Property equipment data analysis method and device based on artificial intelligence
By using AI-based multimodal sensor data analysis and knowledge distillation operations based on edge computing and cloud computing models, the problem of single-parameter fault identification and low efficiency in property equipment inspection has been solved, achieving efficient and accurate fault analysis and handling.
Patent Information
- Application Number
- CN202511021759.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-23
- Publication Date
- 2025-11-07
AI Technical Summary
Existing property equipment inspection methods mainly rely on single parameter thresholds to identify faults, lack the ability to analyze composite parameters, are difficult to deal with new types of faults, and the inspection efficiency depends on path design, making it difficult to maximize.
By employing an AI-based data analysis method, multimodal sensing data is collected using sensing components. Through knowledge distillation operations using edge computing and cloud computing models, fault analysis and response plan determination are performed, achieving fault identification and path optimization based on composite parameters.
It improved the accuracy of property equipment failure analysis and inspection efficiency, reduced reliance on manpower, and achieved standardized analysis and efficient failure handling.
Smart Images
Figure CN120910448A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of artificial intelligence, and in particular to a property equipment data analysis method and device based on artificial intelligence. BACKGROUND
[0002] In property services, various property equipment is usually involved, relating to water, electricity, gas, network and many other aspects. Various equipment needs to be regularly inspected and maintained to exclude faults and avoid affecting service quality. Therefore, daily inspection work is an important content in property services. At the present stage, the inspection work is mainly completed by manual work, that is, the staff holds a specific inspection device and inspects in the community according to the established route to inspect the equipment in turn. In some other cases, there is a solution to replace manual work with robots, that is, a robot inspects the equipment in turn according to the established route. This greatly reduces the dependence on manpower.
[0003] However, the existing inspection methods, whether relying on manpower or robots, have basically the same core principle. The main defect is that it usually only identifies faults by a single parameter threshold, the analysis logic of the fault is very simple, it is not good at fault analysis by composite parameters, and it lacks the ability to cope with new faults and previously unseen types of faults. Moreover, the inspection efficiency is largely dependent on the inspection path, and often cannot maximize the efficiency. SUMMARY
[0004] The present application provides a property equipment data analysis method and device based on artificial intelligence, which realizes data analysis of property equipment based on an artificial intelligence operation model to discover faults.
[0005] In a first aspect, the present application provides a property equipment data analysis method based on artificial intelligence, which is applied to a data analysis system and includes:
[0006] Multi-modal sensing data is collected by a sensing assembly, and the multi-modal sensing data is preprocessed to obtain data to be analyzed;
[0007] The data to be analyzed is input into an edge computing model to make the edge computing model determine an analysis result; the edge computing model is periodically updated by a cloud computing model through knowledge distillation operation;
[0008] According to the analysis result, a disposal scheme for the property equipment is determined.
[0009] Preferably, the multi-modal sensing data includes image data, audio data, and infrared thermal imaging data; the preprocessing of the multi-modal sensing data to obtain the data to be analyzed includes:
[0010] The multi-modal sensor data is subjected to data desensitization processing and feature compression processing to obtain an incremental data set;
[0011] The incremental data set is used as the data to be analyzed.
[0012] Preferably, it further comprises:
[0013] When a new failure mode is detected, the cloud computing model updates the adapter layer parameters using a parameter efficient fine-tuning technique, and creates a processing node in the knowledge graph, and associates a disposal rule with the processing node.
[0014] Preferably, the edge computing model is periodically updated by the cloud computing model through a knowledge distillation operation, comprising:
[0015] When the cloud computing model associates a disposal rule with the processing node, the cloud computing model updates the disposal rule to the edge computing model through a knowledge distillation operation.
[0016] Preferably, the expression of the knowledge distillation operation is:
[0017]
[0018] Wherein, L KD represents the knowledge distillation loss; T represents the temperature coefficient, and T>1; KL represents the KL divergence; σ is the softmax function; z t is the original output of the cloud computing model; z s is the original output of the edge computing model.
