A system for quickly solving sound, light and electricity content after-sales of an exhibition hall based on artificial intelligence technology

Abstract

The application belongs to the technical field of artificial intelligence and intelligent control of exhibition hall, and particularly relates to a system for quickly solving the after-sales of sound, light and electricity content of an exhibition hall based on artificial intelligence technology, comprising the following steps: S1, dividing multi-source perception data into a first data set and a second data set; S2, differential fusion processing to generate first difference set data and first intersection data; S3, dynamic permission calculation to generate a permission boundary vector; S4, executing a scene control instruction to structureally extract conversation data into a fault quadruple knowledge; S5, comparative analysis to generate second difference set data and second intersection data, and updating a fault prediction model. The application can realize the adaptive adjustment, predictive maintenance and self-repairing capability of the sound, light and electricity after-sales system of the exhibition hall through the hierarchical distribution processing of multi-source perception data, differential fusion abnormality identification, dynamic permission boundary vectorization and fault knowledge graph incremental evolution and other technical means.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence and intelligent control technology for exhibition halls, specifically relating to a system based on artificial intelligence technology for quickly resolving after-sales issues related to the audio-visual content of exhibition halls. Background Technology

[0002] With the rapid development of the exhibition and display industry, museums, corporate showrooms, science and technology museums, and other venues are increasingly reliant on audio-visual systems. Existing exhibition hall control systems typically employ a centralized architecture, managing lighting, sound, projection, and other equipment through a central control terminal. Some advanced systems incorporate sensor networks, enabling automated scene switching based on environmental changes, such as adjusting lighting brightness according to personnel presence or controlling equipment start-up and shutdown via preset schedules. Regarding after-sales maintenance, current technologies largely employ a combination of regular inspections and fault reporting; when equipment malfunctions, maintenance personnel conduct on-site troubleshooting and repairs.

[0003] In recent years, artificial intelligence technology has begun to be applied in the field of exhibition hall control. Some systems have introduced machine learning algorithms, which can predict the operating status of equipment based on historical data, or achieve human-machine control through voice interaction. Dual-terminal control architectures are also becoming increasingly common, typically employing a central control terminal and a mobile robot terminal in conjunction to achieve wider scene coverage. Regarding fault diagnosis, existing technical solutions are mostly based on matching using pre-set rule bases, triggering corresponding alarms or repair processes when abnormal parameters are detected. Some systems support remote diagnostic functions, allowing maintenance personnel to access the system via the network for troubleshooting.

[0004] Problems with existing technology: However, existing technical solutions still have many shortcomings. First, regarding access control, the permission boundaries between the central terminal and the robot terminal are usually fixed and preset, unable to be dynamically adjusted according to the risk level of the scenario, personnel characteristics, and equipment health. This leads to either delayed response or the risk of loss of control in complex scenarios. Second, regarding fault diagnosis, existing systems rely on pre-built rule bases and cannot continuously learn and evolve from manual repair experience. When encountering new fault types, they often cannot be effectively identified, and personnel's long-term reliance on AI systems leads to insufficient accumulation of their own experience, creating a vicious cycle. Third, regarding data processing, existing solutions mostly use single threshold judgments or simple statistical analysis, lacking hierarchical and diversity processing and differential fusion analysis of multi-source sensing data. This makes it difficult to capture early hidden fault characteristics of equipment and cannot achieve true predictive maintenance. Finally, regarding knowledge accumulation, the internal dialogue data between terminals and human-machine dialogue data are not effectively structured and utilized. The experience of each fault repair cannot be transformed into reusable knowledge, resulting in repeated troubleshooting when the same problem recurs, leading to low after-sales efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a system for quickly resolving after-sales issues related to the audio-visual content of exhibition halls based on artificial intelligence technology. This system can achieve adaptive adjustment, predictive maintenance, and self-repair capabilities of the audio-visual after-sales system through hierarchical and discrete processing of multi-source sensing data, differential fusion anomaly identification, dynamic permission boundary vectorization, and incremental evolution of fault knowledge graphs. This will improve after-sales response efficiency, reduce operation and maintenance costs, and form a positive cycle of human-machine collaborative evolution.

