classifying the heat exchange performance states through the description vector of the second data set and obtaining the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

In a third aspect, the present disclosure provides a computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing steps in the above-described method when the program is executed.

The present disclosure has at least the beneficial effects: according to the method, the data set description vector of the heat exchange data set of the heat exchanger to be detected is obtained, the data set description vector is determined to be a first data set description vector, the first data set description vector is regulated to achieve data cleaning, in one round of regulation, the description vector to be cleaned is determined based on the first data set description vector, the cleaning intermediate state corresponding to the round of regulation is obtained according to the historical description vector to be cleaned obtained for the historical heat exchange data set of the heat exchanger to be detected, the description vector to be cleaned is cleaned through the cleaning intermediate state, the intermediate description vector corresponding to the round of regulation is obtained, in the first round of regulation, the description vector to be cleaned is the first data set description vector, in the X round of regulation, X is a positive integer. The cleaning process of each round of adjustment is at least combined with a first data set description vector (namely an initial data set description vector to be cleaned), and after data cleaning, an obtained second data set description vector can retain an initial heat exchange performance state in heat exchange data sets of the heat exchanger to be detected, so that performance recognition deviation caused by changing the heat exchange performance state during data cleaning is avoided. Further, when the adjustment is performed for multiple times, the data cleaning results (namely the intermediate description vectors) of each historical adjustment are combined in the latter adjustment, so that the quality of data cleaning can be improved, and the reliability of identifying the heat exchange performance state of the target description vectors is further improved. And classifying the heat exchange performance states by adjusting the description vector of the cleaned second data set to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the aspects of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the technical aspects of the disclosure.

Fig. 1 is a schematic implementation flow chart of a method for detecting performance of a variable flow microchannel heat exchanger according to an embodiment of the disclosure.

Fig. 2 is a schematic diagram of a composition structure of a performance detection device of a variable flow microchannel heat exchanger according to an embodiment of the disclosure.

Fig. 3 is a schematic hardware entity diagram of a computer device according to an embodiment of the disclosure.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present disclosure more apparent, the technical solutions of the present disclosure are further elaborated below in conjunction with the drawings and the embodiments, and the described embodiments should not be construed as limiting the present disclosure, and all other embodiments obtained by those skilled in the art without making inventive efforts are within the scope of protection of the present disclosure.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. The term "first/second/third" is merely to distinguish similar objects and does not represent a particular ordering of objects, it being understood that the "first/second/third" may be interchanged with a particular order or precedence where allowed, to enable embodiments of the disclosure described herein to be implemented in other than those illustrated or described herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing the present disclosure only and is not intended to be limiting of the present disclosure.

Embodiments of the present disclosure provide a variable flow microchannel heat exchanger performance detection method that may be performed by a processor of a computer device. The computer device may refer to a server, a notebook computer, a tablet computer, a desktop computer, an intelligent television, a mobile device, and other devices with data processing capability.

Fig. 1 is a schematic implementation flow chart of a method for detecting performance of a variable flow microchannel heat exchanger according to an embodiment of the disclosure, as shown in fig. 1, where the method includes the following operations:

operation 100, a dataset description vector of a heat exchange dataset of a heat exchanger to be detected is obtained and determined as a first dataset description vector.

The heat exchange data set of the heat exchanger to be detected can be working data of the heat exchanger such as fluid inlet temperature, fluid outlet temperature, fluid flow, pressure, wall temperature, heat conductivity and the like of the variable flow microchannel heat exchanger (Variable Flow Microchannel Heat Exchanger) acquired based on the internet of things technology, and the working data are obtained by arrangement according to preset data arrangement rules (without limitation). The feature extraction can be performed on the heat exchange data set of the heat exchanger to be detected based on the convolutional neural network, so as to obtain a corresponding data set description vector, the data set description vector is determined to be a first data set description vector corresponding to the heat exchange data set of the heat exchanger to be detected, and the description vector is a vector describing feature information (such as temperature feature, flow feature, pressure feature and the like) of the heat exchange data set of the heat exchanger to be detected.

And 200, performing multi-round adjustment on the description vector of the first data set to obtain an intermediate description vector corresponding to each round of adjustment.

In the embodiment of the disclosure, the intermediate description vector is a description vector obtained after one round of cleaning is performed on the description vector of the first data set, and it can be understood that the description vector of the first data set represents a description vector to be cleaned corresponding to the heat exchange data set of the heat exchanger to be detected, and the intermediate description vector represents a cleaning description vector corresponding to the heat exchange data set of the heat exchanger to be detected. The data cleaning may include processes such as error value processing, missing value processing, duplicate value processing, noise reduction, logical consistency check, outlier processing, etc., and since the process is in the prior art, the process is not developed here, and an object of the embodiments of the present disclosure is to alleviate an error in subsequent state recognition caused by excessive cleaning in the data cleaning process.

Specifically, the adjusting process is a process of performing data cleaning on a first data set description vector, the first data set description vector is determined to be a description vector to be cleaned during the first round of adjustment, the first data set description vector and an acquired intermediate description vector are polymerized to be a description vector to be cleaned during the X round of adjustment process, a cleaning intermediate state corresponding to the round of adjustment is obtained through a history description vector set to be cleaned, which is acquired by heat exchange data sets of a history heat exchanger to be detected, the description vector to be cleaned is cleaned through the cleaning intermediate state, the intermediate description vector corresponding to the round of adjustment is obtained, and X is a positive integer. One round of adjustment is performed by adopting any feasible deep neural network, for example, a plurality of data clusters (for example, a plurality of data clusters divided according to acquisition time or a plurality of data clusters divided according to acquisition data types) which are sequentially arranged can be sequentially input into the gating network based on the gating network, the state of the gating network is sequentially expanded, the output of the next round is determined through the output of the previous round, and the gating network can deploy a classification module to obtain an inference result through classification.