[0019] Preferably, the edge computing model is periodically updated by the cloud computing model through a knowledge distillation operation, comprising:
[0020] The edge computing model receives an update package from the cloud computing model and performs hot updating using a dynamic module loading technique.
[0021] Preferably, it further comprises:
[0022] After the execution of the disposal scheme, the execution result is fed back to the cloud computing model, so that the cloud computing model performs reinforcement learning of the knowledge graph.
[0023] In a second aspect, the present application provides a property equipment data analysis device based on artificial intelligence, which is placed in a data analysis system and comprises:
[0024] A data acquisition module is used to acquire multi-modal sensor data using a sensor assembly and to pre-process the multi-modal sensor data to obtain data to be analyzed.
[0025] The data analysis module is configured to input the data to be analyzed into an edge computing model, so that the edge computing model determines an analysis result; and the edge computing model is periodically updated through a cloud computing model via a knowledge distillation operation.
[0026] The processing module is configured to determine a processing scheme for the property equipment according to the analysis result.
[0027] In a third aspect, the present application provides a readable medium comprising execution instructions, when a processor of an electronic device executes the execution instructions, the electronic device executes the method according to any one of the first aspect.
[0028] In a fourth aspect, the present application provides an electronic device comprising a processor and a memory storing execution instructions, when the processor executes the execution instructions stored in the memory, the processor executes the method according to any one of the first aspect.
[0029] The present application provides a property equipment data analysis method and device based on artificial intelligence, which determines the fault of the property equipment and gives a processing scheme by performing operation and analysis on multi-modal data based on the computing model of artificial intelligence technology, thereby replacing manual inspection, saving human resources, improving the work efficiency of property services, realizing the standardization of analysis, improving the accuracy, and no longer relying on the experience of staff.
[0030] The further effects of the above-mentioned non-conventional preferred modes will be described in the following in combination with the specific embodiments. BRIEF DESCRIPTION OF DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present application or the prior technical solutions, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.
[0032] Figure 1 A flowchart of a property equipment data analysis method based on artificial intelligence provided by an embodiment of the present application;
[0033] Figure 2 A flowchart of another property equipment data analysis method based on artificial intelligence provided by an embodiment of the present application;
[0034] Figure 3 A structural diagram of a property equipment data analysis device based on artificial intelligence provided by an embodiment of the present application;
[0035] Figure 4A structural schematic diagram of an electronic device is provided for an embodiment of the present application. DETAILED DESCRIPTION
[0036] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with specific embodiments and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.
[0037] In property services, various property equipment is usually involved, which relates to water, electricity, gas, network and many other aspects. Various equipment needs to be checked and maintained regularly, and faults are excluded to avoid affecting service quality. Therefore, daily inspection work is an important content in property services. At the present stage, the inspection work is mainly completed by manual work, that is, the staff holds a specific inspection device and inspects in the community according to the established route, and checks the equipment in turn. In some other cases, there is a solution of replacing manual work with robots, that is, using robots to check the equipment in turn according to the established route. This greatly reduces the dependence on manpower.
[0038] However, the existing inspection methods, whether relying on manpower or robots, have basically the same core principle. The main defect is that the fault is usually identified only by a single parameter threshold, the analysis logic of the fault is very simple, it is not good at fault analysis by composite parameters, and it lacks the ability to cope with new types of faults that have not been seen before.
[0039] In other words, for the staff, whether to perform composite parameter fault analysis and whether to cope with special situations mainly depend on the ability and experience of the staff. The effect is not stable, and misjudgment is easy to occur, and the accuracy is insufficient. For robots, their judgment ability depends on the inherent calculation logic, and it is still difficult to meet the needs in complex situations at the present stage.
