Air conditioner defect detection method and system integrating large language model and data driving

Abstract

The invention discloses an air conditioner defect detection method and system integrating a large language model and data driving, and belongs to the technical field of equipment defect detection. Comprising the steps of obtaining sensing signal data of the central air conditioner and performing feature extraction to generate a signal feature vector; fusing the signal feature vector and a preset text feature vector, and processing a fusion result through a large language model to obtain an embedded enhanced feature; wherein the text feature vector is generated by embedding a manually detected and recorded short text as an input query into a fine-tuned large language model, and the fine-tuned large language model is obtained by performing fine tuning on the large language model through mask language modeling by utilizing a professional knowledge text of the central air conditioner; and performing linear regression based on the embedded enhanced features to obtain a defect detection result. The defect detection precision of the central air conditioner can be improved, and the problem that air conditioner defect detection is limited due to single data source and large data scale in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment defect detection, and particularly to an air-conditioning defect detection method and system integrating a large language model and data-driven Background Art

[0002] The statements in this section only mention the background art related to the present invention and do not necessarily constitute prior art.

[0003] As a common environmental conditioning method in modern buildings, central air-conditioning has been popularized in public areas such as office areas, campus classrooms, shopping malls, and even private spaces such as family housing. With the development of electronic technology and automation technology, central air-conditioning systems have begun to adopt advanced technologies such as digital control systems and intelligent sensors to improve the energy efficiency performance and control accuracy of the systems. At the same time, innovations such as variable-frequency air-conditioning, total heat exchange, and energy-saving technologies have emerged, greatly improving the energy efficiency and environmental protection performance of central air-conditioning systems. With the continuous improvement of performance and the introduction of new products, fault diagnosis of central air-conditioning has become particularly important.

[0004] A central air-conditioning system usually consists of multiple components and devices, including air-conditioning units, air ducts, cooling water systems, etc. Therefore, the causes of its faults may involve multiple aspects. Once a central air-conditioning system fails, it may affect the environmental conditioning effect in the entire building or area. Existing central air-conditioning fault diagnosis methods have the following problems:

[0005] (1) Central air-conditioning fault diagnosis driven by expert experience and knowledge depends on known faults and cannot timely identify and diagnose unknown faults. For some advanced faults or complex problems, technical personnel with professional knowledge and maintenance experience are required to solve them, relying greatly on manual judgment.

[0006] (2) Pure data-driven air-conditioning fault diagnosis depends on historical data, and there may be problems such as incomplete fault causes in the data set. Moreover, current fault diagnosis methods also have difficulties and limitations in processing and analyzing a large amount of data. Summary of the Invention

[0007] To solve the deficiencies of the prior art, the present invention provides an air-conditioning defect detection method, system, electronic device, computer-readable storage medium, and computer program product integrating a large language model and data-driven, which effectively integrates valuable information from different sources and improves the accuracy of central air-conditioning defect detection.

[0008] In the first aspect, the present invention provides an air-conditioning defect detection method integrating a large language model and data-driven;

[0009] An integrated large language model and data-driven air conditioner defect detection method, comprising:

[0010] Obtain the perception signal data of the central air conditioner and perform feature extraction to generate a signal feature vector;

[0011] Fuse the signal feature vector and a preset text feature vector, and process the fusion result through a large language model to obtain an embedded enhanced feature; wherein, the text feature vector is generated by taking a short text of manual detection records as an input query and embedding it into a fine-tuned large language model, and the fine-tuned large language model is fine-tuned by using the central air conditioner professional knowledge text through masked language modeling;

[0012] Perform linear regression based on the embedded enhanced feature to obtain a defect detection result.

[0013] In some embodiments, the specific method of fine-tuning the large language representation model by using the central air conditioner professional knowledge text through masked language modeling is:

[0014] Take the central air conditioner professional knowledge text as an input sequence and perform random masking, input the randomly masked central air conditioner professional knowledge text sequence into the large language model for processing, generate the probability distribution of each word in the central air conditioner professional knowledge text sequence and output the predicted words.

[0015] In some embodiments, taking a short text of manual detection records as an input query and embedding it into a fine-tuned large language model to generate a text feature vector includes:

[0016] Convert the short text of manual detection records into a token sequence containing multiple sub-words through the large language model and map it to a vector representation, extract the embedding vector and add position encoding;

[0017] Process the embedding vector through a Transformer layer to obtain a text feature vector.

[0018] In some embodiments, the step of obtaining the perception signal data of the central air conditioner and performing feature extraction to generate a signal feature vector includes:

[0019] Normalize the perception signal data in sequence and add learnable scaling and offset parameters to generate central air conditioner signal data;

[0020] Process the central air conditioner signal data through a convolutional neural network to obtain a signal feature vector.