[0006] The specific technical solution adopted by this invention is as follows: A method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology includes the following steps: S1. Collect multi-source sensing data through the global sensing layer, and divide the multi-source sensing data into a first data set and a second data set. The first data set includes environmental status data, and the second data set includes personnel characteristic data and equipment operation data. S2. Perform differential fusion processing on the first data set and the second data set to generate a first difference set data and a first intersection set data, wherein the first difference set data is used to characterize abnormal deviation features, and the first intersection set data is used to characterize common stable features. S3. Based on the first difference set data and the first intersection data, a permission boundary vector between the central control terminal and the AI ​​control robot terminal is generated through the dynamic permission calculation module. The permission boundary vector includes the effective threshold of the control command, the response priority, and the emergency takeover triggering condition. S4. Execute scene control commands through the interactive service layer, and collect internal dialogue data between terminals and human-computer dialogue data during the execution process. Extract the dialogue data into a structured fault quadruple knowledge, which includes fault phenomena, diagnostic logic, repair actions and effect verification fields. S5. Compare and analyze the newly generated fault quadruple knowledge with the historical fault knowledge base to generate the second difference set data and the second intersection data. Identify the new fault types based on the second difference set data, confirm the common repair strategies based on the second intersection data, and update the fault prediction model.

[0007] According to another aspect of the present invention, the multi-source sensing data is divided into a first data set and a second data set, including: Based on the time dimension of data collection, the multi-source sensing data is divided into a static reference dataset and a dynamically changing dataset. Based on the spatial dimension of data collection, the multi-source sensing data is divided into fixed node datasets and mobile node datasets. Based on the modal dimension of data acquisition, the multi-source sensing data is divided into acoustic region dataset, optical region dataset and electromagnetic region dataset; The static reference dataset, dynamically changing dataset, fixed node dataset, moving node dataset, acoustic region dataset, optical region dataset, and electromagnetic region dataset form a multi-level data hierarchy structure.

[0008] According to another aspect of the present invention, the differential fusion process includes: Perform a first feature extraction on the first data set to generate a first feature vector set; Perform a second feature extraction on the second data set to generate a second feature vector set; Calculate the Euclidean distance between the first feature vector set and the second feature vector set, and generate a difference score; When the difference score exceeds a preset deviation threshold, the corresponding data segment is marked as the first difference set data; When the difference score is lower than the preset convergence threshold, the corresponding data segment is marked as the first intersection data; The first difference set data is amplified to enhance the discriminative power of abnormal features.

[0009] According to another aspect of the present invention, generating the permission boundary vector includes: Construct a scenario risk assessment function, with input parameters including the weight coefficients of the first difference set data; Construct a personnel feature weight vector, with input parameters including the visitor's identity attributes, behavioral preferences, and physical characteristics; Construct a device health degradation curve, with input parameters including the historical trend of the device's operating data; The Pareto optimal solution of the permission boundary vector is calculated using a multi-objective optimization algorithm; The Pareto optimal solution is mapped to the threshold for the activation of control commands, the tolerance for response delay, and the emergency takeover triggering condition.

[0010] According to another aspect of the present invention, the dialogue data is structured and extracted into fault quadruple knowledge, including: Semantic parsing technology is used to identify fault description fragments, diagnostic reasoning fragments, operation instruction fragments, and effect feedback fragments in dialogue data. By combining the fault description fragment, diagnostic reasoning fragment, operation instruction fragment, and effect feedback fragment into a complete fault handling chain through timestamp alignment and causal correlation analysis; The fault handling chain is stored in a graph structure and labeled with confidence scores and applicable scenario tags; When the similarity between a newly generated fault handling chain and a historical fault handling chain is lower than a preset threshold, a manual intervention request is triggered.

[0011] According to another aspect of the present invention, the comparative analysis includes: The newly generated fault quadtuple knowledge forms a third data set; The historical fault knowledge base will form the fourth data set; Calculate the third difference between the third data set and the fourth data set to identify new fault types that are not recorded in the historical knowledge base; Calculate the third intersection of the third data set and the fourth data set to identify common fault types that already exist in the historical knowledge base; The newly added fault types in the third difference set are prioritized and corresponding warning instructions are generated.

[0012] According to another aspect of the present invention, a method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology further includes step S6: The AI-controlled robot terminal collects electrical fingerprint data from acoustic, optical, and electronic devices within a preset range during movement. The electrical fingerprint data includes the current harmonic characteristics, voltage fluctuation spectrum, and acoustic response characteristics of the devices during operation. The electronic fingerprint data is divided into a fifth data set; The fifth data set is fused and analyzed with the first data set to generate the fourth difference set data; Based on the fourth difference set data, early latent fault characteristics of the device are identified, and predictive maintenance instructions are generated when the latent fault characteristics exceed a preset health threshold.