Let the input of this round be p, the result of the upper round be q, in one round of adjustment, four intermediate results are obtained, respectively being the first result r ₁ Second result r ₂ Third result r ₃ And fourth result r ₄ Can be specifically based onThe following method is adopted for acquisition:

r ₁ =Relu（S _r11 ·p+S _r12 ·q+M _r1 ）

r ₂ =tanh（S _r21 ·p+S _r22 ·q+M _r2 ）

r ₃ =Relu（S _r31 ·p+S _r32 ·q+M _r3 ）

r ₄ =Relu（S _r41 ·p+S _r42 ·q+M _r4 ）

wherein S is _r11 、S _r12 、S _r21 、S _r22 、S _r31 、S _r32 、S _r41 、S _r42 Weights, M, respectively obtained for one round of adjustment _r1 、M _r2 、M _r3 、M _r4 Respectively, the bias to achieve deployment.

Based on the above results, the cleaning intermediate state (intermediate state of the gate control network) of the present round can be obtained, and if g is set as g, g=r ₃ ·g _ex +r ₁ ·r ₂ Wherein g _ex The intermediate state of the wash obtained for the previous round.

Then the reasoning result Z of this round is z=r ₄ ·tanh（g）。

In general, each round of adjustment is performed on a data cluster, in addition, input data can be a description vector to be cleaned, a history description vector set to be cleaned is obtained by performing heat exchange on a history heat exchange data set to be detected on a heat exchanger, an intermediate cleaning state corresponding to the round of adjustment is obtained, then an intermediate description vector corresponding to the round of adjustment is obtained based on new input and the combination of the information, and next round of required information is transmitted forward, so that the description vector of the data set to be cleaned next time is cleaned conveniently.

The method and the device have the advantages that the description vector of the first data set needs to be subjected to multi-round adjustment, for example, if two rounds of adjustment are performed, the hardware cost of adjustment is low, and the calculation force dependence is low; if the cleaning quality is increased, three rounds of adjustment can be performed, and the hardware cost is increased, but the cleaning is not complicated when each round of adjustment is performed, and the excessive hardware environment is not needed. Then, in the adjustment process, three rounds of adjustment are exemplified:

In the first round of adjustment, the description vector of the first data set is determined to be a description vector to be cleaned, a corresponding cleaning intermediate state is obtained according to the corresponding historical description vector to be cleaned, which is obtained for the heat exchange data set of the historical heat exchanger to be detected, and the description vector to be cleaned is cleaned according to the cleaning intermediate state, so that an intermediate description vector corresponding to the first round of adjustment is obtained. In the data cleaning process of performing first round adjustment on the first data set description vector, a cleaning intermediate state corresponding to the heat exchange data set of the historical heat exchanger to be detected is obtained, the cleaning intermediate state is taken as input together with the description vector of the heat exchange data set to be cleaned corresponding to the heat exchange data set of the current heat exchanger to be detected, and the gating network is enabled to clean the data set to obtain an output cleaning intermediate state corresponding to the heat exchange data set of the current heat exchanger to be detected and an intermediate description vector corresponding to the current round adjustment. Before cleaning, the first data set description vector may be subjected to a dimension reduction operation, for example, the number of constituent elements (i.e., dimensions) of the first data set description vector is transformed by a first full-connection layer, the first data set description vector is processed based on a preset nonlinear function (such as tanh), and the first full-connection layer is used for reducing the dimensions, and the nonlinear function is used for compressing the data values. And adding a first full-connection layer and a nonlinear function, and integrating the description vector to be cleaned obtained after the first data set description vector is in the full-connection layer in the description vector to be cleaned adjusted in the next round.

In the second round of adjustment, the description vector of the first data set and an intermediate description vector obtained by one round of adjustment before the current round of adjustment are polymerized into a description vector to be cleaned (such as vector splicing, adding, connecting and the like, and are polymerized), corresponding cleaning intermediate states are obtained according to the corresponding historical description vector to be cleaned obtained by the heat exchange data set of the historical heat exchanger to be detected, and the description vector to be cleaned is cleaned according to the cleaning intermediate states, so that the intermediate description vector corresponding to the second round of adjustment is obtained.

In the third round of adjustment, the first data set description vector and two intermediate description vectors obtained by two times of adjustment before the first round of adjustment (one intermediate description vector obtained by the first round of adjustment and one intermediate description vector obtained by the second round of adjustment) are polymerized into a description vector to be cleaned, corresponding cleaning intermediate states are obtained according to the corresponding historical description vector to be cleaned obtained by the historical heat exchange data set to be detected, and the description vector to be cleaned is cleaned according to the cleaning intermediate states, so that the intermediate description vector corresponding to the third round of adjustment is obtained.

The X-th round of adjustment after the first round of adjustment is to aggregate each intermediate description vector and the first data set description vector which are obtained by each adjustment before the present round of adjustment into a description vector to be cleaned. In the data cleaning of the X-th round of adjustment of the description vector of the first data set, the difference between the description vector to be cleaned and the first round of adjustment is that the description vector of the first data set corresponding to the heat exchange data set of the heat exchanger to be detected is obtained by aggregating the description vector of the first data set corresponding to the heat exchange data set to be detected and the intermediate description vector obtained before the adjustment of the first round.

And for the mode of aggregating the description vectors, the acquired intermediate description vectors and the first data set description vectors are converted into description vectors to be cleaned with better characterization effects through description vector integration, so that the description vectors to be cleaned are cleaned and simultaneously combined with the first data set description vectors and the previous data cleaning results of each round to judge, the reliability of data cleaning is improved, and the classification of heat exchange performance states is facilitated, and the reliability is improved. The polymerization process is, for example: and vector stitching is carried out on each obtained intermediate description vector and the first data set description vector by a preset stitching bit number (namely, who is defined to be who is behind the former) to obtain the description vector to be cleaned. Or, performing weight balance combination on each intermediate description vector and the first data set description vector (namely, multiplying the aggregate adjustment parameter by the first data set description vector and then summing the results) through the aggregate adjustment parameters corresponding to each obtained intermediate description vector and the aggregate adjustment parameters corresponding to the first data set description vector, so as to obtain the description vector to be cleaned. Each aggregation adjustment parameter is used for describing the importance of the corresponding intermediate description vector or the first data set description vector to the description vector to be cleaned, i.e. the degree of influence of the two description vectors on each other.