[0040] Moreover, the inspection efficiency is largely dependent on the inspection path, and the efficiency cannot be maximized. Because in the community scene, the inspection targets are often scattered and distributed, and the selection of the road may also be subject to many restrictions, so many times the optimal inspection path cannot be accurately designed, resulting in low efficiency.
[0041] Therefore, the present application provides a property equipment data analysis method based on artificial intelligence. Referring to Figure 1As shown, a specific embodiment of the property equipment data analysis method based on artificial intelligence provided by the application is shown. In this embodiment, the method is applied to a data analysis system, which is an analysis system used in intelligent property services. The system covers hardware, software, databases, and integrates artificial intelligence technology, Internet of Things technology, and wireless communication technology. The method comprises:
[0042] Step 101, collecting multi-modal sensor data using a sensor assembly and preprocessing the multi-modal sensor data to obtain data to be analyzed.
[0043] The sensor assembly is a part of the hardware components in the data analysis system, mainly deployed on or near the property equipment. The sensor assembly collects multi-modal sensor data and transmits it back through Internet of Things technology and wireless communication technology. The multi-modal sensor data can specifically include image data, audio data, and infrared thermal imaging data. The multi-modal sensor data can reflect whether the property equipment has overheating, noise, or obvious visual faults. In other words, the problems it can reflect are highly similar to the intuitive inspection of manual inspection, thereby playing a monitoring role in the property equipment to some extent.
[0044] After collecting the multi-modal data, further preprocessing is required. The preprocessing process mainly includes data desensitization processing and feature compression processing of the multi-modal sensor data. Data desensitization processing, such as erasing sensitive information such as faces and license plates in image data. Feature compression processing, such as JPEG2000 image encoding. After preprocessing, an incremental data set can be obtained, and the incremental data set is used as the data to be analyzed.
[0045] Step 102, inputting the data to be analyzed into an edge computing model to make the edge computing model determine an analysis result.
[0046] In this embodiment, there are two operation models based on artificial intelligence technology, a cloud computing model and an edge computing model. The cloud computing model is a large model (such as an LLM model) deployed in a cloud platform, mainly responsible for data training, with multi-modal fusion and decision-making capabilities. It requires high computing resources and cannot be deployed to the edge. On the contrary, the edge computing model is a small model, or a lightweight model, obtained by knowledge distillation from the cloud computing model, which can be adapted to embedded devices (such as ONNX Runtime format devices) and can inherit the decision logic inside the cloud computing model. The edge computing model is usually deployed at the edge and can use its inherited decision logic to directly operate and analyze the data to be analyzed.
[0047] Generally, the cloud computing model needs to be first completed by building and data training. Then the cloud computing model is knowledge distilled to obtain an edge computing model, and is deployed to the edge end for application. Moreover, the edge computing model can be periodically updated through the knowledge distillation operation of the cloud computing model, so as to realize continuous reinforcement learning, continuously improve the function, add new decision logic, and cope with more complex conditions.
[0048] In actual application process, the to-be-analyzed data can be directly input to the edge computing model. The edge computing model determines the analysis result through calculation. As known, the multi-modal sensing data can reflect whether the property equipment exists overheating, noise abnormal sound, or obvious visual failure. In other words, the problems it can reflect are highly similar to the intuitive inspection of artificial inspection, so as to play a monitoring role on the property equipment to a certain extent. Therefore, the edge computing model can determine whether the property equipment exists a fault and what type of fault exists through the to-be-analyzed data. For example, overheating, abnormal sound, and various visual failures such as appearance damage often correspond to a certain type of fault. Further, the superposition of different abnormal conditions is more likely to correspond to a more complex fault type. The edge computing model can show such faults through the analysis result.
[0049] Step 103, determining a disposal scheme for the property equipment according to the analysis result.
[0050] After obtaining the analysis result, it is clear whether the property equipment exists a fault and the specific type of the fault. According to the specific type of the fault, the disposal scheme for the property equipment can be determined, which specifically includes a maintenance scheme, a replacement scheme, and the like. The staff can dispose the property equipment according to the disposal scheme.