[0013] According to another aspect of the present invention, a method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology further includes step S7: Construct a coupled model of personnel characteristics, scene parameters, and equipment losses; The personnel characteristic data in the second dataset is input into the coupling model to generate an assessment value of the impact of scene control commands on equipment lifespan. The impact assessment value is weighted and fused with the first difference set data to generate a comprehensive optimization instruction; The comprehensive optimization instructions are used to dynamically balance user experience and operational costs.

[0014] According to another aspect of the present invention, in any of steps S1 to S7, the data set is divided in a manner including at least one of time series division, spatial region division, data mode division, and processing level division. In any of steps S2, S5, and S6, the generation methods of the difference set data and intersection data include at least one of set operations, vector space mapping, and graph structure comparison. In any of steps S3, S4, and S7, the update methods for the permission boundary vector, fault quadruple knowledge, and coupling model include at least one of incremental learning, full reconstruction, and transfer learning.

[0015] According to another aspect of the present invention, a system for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology is also provided, comprising: The global perception layer is used to perform data acquisition in steps S1 and S6. The AI ​​brain layer is used to perform data processing and model calculations in steps S2, S3, S5, and S7. The interaction service layer is used to perform dialogue data acquisition and instruction execution in step S4. The global perception layer, AI brain layer, and interactive service layer are connected by a data bus to form a closed-loop control and self-learning feedback loop.

[0016] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method described in any one of the foregoing.

[0017] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.

[0018] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any one of the preceding embodiments.

[0019] The technical effects achieved by this invention are as follows: This invention significantly improves data processing accuracy and anomaly identification capabilities through hierarchical diversity processing and differential fusion analysis of multi-source sensing data. The multi-source sensing data is divided into a first data set and a second data set. By calculating the difference score between feature vector sets, a first difference set and a first intersection set are generated. The difference set data represents abnormal deviation features, while the intersection set represents common stable features. Amplifying the difference set data enhances the distinguishability of weak abnormal signals, enabling the identification of early-stage latent fault characteristics before a fault occurs. This data processing method is not limited to specific sensors or data formats, employing various generation methods such as set operations, vector space mapping, and graph structure comparison. Its protection scope covers possible future technological variations, demonstrating strong foresight and adaptability.

[0020] This invention achieves scenario-adaptive access control and continuous evolution of fault diagnosis through dynamic permission boundary vectorization and incremental evolution of the fault knowledge graph. A multi-objective optimization algorithm calculates the Pareto optimal solution for the permission boundary vector, adjusting the permission configuration between the central control terminal and the AI-controlled robot terminal in real time to balance security, response efficiency, and operational costs. Simultaneously, it extracts structured fault quadruple knowledge from inter-terminal dialogues and human-machine interactions. Through difference and intersection analysis of the old and new knowledge bases, the knowledge graph continuously evolves, enabling the system to autonomously learn from manual repair experience. This addresses the technical problem of insufficient experience due to long-term reliance on AI, forming a positive cycle of human-machine collaborative evolution.

[0021] This invention upgrades the after-sales maintenance of showroom audio-visual equipment from post-fault response to pre-fault prevention through mobile edge diagnostic closed-loop and predictive maintenance, significantly reducing the risk of showroom operational interruptions and maintenance costs. An AI-controlled robot terminal collects electrical fingerprint data from audio-visual equipment within a preset range during movement, including current harmonic characteristics, voltage fluctuation spectrum, and acoustic response characteristics. This data is fused and analyzed with fixed node data from the full-domain perception layer, forming a complementary perception network of fixed and mobile devices, enhancing the early detection capability of latent faults. By constructing a coupled model of personnel characteristics, scene parameters, and equipment wear, the impact of scene control commands on equipment lifespan is simultaneously evaluated when generating scene control commands, dynamically balancing and optimizing between user experience and maintenance costs. Through closed-loop control and self-learning feedback loops, control strategies can be autonomously optimized without human intervention, significantly reducing maintenance labor costs and achieving automation, intelligence, and self-evolution of the showroom's audio-visual after-sales system. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method in this invention; Figure 2 This is a schematic diagram of the system structure in this invention. Detailed Implementation