[0021] In some embodiments, the step of fusing the signal feature vector and a preset text feature vector, and processing the fusion result through a large language model to obtain an embedded enhanced feature includes:

[0022] Concatenate the signal feature vector and the preset text feature vector according to the feature dimension to obtain a fused feature vector;

[0023] Input the fused feature vector into the large language model, and perform sequential processing using the multi-head attention mechanism and the feed-forward neural network, and combine the residual connection to output the embedded enhanced feature.

[0024] In some embodiments, the linear regression based on the embedded enhanced feature to obtain the defect detection result is specifically: process the embedded enhanced feature through a multi-layer perceptron to obtain the defect detection result.

[0025] In a second aspect, the present invention provides an integrated large language model and data-driven air conditioner defect detection system;

[0026] An integrated large language model and data-driven air conditioner defect detection system includes:

[0027] A feature extraction module, configured to: obtain the sensed signal data of the central air conditioner and perform feature extraction to generate a signal feature vector;

[0028] A feature fusion module, configured to: fuse the signal feature vector and the preset text feature vector, and process the fusion result through a large language model to obtain an embedded enhanced feature; wherein, the text feature vector is generated by inputting the short text of the manual detection record as a query and embedding it into the fine-tuned large language model, and the fine-tuned large language model is fine-tuned by using the professional knowledge text of the central air conditioner through masked language modeling;

[0029] A fault category discrimination module, configured to: perform linear regression based on the embedded enhanced feature to obtain the defect detection result.

[0030] In a third aspect, the present invention provides an electronic device;

[0031] An electronic device includes a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the above integrated large language model and data-driven air conditioner defect detection method.

[0032] In a fourth aspect, the present invention provides a computer-readable storage medium;

[0033] A computer-readable storage medium, on which a computer program / instructions are stored, and when the computer program / instructions are executed by a processor, the steps of the above integrated large language model and data-driven air conditioner defect detection method are implemented.

[0034] In a fifth aspect, the present invention provides a computer program product;

[0035] A computer program product includes a computer program / instructions which, when executed by a processor, implement the steps of the above-mentioned integrated large language model and data-driven air conditioner defect detection method.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] 1. For the technical solution provided by the present invention, the large language representation model is fine-tuned by using the professional knowledge in the central air conditioner field, enabling the large language representation model to better cope with the complexity and specific requirements of the central air conditioner structure; and the large language model is used to extract feedforward information through a deep neural network, and the multi-head attention mechanism is used to focus on different parts of the input and specific task fine-tuning is performed, solving the problem of restricting diagnosis due to large data scale, and at the same time providing other source data for central air conditioner defect detection, solving the limitations on diagnosis caused by solely expert knowledge-driven and data-driven methods.

[0038] 2. For the technical solution provided by the present invention, unstructured text data is effectively utilized, enabling the model to process valuable information from unstructured texts such as technical documents and maintenance logs, processing and effectively integrating data from different sources, and making up for the problem of poor defect detection accuracy of traditional central air conditioners based on single data.

[0039] 3. For the technical solution provided by the present invention, when the large language model processes the fused feature vectors, the output layer is frozen to keep the weights of this layer unchanged and trained when new data enters, saving time and training volume, solving the problem of poor processing speed, and achieving instant feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0041] Figure 1 It is a schematic flowchart of the integrated large language model and data-driven air conditioner defect detection method provided by the embodiment of the present invention;

[0042] Figure 2 It is a schematic flowchart of the large language model embedding output training structure provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0044] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0045] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0046] Embodiment 1

[0047] The defect detection of existing central air conditioners relies on known faults and historical data, and the prediction accuracy for future faults needs to be improved. Therefore, the present invention provides an integrated large language model and data-driven air conditioner defect detection method, which fuses the real-time perception data of the central air conditioner with the professional text data of the central air conditioner and the manual inspection records, etc. to perform air conditioner defect detection and improve the accuracy of defect detection.

[0048] Next, in combination with Figure 1 - Figure 2 , a detailed description will be given of an integrated large language model and data-driven air conditioner defect detection method disclosed in this embodiment. The integrated large language model and data-driven air conditioner defect detection method includes:

[0049] S1. Obtain the perception signal data of the central air conditioner and perform feature extraction to generate a signal feature vector.

[0050] In this embodiment, the perception signal data includes the temperature signal, pressure signal, flow signal, etc. of the central air conditioner. In order to adapt to different samples and ensure the expression ability of the model, the perception signal data is preprocessed in S1.