[0051] In addition, after the disposal scheme is executed, the execution result (such as "abnormal sound is eliminated after bearing replacement") can be further fed back to the cloud computing model, so that the cloud computing model performs reinforcement learning of the knowledge graph. Thus, the cloud computing model can understand the actual effect of the disposal scheme, and further determine whether the analysis result and the disposal scheme are accurate and whether the problems existing in the property equipment are actually solved, so as to facilitate the cloud computing model to further perform reinforcement learning and update the knowledge graph, thereby improving the function.
[0052] From the above cases, it can be known that if the actual execution result is "abnormal sound is eliminated after bearing replacement", it indicates that the analysis result is accurate and the treatment scheme is effective. On the contrary, if the abnormal sound is not eliminated after the bearing replacement, it indicates that the analysis result may be incorrect or the treatment scheme is ineffective. In this case, manual analysis and treatment can be intervened to determine the actual fault and effective treatment scheme, and then fed back to the cloud computing model (equivalent to providing labeled data) to make it further learn and train data to better cope with similar situations next time.
[0053] From the above technical solutions, it can be known that the embodiment has the beneficial effects that: the computing model based on artificial intelligence technology is used to analyze the multi-modal data, so as to determine the fault of the property equipment and give a treatment scheme, thereby replacing manual inspection, saving human resources, improving the work efficiency of property services, realizing the standardization of analysis, and improving the accuracy without relying on the experience of the staff.
[0054] Figure 1 The method described in the present application is only a basic embodiment, and certain optimization and expansion can be made on the basis thereof to obtain other preferred embodiments of the method.
[0055] As shown in Figure 2 , another specific embodiment of the property equipment data analysis method based on artificial intelligence is provided. The embodiment is further described on the basis of the foregoing embodiment. In the embodiment, the method comprises the following steps:
[0056] In step 201, multi-modal sensing data is collected by a sensing assembly, and the multi-modal sensing data is preprocessed to obtain data to be analyzed.
[0057] The sensing assembly is a part of the hardware components of the data analysis system, which is mainly deployed on or near the property equipment. The sensing assembly collects multi-modal sensing data and transmits it back through Internet of Things technology and wireless communication technology. The multi-modal sensing data can specifically include image data, audio data, and infrared thermal imaging data. The multi-modal sensing data can reflect whether the property equipment has overheating, noise, or obvious visual faults. In other words, the problems it can reflect are highly similar to the direct observation of manual inspection, thereby playing a monitoring role in the property equipment to a certain extent.
[0058] After the multi-modal data is collected, further preprocessing is required. The preprocessing process mainly includes data desensitization processing and feature compression processing of the multi-modal sensor data. Data desensitization processing, such as erasing sensitive information such as faces and license plates in image data. Feature compression processing, such as JPEG2000 image encoding. After preprocessing, an incremental data set can be obtained, and the incremental data set is used as the data to be analyzed.
[0059] Step 202, input the data to be analyzed into the edge computing model, so that the edge computing model determines the analysis result.
[0060] In this embodiment, there are two operation models based on artificial intelligence technology, a cloud computing model and an edge computing model. The cloud computing model is a large model (such as an LLM model) deployed in a cloud platform, mainly responsible for data training, multi-modal fusion, and decision-making capabilities. It requires high computing resources and cannot be deployed to the edge. On the contrary, the edge computing model is a small model, or lightweight model, obtained by knowledge distillation from the cloud computing model, which can be adapted to embedded devices (such as ONNX Runtime format devices) and can inherit the decision logic inside the cloud computing model. The edge computing model is usually deployed on the edge and can use its inherited decision logic to directly operate and analyze the data to be analyzed.