[0023] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0024] It should be noted that the terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] According to an embodiment of the present invention, a method embodiment for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0026] like Figure 1 As shown, a method for quickly resolving after-sales issues related to the audio-visual content of an exhibition hall based on artificial intelligence technology includes the following steps: S1. Collect multi-source sensing data through the global sensing layer, and divide the multi-source sensing data into a first data set and a second data set. The first data set includes environmental status data, and the second data set includes personnel characteristic data and equipment operation data. S2. Perform differential fusion processing on the first data set and the second data set to generate the first difference set data and the first intersection set data, wherein the first difference set data is used to characterize abnormal deviation features, and the first intersection set data is used to characterize common stable features. S3. Based on the first difference set data and the first intersection set data, the permission boundary vector between the central control terminal and the AI ​​control robot terminal is generated through the dynamic permission calculation module. The permission boundary vector includes the effective threshold of the control command, the response priority, and the emergency takeover triggering condition. S4. Execute scene control commands through the interaction service layer, and collect internal dialogue data between terminals and human-computer dialogue data during the execution process. Extract the dialogue data into fault quadruple knowledge in a structured manner. The fault quadruple knowledge includes fault phenomenon, diagnostic logic, repair action and effect verification fields. S5. Compare and analyze the newly generated fault quadruple knowledge with the historical fault knowledge base to generate second difference set data and second intersection data. Identify new fault types based on the second difference set data, confirm common repair strategies based on the second intersection data, and update the fault prediction model.

[0027] In step S1, the global perception layer includes fixed sensor nodes and mobile perception nodes. Fixed sensor nodes are deployed in various functional areas of the exhibition hall and include temperature and humidity sensors, light intensity sensors, sound collectors, and current and voltage monitoring modules. Mobile perception nodes are integrated into the AI-controlled robot terminal and include camera modules, microphone arrays, infrared sensors, and electronic fingerprint acquisition modules.

[0028] The multi-source sensing data is divided using a multi-dimensional classification strategy. The first dataset contains environmental state data, including environmental parameters of the exhibition hall space, such as temperature distribution, humidity distribution, light intensity distribution, and background noise level. The second dataset contains personnel characteristic data, obtained through image recognition and behavior analysis, including the number of people, personnel distribution density, personnel movement trajectory, and personnel dwell time. The equipment operation data includes operating status parameters of various audio-visual equipment, such as power consumption, operating time, temperature changes, and signal strength.

[0029] Furthermore, environmental status data reflects the objective physical environment of the exhibition hall and has relative stability; personnel characteristic data and equipment operation data reflect dynamic changing factors and have higher volatility. By separating and processing these data, appropriate analysis algorithms can be used for data with different characteristics, thereby improving data processing efficiency.

[0030] Furthermore, effective classification and management of data can be achieved, providing structured input for subsequent differential fusion processing, reducing the complexity of data processing. Through the separation of data sets, there is an implicit correlation between environmental status data and personnel characteristic data. When environmental parameters change slightly, personnel behavior patterns will exhibit predictable response characteristics, which can be used to predict scene switching needs in advance.

[0031] In step S2, the core of differential fusion processing lies in identifying the differences and commonalities among the data. The first feature extraction uses time-frequency analysis to vectorize the environmental state data. The second feature extraction uses multimodal fusion to vectorize the personnel feature data and equipment operation data.

[0032] Furthermore, the difference score is calculated using a vector space distance metric. Let the first feature vector set be... The second feature vector set is Then the difference score It can be represented as: in, Denotes the vector norm, when and When they are exactly the same, , indicating no difference; when and When completely orthogonal, Approaching the maximum value indicates the greatest difference. This normalization process ensures that the difference score is within a reasonable range and is not affected by the data volume.

[0033] An adaptive method is used to set the deviation threshold and convergence threshold. The deviation threshold is used to identify abnormal deviation features. When the difference score exceeds the threshold, it indicates that the current data differs significantly from the reference state and is marked as the first difference set data. The convergence threshold is used to identify common stable features. When the difference score is below the threshold, it indicates that the current data is highly consistent with the reference state and is marked as the first intersection set data. Feature enhancement algorithms are used to amplify the first difference set data, enhancing the discriminative power of abnormal features through linear or nonlinear transformations, making weak abnormal signals easier to identify.

[0034] Furthermore, through differential fusion processing, abnormal deviation features and common stable features can be accurately identified, providing a basis for subsequent permission boundary calculations. Through amplification processing, it can be used to predict potential equipment failures and achieve early warning of failures.

[0035] In step S3, the dynamic permission calculation module calculates the permission boundary between the central control terminal and the AI-controlled robot terminal in real time based on the first difference set data and the first intersection set data. The permission boundary vector is a multi-dimensional vector, with each dimension corresponding to the effective threshold of the control command, the response priority, and the emergency takeover triggering condition.