The cleaning process of each round of adjustment is at least combined with a first data set description vector (namely an initial data set description vector to be cleaned), and after data cleaning, an obtained second data set description vector can keep an initial heat exchange performance state in a heat exchange data set of a heat exchanger to be detected, so that deviation of performance identification caused by changing the heat exchange performance state during data cleaning is avoided. Further, when the adjustment is performed for multiple times, the data cleaning results (namely the intermediate description vectors) of each historical adjustment are combined in the latter adjustment, so that the quality of data cleaning can be improved, and the reliability of identifying the heat exchange performance state of the target description vectors is further improved. And classifying the heat exchange performance states by adjusting the description vector of the cleaned second data set to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

And 300, determining the intermediate description vector obtained by the final wheel adjustment into a target description vector, and cleaning the first data set description vector according to the target data cleaning parameters obtained by mining the target description vector to obtain a second data set description vector.

The target data cleaning parameter is a mask outputted through cleaning, and the first data set description vector is cleaned through the target data cleaning parameter, so that a second data set description vector corresponding to the heat exchange data set of the heat exchanger to be detected, namely the data set description vector after data cleaning, is obtained.

When the target data cleaning parameters are obtained, firstly, performing dimension reduction operation on the target description vectors obtained through final wheel adjustment to obtain intermediate description vectors with the number of vector elements corresponding to that of the first data set description vectors, then performing numerical interval normalization on the intermediate description vectors to obtain normalization vectors, determining the normalization vectors into the target data cleaning parameters, and finally calculating the inner product of the target data cleaning parameters and the first data set description vectors to obtain a second data set description vector. And transforming the target description vector into a target data cleaning parameter, wherein the description vector can be processed based on a preset nonlinear function (such as sigmoid) through the description vector dimension reduction operation of the second full-connection layer transformation, the second full-connection layer is used for dimension reduction on the dimension, and the logic function is used for normalizing the value. The second data set description vector is a data set description vector obtained by cleaning the data obtained by carrying out various processes on the first data set description vector, and the heat exchange performance state information can be reserved as the information of the first data set description vector is integrated in the previous process.

And operation 400, classifying the heat exchange performance states through the second data set description vector to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

In the embodiment of the disclosure, data set description vector reasoning is performed on the second data set description vector to obtain a classification support coefficient containing a heat exchange performance state in the second data set description vector, and if the classification support coefficient is greater than a preset heat exchange performance state classification reference support coefficient, the heat exchange performance state of the heat exchanger to be detected is determined to be contained in the heat exchange data set. The support coefficient represents the probability of corresponding heat exchange performance state, and can be represented by probability. The heat exchange performance state is state information obtained by analyzing heat exchange performance, for example, classification of heat transfer effects, such as strong heat transfer performance, weak heat transfer performance and general heat transfer performance; or classification of fluid movement velocity states; or the state of resistance of the fluid in the microchannel (e.g., normal, too high resistance); or heat transfer performance state such as heat transfer efficiency of the heat exchanger.

For example, when classifying the heat exchange performance state, a support vector machine may be used to analyze the second data set description vector to obtain a classification support coefficient of the second data set description vector, where the classification support coefficient includes a corresponding heat exchange performance state, and the classification support coefficient is compared with a preset heat exchange performance state classification reference support coefficient, and when the classification support coefficient is greater than the preset heat exchange performance state classification reference support coefficient, the classification support coefficient indicates that the classification support coefficient is a corresponding heat exchange performance state, and it is determined that the heat exchange data set of the heat exchanger to be detected includes a corresponding heat exchange performance state.

As one implementation, the above example may be implemented based on a heat exchange performance state detection algorithm that may be trained using the following operations:

operation S101, a heat exchanger data sample set is acquired.

Each heat exchanger data sample in the set of heat exchanger data samples includes a first heat exchange performance sample description vector and a sample annotation drawn on a heat exchanger data sample to be cleaned, the sample annotation including a sample heat exchange performance status flag describing whether the heat exchange performance status corresponds to the corresponding heat exchanger data sample to be cleaned.

And S102, determining a heat exchanger data sample in the heat exchanger data sample set, and inputting a corresponding first heat exchange performance sample description vector into a heat exchange performance state detection algorithm to obtain an inferred heat exchange performance state mark for classifying the heat exchange performance state.

And S103, correcting the algorithm learnable variable in the heat exchange performance state detection algorithm by reasoning the loss between the heat exchange performance state marks and the corresponding sample heat exchange performance state marks.

The training process of the heat exchange performance state detection algorithm is one round of adjustment, and the aim of each round of adjustment is to optimize the variable of the algorithm, such as learning rate and super-parameters, and when the algorithm converges, the trained heat exchange performance state detection algorithm is obtained. The algorithm learnable variables specifically comprise a first learnable variable used for cleaning and a second learnable variable used for supporting coefficient determination, and the reasoning heat exchange performance state marks for classifying the heat exchange performance states can be obtained by cleaning the first heat exchange performance sample description vector through the first learnable variable to obtain a corresponding second heat exchange performance sample description vector; and carrying out data set description vector reasoning on the second heat exchange performance sample description vector through a second learnable variable to obtain a classification support coefficient supporting the heat exchange performance state, and comparing the classification support coefficient with a preset heat exchange performance state classification reference support coefficient to obtain a reasoning heat exchange performance state mark.

Each round of adjustment training of the heat exchange performance state detection algorithm comprises correction of a first variable and a second variable, wherein the correction mainly comprises obtaining a first loss value corresponding to cleaning according to a cleaning result and first loss between sample notes, and obtaining a second loss value corresponding to heat exchange performance state classification based on second loss between reasoning sensitive content and sample notes. The sample annotation also comprises pure heat exchange performance sample description vectors extracted from the corresponding data samples of the heat exchanger to be cleaned, wherein the pure heat exchange performance sample description vectors are data set description vectors which do not need cleaning, and the first learnable variable is corrected through first loss between the second heat exchange performance sample description vectors and the corresponding pure heat exchange performance sample description vectors; obtaining second losses between the inferred heat exchange performance state markers and the corresponding sample heat exchange performance state markers, and fusing (e.g., weighted summation or direct summation) the first losses and the second losses to obtain fusion losses; optimizing the first and second learnable variables by fusing the losses. Wherein the first loss may be calculated based on a mean square error and the second loss may be calculated based on a cross entropy.