[0061] Generally, the cloud computing model needs to be first built and trained. Then, the cloud computing model is knowledge distilled to obtain the edge computing model and deployed to the edge for application. Moreover, the edge computing model can be periodically updated through knowledge distillation operation of the cloud computing model, thereby realizing continuous reinforcement learning, constantly improving functions, adding new decision logic, and coping with more complex situations.
[0062] In actual application, the data to be analyzed can be directly input into the edge computing model. The edge computing model determines the analysis result after calculation. As mentioned above, multi-modal sensor data can reflect whether the property equipment is overheating, noisy, or has obvious visual faults. In other words, it can reflect problems similar to visual inspection by manual inspection, thereby monitoring the property equipment to a certain extent. Therefore, the edge computing model can determine whether the property equipment has a fault and what type of fault exists through the data to be analyzed. For example, overheating, abnormal noise, and various visual faults such as appearance damage often correspond to a certain type of fault. Further, the superposition of different abnormal conditions is more likely to correspond to a more complex fault type. The edge computing model can show such faults through the analysis result.
[0063] Step 203, determining a treatment scheme for the property equipment according to the analysis result.
[0064] After obtaining the analysis result, it is determined whether the property equipment has a fault and the specific type of the fault. According to the specific type of the fault, a treatment scheme for the property equipment can be determined, including a repair scheme, a replacement scheme, and the like. The staff can dispose the property equipment according to the treatment scheme.
[0065] Step 204, when a new fault mode is detected, the cloud computing model updates the adapter layer parameters using the parameter efficient fine-tuning technology, and creates a processing node in the knowledge graph, and associates a treatment rule with the processing node.
[0066] It can be understood that the property equipment can have various fault modes, and new fault modes can appear at any time with long-term application. Therefore, the computing model is difficult to learn all fault types and establish all treatment rules in the early training process. Reinforcement learning needs to be continuously carried out in the subsequent application process. The reinforcement learning process mainly occurs in the cloud computing model. After the cloud computing model completes the reinforcement learning, the edge computing model needs to be updated through the knowledge distillation technology.
[0067] For example, as known in the prior art, after the execution of the treatment scheme, the execution result can be fed back to the cloud computing model, so that the cloud computing model performs reinforcement learning of the knowledge graph. Thus, the cloud computing model can understand the actual effect of the treatment scheme, and further determine whether the analysis result and the treatment scheme are accurate and whether the problems existing in the property equipment are actually solved, so as to facilitate the cloud computing model to perform further reinforcement learning and update the knowledge graph, thereby improving its function.
[0068] In the embodiment, the cloud computing model can update the adapter layer (Adapter Layers) parameters using the parameter efficient fine-tuning technology (PEFT), and create a processing node in the knowledge graph, and associate a treatment rule with the processing node. The cloud computing model thus completes the reinforcement learning and updates the decision logic in itself.
[0069] Step 205, when the cloud computing model associates a treatment rule with the processing node, the cloud computing model updates the treatment rule to the edge computing model through knowledge distillation operation.
[0070] After the cloud computing model associates a treatment rule with the newly added processing node and updates the decision logic in itself, the new treatment rule needs to be updated to the edge computing model through knowledge distillation, so that the decision logics of the two are synchronized, and the edge computing model can also perform operation according to the new decision logic.
[0071] The expression of the knowledge distillation operation is:
[0072]
[0073] Wherein, L KD represents the knowledge distillation loss; T represents the temperature coefficient, and T>1; KL represents the KL divergence; sigma is a softmax function; z t is the original output of the cloud computing model; z s is the original output of the edge computing model.
[0074] And it also needs to be explained that in the updating process, the edge computing model receives an update package from the cloud computing model and performs hot updating by using a dynamic module loading technology. Hot updating is also non-stop updating, which means that the ongoing operation will not be affected during the updating process, and the service will not be interrupted.