[0051] As an implementation manner, S1 specifically includes:

[0052] S101. Process the temperature signal, pressure signal and flow signal, calculate the standard deviation and mean of each sample therein, and normalize the input samples based on this.

[0053] Exemplarily, for each sample i of the input signal x, calculate the mean and standard deviation; the mean is expressed as:

[0054]

[0055] The standard deviation is expressed as:

[0056]

[0057] Among them, N represents the number of samples, and x ij is the j-th eigenvalue of the i-th sample.

[0058] Then, the input samples are normalized based on the calculated mean and standard deviation, expressed as:

[0059]

[0060] Then, learnable scaling and offset parameters are added to the normalized data to obtain the central air-conditioning signal data.

[0061]

[0062] In the formula, β represents the learnable offset parameter, which is a very small constant used to avoid division-by-zero errors in the output; γ represents the scaling factor for scaling and can adjust the output.

[0063] Based on this, the model can learn more complex feature representations to restore the expressive ability of the model.

[0064] S102. Input the central air-conditioning signal data into the trained convolutional neural network (CNN) for feature extraction to obtain the signal feature vector.

[0065] In this step, the network architecture of the convolutional neural network is not improved. Its specific network architecture and data processing flow are both existing technologies and will not be elaborated here.

[0066] S2. Fuse the signal feature vector and the preset text feature vector, and process the fusion result through the large language model to obtain the embedded enhanced feature.

[0067] Furthermore, using the central air-conditioning professional knowledge text, the large language model is fine-tuned through masked language modeling. Taking the artificial detection record short text as the input query and embedding it into the fine-tuned large language model to generate the text feature vector.

[0068] To improve the accuracy of central air-conditioning defect detection, in this step, fine-tuning is carried out in the large language representation model with the help of central air-conditioning professional text data. At the same time, combining the artificial detection text to construct a text feature vector that retains the context semantics, providing multi-source basic data for central air-conditioning defect detection and ensuring the comprehensiveness and authenticity of the detection.

[0069] Moreover, the large language model is usually based on the self-attention mechanism, can process data in parallel, improve the computing efficiency, and has more advantages than traditional models such as the recurrent neural network (RNN), such as ChatGPT2.

[0070] As an implementation, using central air-conditioning professional knowledge texts, fine-tune the large language model through masked language modeling, and use manually detected short texts as input queries to embed into the fine-tuned large language model to generate text feature vectors, including:

[0071] Step 1: Use the central air-conditioning professional knowledge text as the input sequence X = [x 1 , x 2 , x 3 ,..., x M . Randomly select some words in the input sequence for masking. The input sequence after masking is represented as X' = [x 1 , x 2 , x 3 ,..., x m , MASK,..., x M .

[0072] Here, M represents the length of the input sequence, MASK represents the mask, and the proportion of masked words is 10% - 15%.

[0073] Here, the central air-conditioning professional knowledge text includes central air-conditioning and ventilation system terms, central air-conditioning and ventilation system maintenance specifications, fault diagnosis manuals, fault diagnosis records, etc.

[0074] Step 2: Load the large language model, input the masked input sequence into the large language representation model for processing, fine-tune the large language model through masked language modeling technology, and freeze the output layer.

[0075] Specifically, for the masked input sequence X', the large language model will generate the probability distribution of each word; for each position i in X', output the predicted probability of the vocabulary corresponding to that position According to the probability comparison, output the predicted words.

[0076] Let the large language model perform context prediction and understanding on the masked words, and then learn the language features and terms in the central air-conditioning field, so as to achieve fine-tuning in the professional field.

[0077] Step 3: Embed the manually detected short text as an input query into the fine-tuned large language representation model to obtain a text feature vector that can reflect its context content.

[0078] Furthermore, use the manually detected short text as input data and transfer it to the fine-tuned large language model. The large language model converts the manually detected short text into a text feature vector. First, decompose the manually detected short text into subwords through the large language model, and use the tokenizer BPE in the large language model to convert the text containing multiple subwords into a token sequence {t1 ,t 2 ,t 3 ,...,t n}, each tag t i is mapped to a vector representation E(t i ), E(t i ) is the embedding matrix W[t i Extract the marker t from i The embedding vector is then input into the Transformer layer of the large language model, and feature extraction and learning are performed through forward propagation and self-attention mechanism. The input vector is x i =E(t i )+P(t i ), where P(t i ) is a position code used to maintain sequence information.

[0079] Specifically, the key vector, query vector and value vector in the self-attention mechanism are defined as K = Q = V = W K X, where X represents the total input of the embedding, and W K Represents the learned weight matrix.