As another embodiment, the heat exchange performance state classification is performed by the second data set description vector, so as to obtain the heat exchange performance state of the heat exchange data set of the heat exchanger to be detected, which may also be implemented by the following operations:

in operation S201, a heat exchange data set and a heat exchange performance state relation network of the heat exchanger to be detected are obtained.

The heat exchange performance state relation network is a structure formed (woven) based on the relation (such as subordinate relation and parallel relation) among various collected heat exchange performance state information, and can be also called as a heat exchange performance state tree structure. The heat exchange performance state relationship network comprises a plurality of knots, one corresponding to each heat exchange performance state, and one knots may have a derivative knot, if having a derivative knot, the knot is an upper knot or a parent knot, and the derivative knot is a lower knot or a child knot.

And S202, inputting the heat exchange data set of the heat exchanger to be detected into a heat exchange characteristic recognition algorithm to perform heat exchange characteristic recognition, and obtaining data cluster description vectors of a plurality of heat exchange data clusters of the heat exchange data set of the heat exchanger to be detected.

As described in operation 200, a round of adjustment is performed on a data cluster, which may be referred to as a heat exchange data cluster, which is obtained by clustering a heat exchange data set of a heat exchanger to be detected, for example, a data cluster that is partitioned based on a data acquisition period, or a data cluster that is partitioned according to a data type (such as different heat exchange indexes of a heat exchanger), a second data set description vector represents a cross-correlation feature (or a called context feature) of the heat exchange data set of the heat exchanger to be detected, a data cluster description vector represents a heat exchange performance feature of a corresponding heat exchange data cluster in the heat exchange data set of the heat exchanger to be detected, the heat exchange data set to be detected is input into a heat exchange feature recognition algorithm, and the heat exchange feature recognition algorithm may complete heat exchange feature recognition in combination with the context of the heat exchange data set of the heat exchanger to be detected, so as to obtain a data cluster description vector. The heat exchange characteristic recognition algorithm is obtained by optimizing heat exchange characteristic recognition of a preset heat exchange characteristic recognition algorithm, the preset heat exchange characteristic recognition algorithm comprises a heat exchange data clustering module, a heat exchange characteristic recognition module and a heat exchange characteristic fusion module, a heat exchange characteristic recognition algorithm is input into a heat exchange data set to be detected for heat exchange characteristic recognition, and data cluster description vectors of a plurality of heat exchange data clusters of a plurality of heat exchange data sets to be detected of the heat exchange data set to be detected are obtained, and the method specifically comprises the following steps:

In operation S2021, the heat exchange data set to be detected is input to the heat exchange data clustering module to perform data clustering, so as to obtain a plurality of heat exchange data clusters.

In operation S2022, the plurality of heat exchange data clusters are input to the heat exchange feature recognition module to perform heat exchange feature recognition, so as to obtain basic heat exchange feature vectors, data cluster position vectors and data cluster interval vectors corresponding to the plurality of heat exchange data clusters.

The basic heat exchange characteristic vector represents the characteristic of the corresponding heat exchange data cluster, the data cluster interval vector represents the characteristic of the corresponding heat exchange data cluster in the data interval (such as in which period or in which heat exchange data type) of the heat exchange data set of the heat exchanger to be detected, and the data cluster position vector represents the position characteristic of the corresponding heat exchange data cluster in the data area.

In operation S2023, the basic heat exchange feature vector, the data cluster position vector and the data cluster interval vector are input into a heat exchange feature fusion module to perform cross-correlation heat exchange feature fusion on the plurality of heat exchange data clusters, so as to obtain a data cluster description vector.

For example, the heat exchange feature fusion module may include a bidirectional encoding module, and input the basic heat exchange feature vector, the data cluster position vector and the data cluster interval vector into the bidirectional encoding module, and fuse (e.g. splice or add) the cross-correlation heat exchange features (i.e. the context features) of each heat exchange data cluster, so as to obtain the data cluster description vector of each heat exchange data cluster.

In operation S203, the description vector extraction module extracts the description vector of the heat exchange performance state relation network to obtain a heat exchange performance description vector of the heat exchange performance state relation network, where the heat exchange performance description vector represents a state derivative relation of the heat exchange performance state relation network.

The heat exchange performance description vector represents a state derivative relation of a heat exchange performance state relation network, the state derivative relation of the heat exchange performance state relation network comprises a structure of the heat exchange performance state relation network and heat exchange performance state information, and the hierarchy of the heat exchange performance state relation network is the association between the structure derivative relation of the heat exchange performance state and each heat exchange performance state. The description vector extraction module may be obtained by performing description vector extraction optimization on a preset description vector extraction module, where the preset description vector extraction module includes a plurality of description vector fusion units mapped (i.e. in one-to-one correspondence) with a plurality of network nodes in a preset heat exchange performance state relation network.

For example, the description vector extraction module performs description vector extraction on the heat exchange performance state relation network to obtain a heat exchange performance description vector of the heat exchange performance state relation network, including:

and (S2031) performing heat exchange characteristic identification on the heat exchange performance state of each of the plurality of knots to obtain a knots heat exchange description vector corresponding to each knot.

The mesh heat exchange description vector represents semantic information of the heat exchange performance state of the mesh, the description vector extraction module further comprises a heat exchange performance state heat exchange characteristic identification module, the heat exchange performance state of each mesh is input into the heat exchange performance state heat exchange characteristic identification module to conduct heat exchange characteristic identification, and the mesh heat exchange description vector corresponding to each mesh is obtained.