[0075] As Figure 3 shown, it is a specific embodiment of the property equipment data analysis device based on artificial intelligence. The device in this embodiment is an entity device for executing Figures 1-2 the method. The device is placed in a data analysis system. Its technical solution is essentially consistent with the above-embodiment, and the corresponding description in the above-embodiment is also applicable to this embodiment. The device in this embodiment includes:
[0076] A data acquisition module 301 is configured to acquire multi-modal sensing data by using a sensing component, and pre-process the multi-modal sensing data to obtain to-be-analyzed data.
[0077] A data analysis module 302 is configured to input the to-be-analyzed data into an edge computing model, so that the edge computing model determines an analysis result; the edge computing model is periodically updated by a cloud computing model through a knowledge distillation operation.
[0078] A disposal module 303 is configured to determine a disposal scheme for the property equipment according to the analysis result.
[0079] In addition, on the basis of the embodiment shown in Figure 3 , preferably, it further includes:
[0080] Preferably, the multi-modal sensing data includes image data, audio data, and infrared thermal imaging data; the data acquisition module 301 includes:
[0081] An acquisition unit 311 is configured to acquire multi-modal sensing data.
[0082] The preprocessing unit 312 is configured to perform data desensitization processing and feature compression processing on the multi-modal sensing data, and obtain an incremental data set; and the incremental data set is used as the to-be-analyzed data.
[0083] Further comprising:
[0084] The first updating module 304 is configured to, when a new fault mode is detected, update the adapter layer parameters of the cloud computing model by using a parameter efficient fine-tuning technology, and create a processing node in a knowledge graph, and associate a disposal rule with the processing node.
[0085] The second updating module 305 is configured to, when the cloud computing model associates the disposal rule with the processing node, update the disposal rule to the edge computing model by using a knowledge distillation operation.
[0086] The expression of the knowledge distillation operation is as follows:
[0087]
[0088] Wherein, L KD represents a knowledge distillation loss; T represents a temperature coefficient, and T>1; KL represents a KL divergence; sigma is a softmax function; z t is an original output of the cloud computing model; z s is an original output of the edge computing model.
[0089] The feedback module 306 is configured to, after the disposal scheme is executed, feed back an execution result to the cloud computing model, so that the cloud computing model performs reinforcement learning on the knowledge graph.
[0090] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. At the hardware level, the electronic device includes a processor, and optionally further includes an internal bus, a network interface, and a memory. The memory can include a memory such as a random-access memory (RAM), and can also include a non-volatile memory such as at least one disk memory. Of course, the electronic device can also include other hardware required by a business.
[0091] The processor, the network interface and the memory can be connected with each other through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus or the like. The bus can be divided into an address bus, a data bus, a control bus and the like. For the convenience of representation, Figure 4 Only one bidirectional arrow is used to represent the bus, but it does not mean that there is only one bus or only one type of bus.
[0092] The memory is used to store execution instructions. Specifically, the execution instructions are computer programs that can be executed. The memory can include an internal memory and a non-volatile memory, and provide the processor with execution instructions and data.
[0093] In a possible implementation manner, the processor reads corresponding execution instructions from the non-volatile memory into the internal memory and then runs, or obtains corresponding execution instructions from other devices, to form the property equipment data analysis apparatus based on artificial intelligence at a logical level. The processor executes the execution instructions stored in the memory, so as to realize the property equipment data analysis method based on artificial intelligence provided in any embodiment of the present application through the executed execution instructions.
[0094] The above-mentioned property equipment data analysis apparatus based on artificial intelligence of the present application Figure 3The method executed by the property equipment data analysis device based on artificial intelligence provided by the embodiment shown can be applied to a processor or implemented by the processor. The processor can be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software. The processor mentioned above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. Each method, step and logic block disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor.
[0095] The steps of the method disclosed in the embodiment of the present application can be directly embodied as hardware decoding processor execution completion, or executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a random memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium in the memory is read by the processor, and the hardware thereof is combined to complete the steps of the above method.