[0080] By matching the query with the key, the relevance of different parts of the input sequence is determined to obtain the attention weights, and the weight vector is used to generate the output d k Indicates the dimension of the key.

[0081] Integrate multiple layers of output to generate the final text feature vector representation:

[0082] H = LayerNorm (Z + X).

[0083] The output after forward propagation enters the self-attention mechanism to capture contextual connections and understand the text.

[0084] In order to detect central air conditioner defects based on multiple data sources, in this embodiment, the signal feature vector containing the real-time operation characteristics of the central air conditioner and the text feature vector retaining the context semantics and containing the professional knowledge of the central air conditioner are fused for subsequent processing. As an implementation method, S2 includes:

[0085] S201, concatenate the signal feature vector and the text feature vector in the feature dimension to form a new feature matrix, namely, a fused feature vector.

[0086] Furthermore, before executing S201, the signal feature vector and the text feature vector are transformed into a dimension reduction and standardized. Specifically, the principal component analysis technique is used to reduce the feature dimension of the high-dimensional data.

[0087] S202. Input the fused feature vector into the feature fusion module in the large language model for processing, and output the embedded enhanced feature.

[0088] The fine-tuned large language model has domain adaptation ability and can optimize domain tasks. By processing the fused feature vector through the large language model and strengthening the fused feature vector in terms of domain knowledge, the feature vector can perform better in the specific context.

[0089] Furthermore, combined with Figure 2 , the feature fusion module includes a multi-head attention mechanism, a normalization layer, a feed-forward neural network, and a normalization layer connected in sequence. The input and output of the multi-head attention mechanism are added and then input into the first normalization layer. The input and output of the feed-forward neural network are added and then input into the second normalization layer. The feed-forward neural network includes an input layer, a hidden layer, and an output layer.

[0090] Furthermore, the processing flow of the large language model for the fused feature vector is as follows:

[0091] (1) Input the fused feature vector into the multi-head attention mechanism. Through multiple attention heads, the large language model learns information from different subspaces. Each head focuses on different parts of the fused feature vector from multiple angles of information and processes the linear calculations of multiple feature dimensions.

[0092] Exemplarily, let the fused feature vector be represented as Y, the dimension be represented as n×d, where n represents the number of samples and d represents the feature dimension; let Ensure that the output dimension of each head is the same. Define the query vector Q, the key vector K, and the weight vector V, which are represented as: Q = YW Q , K = YW K , V = YW V . Then split Q, K, and V into multiple heads, and each head generates an attention output And splice and linearly transform the results obtained by each independent attention head to obtain the output vector.

[0093] (2) Add the fused feature vector to the output processed by the multi-head attention mechanism to achieve residual connection and normalize the features of each sample after residual connection.

[0094] (3) Input the normalized data into the feed-forward neural network. The feed-forward neural network receives the data. The input layer corresponds each feature to a node and transmits it to multiple hidden layers. The neurons in the hidden layer use the weighted sum activation function to process the input. The output of each layer is used as the input of the next layer and is passed layer by layer until the final output is obtained.

[0095] Specifically, the calculation process of the output of each neuron is represented as:

[0096]

[0097] In the formula, represents the weight, represents the bias, φ represents the activation function, l represents the specific hidden layer, i represents the neuron number of the (l - 1)th layer, and j represents the neuron number of the lth layer.

[0098] S3. Perform linear regression based on the embedded enhanced features to obtain the defect detection result.

[0099] Furthermore, process the embedded enhanced features through a multi-layer perceptron to obtain the defect detection result; specifically, first, receive the embedded enhanced features through the input layer, then perform non-linear transformation on the embedded enhanced features through the activation functions in multiple hidden layers to achieve feature extraction; finally, obtain the output vector through the output layer, and convert the output vector into class probabilities through the activation function Softmax to achieve multi-fault classification.

[0100] Exemplarily, given an embedded enhanced feature represented as z = [z 1 , z 2 , z 3 ,..., z k , then the conversion process is expressed as:

[0101]

[0102] In the formula, k represents the dimension of the embedded enhanced feature.

[0103] Embodiment 2

[0104] This embodiment discloses an integrated large language model and data-driven air conditioner defect detection system, including:

[0105] A feature extraction module, configured to: obtain the sensing signal data of the central air conditioner and perform feature extraction to generate a signal feature vector;

[0106] A feature fusion module, configured to: fuse the signal feature vector and a preset text feature vector, and process the fusion result through a large language model to obtain an embedded enhanced feature; wherein, the text feature vector is generated by using the central air conditioner professional knowledge text, fine-tuning the large language representation model through masked language modeling, and embedding the artificial detection record short text as an input query into the fine-tuned large language representation model;

[0107] A fault category discrimination module, configured to: perform linear regression based on the embedded enhanced features to obtain the defect detection result.