In operation S2032, the mesh heat exchange description vector corresponding to the end mesh (i.e., the last mesh at the bottom layer) of the heat exchange performance state relation mesh is input to the description vector fusion unit corresponding to the end mesh, so as to obtain the mesh fusion description vector corresponding to the end mesh.

Operation S2033, starting from the upper level knots of the end knots (i.e., knots from which the end knots are derived), a plurality of knots are reviewed.

In operation S2034, the mesh junction fusion description vector corresponding to the lower mesh junction of the currently processed mesh junction and the mesh junction heat exchange description vector corresponding to the currently processed mesh junction are input to the description vector fusion unit corresponding to the currently processed mesh junction, so as to obtain the mesh junction fusion description vector corresponding to the currently processed mesh junction.

The internode fusion description vector can characterize structure derivative relation from the terminal internode to the corresponding internode and merge semantic information.

In operation S2035, after the multiple knots are browsed (i.e., all traversed), the knots fusion description vector corresponding to the top knots (i.e., the knots located at the top layer, which may be regarded as the root) in the heat exchange performance state relation net is determined as the heat exchange performance description vector.

For example, the description vector extraction module performs description vector extraction on the heat exchange performance state relation network to obtain a heat exchange performance description vector of the heat exchange performance state relation network, including: performing description vector extraction on the heat exchange performance state relation network through a description vector extraction module to obtain a heat exchange performance state representation vector of the heat exchange performance state relation network, performing description vector extraction on the heat exchange performance state relation network through the description vector extraction module to obtain the heat exchange performance state representation vector of the heat exchange performance state relation network, wherein the description vector comprises performing heat exchange characteristic identification on the heat exchange performance state of each network node in a plurality of network nodes to obtain a network node heat exchange description vector corresponding to each network node, inputting the network node heat exchange description vector corresponding to the end network node of the heat exchange performance state relation network into a description vector fusion unit corresponding to the end network node in the description vector extraction module, the method comprises the steps of obtaining a network knot combination representation vector corresponding to a terminal network knot, browsing a plurality of network knots in sequence from an upper network knot of the terminal network knot, loading a network knot combination representation vector corresponding to a lower network knot of the network knot which is processed at present and a network knot heat exchange description vector corresponding to the network knot which is processed at present into a description vector extraction module, obtaining a network knot combination representation vector corresponding to the network knot which is processed at present, and determining the network knot combination representation vector corresponding to the network knot at the top end in a heat exchange performance state relation network as a heat exchange performance state representation vector after a plurality of network knots are browsed. The description vector extraction module is used for extracting the description vector of the heat exchange performance state relation network based on a description vector fusion unit which comprises a plurality of network nodes which are mutually mapped with the heat exchange performance state relation network, so that the structure derivative relation of the heat exchange performance state relation network from the tail end network node to the top end network node and the heat exchange performance semantic information are combined, and the characteristics of the heat exchange performance state relation network are characterized more accurately.

S204, carrying out description vector aggregation on the heat exchange performance description vector, the data cluster description vector and the second data set description vector to obtain a target aggregation vector.

The target aggregation vector is a feature vector obtained by aggregating the level representation vector of the heat exchange performance representation vector of the heat exchange data set of the heat exchanger to be detected and the heat exchange performance state relation network and the heat exchange performance semantic vector, and the feature correlation between the heat exchange data set of the heat exchanger to be detected and the heat exchange performance state relation network can be mined.

In an embodiment of the present disclosure, performing description vector aggregation on a heat exchange performance description vector, a data cluster description vector and a second data set description vector to obtain a target aggregate vector, including:

s2041, inputting the heat exchange performance description vector and the data cluster description vector into a description vector aggregation module to perform description vector aggregation, and obtaining a basic aggregation vector.

The basic aggregation vector is a representation of a heat exchange data set of a heat exchanger to be detected, which is embedded in a state derivative relation of a heat exchange performance state relation network, for example, the description vector aggregation module can be obtained by performing description vector aggregation optimization on a preset description vector aggregation module, the preset description vector aggregation module comprises an association identification unit, a weight calculation unit and a weight distribution unit, and the heat exchange performance description vector and the data cluster description vector are input into the description vector aggregation module to perform description vector aggregation, so that the basic aggregation vector is obtained, and the method comprises the following steps:

S20411, inputting the data cluster description vector and the heat exchange performance description vector into an association recognition unit for association relationship recognition to obtain a target association vector.

The target association vector characterizes the association degree between the data cluster description vector and the heat exchange performance description vector, for example, the association identifying unit may include a correlation determining unit, and inputs the data cluster description vector and the heat exchange performance description vector into the correlation determining unit to perform correlation determination, so as to obtain the target association vector. The correlation determination unit may be a tanh function.

S20412, inputting the target association vector into a weight calculation unit for calculation to obtain the association evaluation weight of the data cluster description vector.

For example, the weight calculating unit may be a normalized exponential function, through which the target association vector is calculated (normalization is completed), to obtain the association evaluation weight of the data cluster description vector.

S20413, inputting the relevance evaluation weight and the data cluster description vector into a weight distribution unit for weight distribution, and obtaining a basic aggregation vector.

S2042, inputting the basic aggregate vector, the second data set description vector and the heat exchange performance description vector into a vector splicing unit for vector splicing to obtain a target aggregate vector.

S205, inputting the target aggregate vector into a heat exchange performance state classification algorithm to classify the heat exchange performance states, and obtaining the heat exchange performance states corresponding to the heat exchange data set of the heat exchanger to be detected.