[0096] The embodiment of the present application further proposes a readable medium, which stores execution instructions. When the stored execution instructions are executed by the processor of the electronic device, the electronic device can execute the property equipment data analysis method based on artificial intelligence provided in any embodiment of the present application, and is specifically used to execute the method shown in the above embodiment. Figure 1 Or Figure 2 The method shown in the above embodiment.
[0097] The electronic device described in each of the above embodiments can be a computer.
[0098] Those skilled in the art should understand that the embodiments of the present application can be provided as a method or a computer program product. Therefore, the present application can adopt a complete hardware embodiment, a complete software embodiment, or a combination of software and hardware.
[0099] The various embodiments in the present application are described in a progressive manner, and the same or similar parts among the various embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.
[0100] It should also be noted that the terms "comprising", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a list of elements not only includes those elements, but also includes other elements not explicitly listed, or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.
[0101] The above only describes the embodiments of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.
Claims
1. An artificial intelligence-based property equipment data analysis method, characterized by, The method is applied to a data analysis system, and comprises the following steps: Collecting multi-modal sensing data by using a sensing component, and preprocessing the multi-modal sensing data to obtain to-be-analyzed data; Inputting the to-be-analyzed data into an edge computing model to make the edge computing model determine an analysis result; the edge computing model is periodically updated by a cloud computing model through knowledge distillation operation; According to the analysis result, determining a disposal scheme for a property equipment.
2. The method of claim 1, wherein, The multi-modal sensing data comprises image data, audio data, and infrared thermal imaging data; The preprocessing of the multi-modal sensing data to obtain to-be-analyzed data comprises the following steps: The multi-modal sensing data is subjected to data desensitization processing and feature compression processing to obtain an incremental data set; The incremental data set is used as the to-be-analyzed data.
3. The method of claim 1, wherein, Further comprising the following steps: When a new fault mode is detected, the cloud computing model updates the adapter layer parameters by using a parameter efficient fine-tuning technology, and creates a processing node in a knowledge graph, and associates a disposal rule with the processing node.
4. The method of claim 3, wherein, The periodic updating of the edge computing model by the cloud computing model through the knowledge distillation operation comprises the following steps: When the cloud computing model associates the disposal rule with the processing node, the cloud computing model updates the disposal rule to the edge computing model through the knowledge distillation operation.
5. The method of claim 4, wherein, The expression of the knowledge distillation operation is as follows: wherein L KD represents the knowledge distillation loss; T represents a temperature coefficient, and T>1; KL represents the KL divergence; s is a softmax function; z t is the original output of the cloud computing model; z s is the original output of the edge computing model.
6. The method of claim 5, wherein, The periodic updating of the edge computing model by the cloud computing model through the knowledge distillation operation comprises the following steps: The edge computing model receives an update package from the cloud computing model, and performs hot updating by using a dynamic module loading technology.
7. The method according to any one of claims 1 to 6, characterized in that Further comprising the following steps: After the disposal scheme is executed, the execution result is fed back to the cloud computing model, so that the cloud computing model performs reinforcement learning of the knowledge graph.
8. An artificial intelligence-based property equipment data analysis device, characterized by, The device is placed in a data analysis system, and comprises the following steps: A data collection module is configured to collect multi-modal sensing data by using a sensing component, and to preprocess the multi-modal sensing data to obtain to-be-analyzed data; A data analysis module is configured to input the to-be-analyzed data into an edge computing model to make the edge computing model determine an analysis result; the edge computing model is periodically updated by a cloud computing model through knowledge distillation operation; A disposal module is configured to determine a disposal scheme for a property equipment according to the analysis result. 9.A computer readable storage medium, the storage medium storing a computer program, the computer program being configured to execute the artificial intelligence-based property equipment data analysis method according to any one of claims 1 to 7. 10.An electronic device, the electronic device comprising: a processor; a memory configured to store executable instructions for the processor; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the artificial intelligence-based property equipment data analysis method according to any one of claims 1 to 7.