[0108] It should be noted here that the above feature extraction module, feature fusion module, and fault category discrimination module correspond to the steps in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0109] Embodiment 3

[0110] Embodiment 3 of the present invention provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps of the above integrated large language model and data-driven air conditioner defect detection method are completed.

[0111] Embodiment 4

[0112] Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above integrated large language model and data-driven air conditioner defect detection method are completed.

[0113] Embodiment 5

[0114] Embodiment 5 of the present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the above integrated large language model and data-driven air conditioner defect detection method are implemented.

[0115] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0116] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing device, where they perform a series of operation steps to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process or multiple processes and / or one block or multiple blocks in the flow. Figure 1 one process or multiple processes and / or Figure 1 one block or multiple blocks.

[0118] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0119] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An air conditioning defect detection method integrating a large language model and data-driven, characterized in that: include: Obtain the perception signal data of the central air conditioner and perform feature extraction to generate a signal feature vector; The signal feature vector and the preset text feature vector are fused, and the fusion result is processed by the large language model to obtain the embedding enhancement feature; wherein the text feature vector is generated by embedding the short text of the manual detection record as the input query into the fine-tuned large language model, and the fine-tuning large language model is to use the central air conditioning professional knowledge text to fine-tune the large language model through mask language modeling; Linear regression is performed based on the embedded enhanced features to obtain defect detection results.

2. The air conditioner defect detection method integrating a large language model and data-driven according to claim 1, characterized in that: Using the central air conditioning professional knowledge text, the large language model is fine-tuned through masked language modeling. Specifically: The central air-conditioning professional knowledge text is used as the input sequence and randomly masked. The randomly masked central air-conditioning professional knowledge text sequence is input into the large language model for processing, the probability distribution of each word in the central air-conditioning professional knowledge text sequence is generated, and the predicted words are output.

3. The air conditioner defect detection method integrating a large language model and data-driven according to claim 1, characterized in that: The short text of the manual detection record is used as the input query embedding into the fine-tuned large language model to generate the text feature vector including: The short text of the manual detection record is converted into a token sequence containing multiple subwords through a large language model and mapped to a vector representation, and the embedding vector is extracted and embedded into the positional encoding; The embedded vector is processed through the Transformer layer to obtain the text feature vector.

4. The air conditioner defect detection method integrating a large language model and data-driven as claimed in claim 1, characterized in that: The acquiring of the perception signal data of the central air conditioner and performing feature extraction to generate a signal feature vector comprises: Normalize the perception signal data in turn and add learnable scaling and offset parameters to generate central air conditioning signal data; The central air-conditioning signal data is processed through a convolutional neural network to obtain the signal feature vector.

5. The air conditioner defect detection method integrating a large language model and data-driven according to claim 1, characterized in that: The step of fusing the signal feature vector with the preset text feature vector and processing the fusion result through the large language model to obtain the embedded enhanced feature includes: The signal feature vector and the preset text feature vector are concatenated according to the feature dimension to obtain a fused feature vector; The fused feature vector is input into the large language model, processed sequentially using the multi-head attention mechanism and feedforward neural network, and the enhanced features are embedded in combination with the residual connection output.

6. The air conditioner defect detection method integrating a large language model and data-driven according to claim 1, characterized in that: The method of performing linear regression based on embedded enhanced features to obtain defect detection results specifically includes: processing the embedded enhanced features through a multi-layer perceptron to obtain defect detection results.

7. An air conditioning defect detection system integrating a large language model and data-driven, characterized in that: include: The feature extraction module is configured to: obtain the perception signal data of the central air conditioner and perform feature extraction to generate a signal feature vector; The feature fusion module is configured to: fuse the signal feature vector and the preset text feature vector, process the fusion result through the large language model, and obtain the embedding enhancement feature; wherein the text feature vector is generated by embedding the short text of the manual detection record as the input query into the fine-tuned large language model, and the fine-tuning large language model is to use the central air-conditioning professional knowledge text to fine-tune the large language model through mask language modeling; The fault category discrimination module is configured to perform linear regression based on the embedded enhanced features to obtain defect detection results.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the air conditioning defect detection method integrating a large language model and data-driven according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the air conditioning defect detection method integrating a large language model and data-driven are implemented as described in any one of claims 1-6.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the air conditioning defect detection method integrating a large language model and data-driven are implemented as described in any one of claims 1-6.

Images