The target heat exchange performance state is one or more heat exchange performance states corresponding to the heat exchange data set of the heat exchanger to be detected in a plurality of heat exchange performance states of the heat exchange performance state relation network. The heat exchange performance state classification algorithm can be obtained by optimizing the heat exchange performance state classification of a preset heat exchange performance state classification algorithm, wherein the preset heat exchange performance state classification algorithm comprises a full connection layer and an output layer. The output layer is used for classifying the heat exchange performance states of the integration vectors to obtain target heat exchange performance states. Optionally, the heat exchange performance state may be final heat exchange performance state information corresponding to any end mesh heat exchange performance state in the heat exchange performance state relation network, and after inputting the target aggregate vector into a heat exchange performance state classification algorithm to classify the heat exchange performance states, obtaining a heat exchange performance state corresponding to the heat exchange data set of the heat exchanger to be detected, the method may further include: acquiring heat exchange performance state matching information, wherein the heat exchange performance state matching information represents the corresponding relation between the final heat exchange performance state information and the multi-level heat exchange performance state information corresponding to the final heat exchange performance state information; and determining target multi-level heat exchange performance state information corresponding to the heat exchange performance state according to the heat exchange performance state matching information.

For example, the final heat exchange performance state information includes a bottom layer heat exchange performance state, the multi-level heat exchange performance state information corresponding to the final heat exchange performance state information includes a multi-level heat exchange performance state corresponding to a bottom layer heat exchange performance state flag, the heat exchange performance state matching information characterizes a correspondence between the bottom layer heat exchange performance state flag and the multi-level heat exchange performance state flag corresponding to the bottom layer heat exchange performance state flag, the bottom layer heat exchange performance state flag is a heat exchange performance state flag corresponding to any end net junction heat exchange performance state in a heat exchange performance state relation net, and the multi-level heat exchange performance state flag is a multi-level heat exchange performance state flag corresponding to a heat exchange performance state from the corresponding end net junction heat exchange performance state to the top end net junction heat exchange performance state. Optionally, when the heat exchange performance state includes a plurality of bottom layer heat exchange performance state markers, determining, by the heat exchange performance state matching information, target multi-level heat exchange performance state information corresponding to the heat exchange performance state includes: determining a plurality of target multistage heat exchange performance state marks corresponding to the plurality of bottom layer heat exchange performance state marks through the heat exchange performance state matching information; after a plurality of target multi-stage heat exchange performance state marks corresponding to the plurality of bottom heat exchange performance state marks are determined through the heat exchange performance state matching information, a target heat exchange performance state relation network is generated through the plurality of target multi-stage heat exchange performance state marks.

Based on the information, the lowest heat exchange performance state is marked through a heat exchange performance state classification algorithm, and target multi-level heat exchange performance state information corresponding to the heat exchange performance state is determined according to the corresponding relation of the multi-level heat exchange performance state information corresponding to the last heat exchange performance state information, so that the accuracy of the heat exchange performance state classification is improved. The heat exchange characteristic recognition algorithm, the description vector extraction module, the description vector aggregation module, the vector splicing unit and the heat exchange performance state classification algorithm are obtained by training a preset heat exchange characteristic recognition algorithm, a preset description vector extraction module, a preset description vector aggregation module, a preset vector splicing unit and a preset heat exchange performance state classification algorithm.

The training process may specifically include:

s1, acquiring a heat exchange data training template of the heat exchanger and a preset heat exchange performance state corresponding to the heat exchange data training template of the heat exchanger.

The preset heat exchange performance state is a preset heat exchange performance state mark which is annotated in advance for the heat exchange data training template of the heat exchanger.

S2, inputting the heat exchange data training templates of the heat exchanger into a preset heat exchange characteristic recognition algorithm to perform heat exchange characteristic extraction, and obtaining training description vectors of a second data set of the heat exchange data training templates of the heat exchanger and training template data cluster description vectors of a plurality of training template heat exchange data clusters of the heat exchange data training templates of the heat exchanger.

And S3, extracting the description vector of the heat exchange performance state relation network through a preset description vector extraction module to obtain a heat exchange performance training description vector of the heat exchange performance state relation network.

S4, inputting the heat exchange performance training description vector and the training template data cluster description vector into a preset description vector aggregation module to carry out description vector aggregation, and obtaining a basic training aggregation vector.

And S5, inputting the basic training aggregate vector, the second data set training description vector and the heat exchange performance training description vector into a preset vector splicing unit for vector splicing to obtain a training template target aggregate vector.

S6, inputting the training template target aggregate vector into a preset heat exchange performance state classification algorithm to classify the heat exchange performance states, and obtaining the training heat exchange performance states corresponding to the heat exchange data training templates of the heat exchanger.

S7, determining algorithm loss through a preset heat exchange performance state and a training heat exchange performance state.

S8, optimizing a preset heat exchange characteristic recognition algorithm, a preset description vector extraction module, a preset description vector aggregation module, a preset vector splicing unit and a preset heat exchange performance state classification algorithm through algorithm loss to obtain the heat exchange characteristic recognition algorithm, the description vector extraction module, the description vector aggregation module, the vector splicing unit and the heat exchange performance state classification algorithm.

The training heat exchange performance state comprises a training heat exchange performance state mark of a heat exchange data training template of the heat exchanger; algorithm losses include heat exchange performance status flag losses; through the preset heat exchange performance state and the training heat exchange performance state, the algorithm loss determination comprises the following steps: and determining the loss of the heat exchange performance state mark based on the preset heat exchange performance state mark and the training heat exchange performance state mark. For example, determining the heat exchange performance status flag loss based on the preset heat exchange performance status flag and the training heat exchange performance status flag includes determining the heat exchange performance status flag loss between the preset heat exchange performance status flag and the training heat exchange performance status flag by a preset loss function, such as a cross entropy function. The heat exchange performance state marking loss characterizes the loss between a preset heat exchange performance state marking and a training heat exchange performance state marking.

Optionally, optimizing a preset heat exchange feature recognition algorithm, a preset description vector extraction module, a preset description vector aggregation module, a preset vector splicing unit and a preset heat exchange performance state classification algorithm through algorithm loss to obtain the heat exchange feature recognition algorithm, the description vector extraction module, the description vector aggregation module, the vector splicing unit and the heat exchange performance state classification algorithm, wherein the method comprises the following steps: through algorithm loss, iteratively optimizing network internal configuration variables of a preset heat exchange characteristic identification algorithm, a preset description vector extraction module, a preset description vector aggregation module, a preset vector splicing unit and a preset heat exchange performance state classification algorithm; repeating S2 until the heat exchange performance state classification optimization iteration of the configuration variables in the network of the preset heat exchange characteristic identification algorithm, the preset description vector extraction module, the preset description vector aggregation module, the preset vector splicing unit and the preset heat exchange performance state classification algorithm is achieved through algorithm loss, iteratively optimizing the preset heat exchange characteristic identification algorithm, the preset description vector extraction module, the preset description vector aggregation module, the preset vector splicing unit and the preset heat exchange performance state classification algorithm, and performing the heat exchange performance state classification optimization iteration of the configuration variables in the network of the preset heat exchange characteristic identification algorithm, the preset description vector extraction module, the preset description vector aggregation module, the preset vector splicing unit and the preset heat exchange performance state classification algorithm through algorithm loss until the heat exchange performance state classification optimization iteration of the configuration variables in the network of the preset description vector extraction module, the preset description vector aggregation module, the preset vector splicing unit and the preset heat exchange performance state classification algorithm is achieved when the heat exchange performance state classification convergence requirements are met.

Based on the method, the preset heat exchange characteristic recognition algorithm, the preset description vector extraction module, the preset description vector aggregation module, the preset vector splicing unit and the preset heat exchange performance state classification algorithm are subjected to joint training, so that the efficiency is high, and the algorithm with more accurate heat exchange performance state classification is obtained.

According to the embodiment, the heat exchange performance state classification is carried out, the context heat exchange characteristic recognition is carried out on the heat exchange data set of the heat exchanger to be detected through the heat exchange characteristic recognition algorithm with the heat exchange data clustering module, the heat exchange characteristic recognition module and the heat exchange characteristic fusion module, the context heat exchange characteristic integration of the heat exchange data set of the heat exchanger to be detected is completed, the heat exchange characteristic of the data cluster is more accurate, the description vector extraction operator containing a plurality of description vector fusion units which are mutually mapped with a plurality of network junctions in the heat exchange performance state relation network is used for carrying out description vector extraction on the heat exchange performance state relation network, the structure derivative relation and the heat exchange characteristic of the heat exchange performance state relation network from the end network junction to the top network junction are combined, the heat exchange characteristic of the heat exchange performance state relation network is characterized more accurately, meanwhile, the description vector aggregation module and the vector splicing unit are used for integrating the heat exchange characteristic of the heat exchange data set of the heat exchanger to be detected again, the representation capability of the target aggregation vector on the heat exchange characteristic of the heat exchange data set is further increased, and the heat exchange performance state detection is carried out through the bottom layer association between the heat exchange characteristic of the heat exchange data set to be detected and the heat exchange performance state in the target aggregation vector, and the layer association between the heat exchange performance state states is used for carrying out the heat exchange performance state detection, and reliable result.

Based on the foregoing embodiments, the embodiments of the present disclosure provide a variable flow microchannel heat exchanger performance detection apparatus, where each unit included in the apparatus and each module included in each unit may be implemented by a processor in a computer device; of course, the method can also be realized by a specific logic circuit; in practice, the processor may be a central processing unit (Central Processing Unit, CPU), microprocessor (Microprocessor Unit, MPU), digital signal processor (Digital Signal Processor, DSP) or field programmable gate array (Field Programmable Gate Array, FPGA), etc.

Fig. 2 is a schematic structural diagram of a performance detection device for a variable flow microchannel heat exchanger according to an embodiment of the disclosure, where, as shown in fig. 2, the performance detection device 210 for a variable flow microchannel heat exchanger includes:

the feature extraction module 211 is configured to obtain a data set description vector of a heat exchange data set of the heat exchanger to be detected, as a first data set description vector;

a feature adjustment module 212, configured to perform a multi-round adjustment on the first dataset description vector, where a round of adjustment includes: determining a description vector to be cleaned based on the description vector of the first data set, obtaining a cleaning intermediate state corresponding to the current round of adjustment according to the historical description vector to be cleaned obtained from the heat exchange data set of the historical heat exchanger to be detected, and cleaning the description vector to be cleaned according to the cleaning intermediate state to obtain an intermediate description vector corresponding to the current round of adjustment; in the first round of adjustment, the description vector to be cleaned is the description vector of the first data set, in the X-th round of adjustment, the description vector to be cleaned is obtained by aggregating the description vector of the first data set and the obtained intermediate description vector, and X is a positive integer;

The data cleaning module 213 is configured to determine an intermediate description vector obtained by adjusting a final wheel to be a target description vector, and clean the first dataset description vector according to a target data cleaning parameter obtained by mining the target description vector to obtain a second dataset description vector;

and the performance detection module 214 is configured to classify the heat exchange performance states according to the second data set description vector, and obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected.

The description of the apparatus embodiments above is similar to that of the method embodiments above, with similar advantageous effects as the method embodiments. In some embodiments, functions or modules included in the apparatus provided by the embodiments of the present disclosure may be used to perform the methods described in the embodiments of the method, and for technical details not disclosed in the embodiments of the apparatus of the present disclosure, please understand with reference to the description of the embodiments of the method of the present disclosure.

It should be noted that, in the embodiment of the present disclosure, if the above-mentioned performance detection method of the variable flow micro-channel heat exchanger is implemented in the form of a software functional module, and sold or used as a separate product, the performance detection method may also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present disclosure may be essentially or portions contributing to the related art, and the software product may be stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present disclosure. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, an optical disk, or other various media capable of storing program codes. Thus, embodiments of the present disclosure are not limited to any specific hardware, software, or firmware, or any combination of the three.

The disclosed embodiments provide a computer device comprising a memory storing a computer program executable on the processor and a processor implementing some or all of the steps of the above method when the processor executes the program.

The disclosed embodiments provide a computer readable storage medium having stored thereon a computer program which when executed by a processor performs some or all of the steps of the above method. The computer readable storage medium may be transitory or non-transitory.

The disclosed embodiments provide a computer program comprising computer readable code which, when run in a computer device, performs some or all of the steps for implementing the methods described above.

Embodiments of the present disclosure provide a computer program product comprising a non-transitory computer-readable storage medium storing a computer program which, when read and executed by a computer, performs some or all of the steps of the above-described method. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In some embodiments, the computer program product is embodied as a computer storage medium, in other embodiments the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It should be noted here that: the above description of various embodiments is intended to emphasize the differences between the various embodiments, the same or similar features being referred to each other. The above description of apparatus, storage medium, computer program and computer program product embodiments is similar to that of method embodiments described above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the disclosed apparatus, storage medium, computer program and computer program product, please refer to the description of the embodiments of the disclosed method.

Fig. 3 is a schematic diagram of a hardware entity of a computer device according to an embodiment of the present disclosure, as shown in fig. 3, the hardware entity of the computer device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program executable on the processor 1001, the processor 1001 implementing the steps in the method of any of the embodiments described above when the program is executed.

The memory 1002 stores a computer program executable on a processor, and the memory 1002 is configured to store instructions and applications executable by the processor 1001, and may also cache data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or already processed by each module in the processor 1001 and the computer device 1000, which may be implemented by a FLASH memory (FLASH) or a random access memory (Random Access Memory, RAM).

The steps of the variable flow microchannel heat exchanger performance detection method of any one of the above are implemented when the processor 1001 executes a program. The processor 1001 generally controls the overall operation of the computer device 1000.

Embodiments of the present disclosure provide a computer storage medium storing one or more programs executable by one or more processors to implement the steps of the variable flow microchannel heat exchanger performance detection method of any of the embodiments above.

It should be noted here that: the description of the storage medium and apparatus embodiments above is similar to that of the method embodiments described above, with similar benefits as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and apparatus of the present disclosure, please refer to the description of the embodiments of the method of the present disclosure for understanding. The processor may be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronic devices implementing the above-described processor functions may be other, and embodiments of the present disclosure are not particularly limited.

The computer storage medium/Memory may be a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable programmable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable Read Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic random access Memory (Ferromagnetic Random Access Memory, FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Read Only optical disk (Compact Disc Read-Only Memory, CD-ROM); but may also be various terminals such as mobile phones, computers, tablet devices, personal digital assistants, etc., that include one or any combination of the above-mentioned memories.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present disclosure, the size of the sequence numbers of the steps/processes described above does not mean the order of execution, and the order of execution of the steps/processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation of the embodiments of the present disclosure. The foregoing embodiment numbers of the present disclosure are merely for description and do not represent advantages or disadvantages of the embodiments. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the several embodiments provided in the present disclosure, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

Alternatively, the above-described integrated units of the present disclosure may be stored in a computer-readable storage medium if implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, the technical solutions of the present disclosure may be embodied essentially or in part in a form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present disclosure. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.

The foregoing is merely an embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and any person skilled in the art can easily think about the changes or substitutions within the technical scope of the present disclosure, and should be covered by the protection scope of the present disclosure.

Claims

1. A method for detecting performance of a variable flow microchannel heat exchanger, the method being applied to a computer device, the method comprising:

performing a plurality of rounds of adjustment on the first dataset description vector, one round of adjustment comprising: determining a description vector to be cleaned based on the description vector of the first data set, obtaining a cleaning intermediate state corresponding to the current round of adjustment according to the historical description vector to be cleaned obtained from the heat exchange data set of the historical heat exchanger to be detected, and cleaning the description vector to be cleaned according to the cleaning intermediate state to obtain an intermediate description vector corresponding to the current round of adjustment; in the first round of adjustment, the description vector to be cleaned is the description vector of the first data set, in the X-th round of adjustment, the description vector to be cleaned is obtained by aggregating the description vector of the first data set and the obtained intermediate description vector, and X is a positive integer;

classifying heat exchange performance states through the second data set description vector to obtain heat exchange performance states of the heat exchange data set of the heat exchanger to be detected;

wherein for one round of adjustment:

2. The method for detecting performance of a variable flow microchannel heat exchanger according to claim 1, wherein the aggregation process of the description vectors to be cleaned comprises:

3. The method for detecting performance of a variable flow microchannel heat exchanger according to claim 1, wherein the classifying the heat exchange performance states by the second data set description vector to obtain the heat exchange performance states of the heat exchange data set to be detected comprises:

4. A variable flow microchannel heat exchanger performance detection method according to claim 3, wherein the method is implemented according to a trained heat exchange performance state detection algorithm, the heat exchange performance state detection algorithm being trained by:

5. The variable flow microchannel heat exchanger performance detection method of claim 4, wherein the algorithmic learnable variables comprise a first learnable variable for purging and a second learnable variable for supporting coefficient determination; the obtaining the inferred heat exchange performance state label for classifying the heat exchange performance state comprises the following steps:

the sample annotation also comprises pure heat exchange performance sample description vectors extracted from the corresponding data samples of the heat exchanger to be cleaned, wherein the pure heat exchange performance sample description vectors are data set description vectors which do not need cleaning; the correcting the algorithm learning variable in the heat exchange performance state detection algorithm through the loss between the inferred heat exchange performance state mark and the corresponding sample heat exchange performance state mark comprises the following steps:

optimizing the first and second learnable variables by the fusion penalty.

6. The method for detecting performance of a variable flow microchannel heat exchanger according to claim 1, wherein the classifying the heat exchange performance states by the second data set description vector to obtain the heat exchange performance states of the heat exchange data set to be detected comprises:

7. The method for detecting performance of a variable flow microchannel heat exchanger according to claim 6, wherein the description vector extraction module includes a plurality of description vector fusion units mapped to a plurality of nodes in the heat exchange performance state relation network, and the description vector extraction module performs description vector extraction on the heat exchange performance state relation network to obtain a heat exchange performance description vector of the heat exchange performance state relation network, and the method includes:

8. The utility model provides a variable flow microchannel heat exchanger performance detection device which characterized in that includes:

the performance detection module is used for classifying the heat exchange performance states through the second data set description vector to obtain the heat exchange performance states of the heat exchange data set of the heat exchanger to be detected;

wherein for one round of adjustment:

9. A computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the program is executed.