[0036] The scenario risk assessment function is constructed based on the weighting coefficients of the first difference set data. A high proportion of the first difference set data indicates a greater number of abnormal factors in the current scenario, increasing the scenario risk level, and consequently increasing the permission weight of the central control terminal. The personnel characteristic weight vector is generated based on the visitor's identity attributes, behavioral preferences, and physical characteristics; different personnel types correspond to different permission configuration strategies. The equipment health decay curve is generated based on the historical trend of equipment operating data, reflecting the change in equipment performance over time.

[0037] Furthermore, the multi-objective optimization algorithm employs the Pareto optimal solution calculation method, assuming the objective function includes a security objective. Response efficiency target and operation and maintenance cost targets Then the Pareto optimal solution is the set of solutions that satisfy the following conditions: For any solution x, there does not exist another solution y such that... For all i, it holds, and there exists at least one j such that ; Pareto optimal solutions ensure the best balance among multiple objectives, but it is impossible to further optimize any one objective without compromising others. Mapping Pareto optimal solutions to specific parameters of the permission boundary vector enables dynamic adaptive adjustment of permissions.

[0038] Furthermore, based on the above, it is possible to dynamically calculate the permission boundaries, improve the system's adaptability to complex scenarios, and through the dynamic adjustment of the permission boundary vector, the AI-controlled robot terminal can autonomously handle more tasks in specific scenarios, reduce the intervention frequency of the central control terminal, and improve the overall system response efficiency.

[0039] In step S4, the interaction service layer includes a central control terminal, an AI-controlled robot terminal, and a human-machine interface. Inter-terminal dialogue data refers to communication data such as instruction transmission, status feedback, and permission negotiation between the central control terminal and the AI-controlled robot terminal. Human-machine dialogue data refers to data such as voice interaction, instruction input, and problem feedback between on-site personnel and the system.

[0040] Semantic parsing technology employs natural language processing methods to identify key information fragments in dialogue data. Fault description fragments contain textual or audio descriptions of the fault phenomena; diagnostic reasoning fragments contain the system's analytical logic regarding the causes of the fault; operation instruction fragments contain instructions for executing repair actions; and effect feedback fragments contain verification information about the effectiveness of the repair.

[0041] Furthermore, timestamp alignment ensures the temporal relevance of each segment. Causal correlation analysis uses logical reasoning to determine the causal relationships between segments, combining scattered segments into a complete fault handling chain. The fault handling chain is stored in a graph structure, where nodes represent segments and edges represent the relationships between segments.

[0042] The confidence score is calculated based on the reliability of the data source, logical consistency, and time sequence reasonableness. The applicable scenario label indicates the environmental conditions, equipment type, and personnel configuration applicable to this fault handling chain.

[0043] Furthermore, unstructured dialogue data is transformed into structured fault knowledge, enabling the effective accumulation and reuse of fault experience. Through the structured extraction of dialogue data, the fault repair logic has cross-scenario universality. This universal knowledge can be transferred and applied to other exhibition hall scenarios, improving the system's generalization ability.

[0044] In step S5, the newly generated fault quadruple knowledge constitutes the third data set, and the historical fault knowledge base constitutes the fourth data set. The calculation of the third difference set uses set difference operation to identify fault types that exist in the third data set but not in the fourth data set, i.e., newly added fault types. The calculation of the third intersection set uses set intersection operation to identify fault types that exist in both the third and fourth data sets, i.e., common fault types.

[0045] The priority ranking of newly added fault types is based on factors such as fault occurrence frequency, fault impact scope, fault repair difficulty, and potential fault risks. Warning commands are generated according to the priority ranking results, triggering warning notifications for high-priority newly added fault types to prompt maintenance personnel to pay close attention.

[0046] Furthermore, the fault prediction model is updated using an incremental learning method. By incorporating new fault types and corresponding repair strategies into the model training data, the model parameters are adjusted to improve the model's ability to identify new faults. Common repair strategies are identified by statistically analyzing the frequency and success rate of each repair strategy in the third intersection set, and the optimal strategy is selected as the standard repair solution.

[0047] Furthermore, based on the above, the fault knowledge base can be continuously evolved, enabling the system to learn autonomously from new faults.

[0048] As an optional embodiment, the multi-source sensing data is divided into a first data set and a second data set, including: Based on the time dimension of data collection, the multi-source sensing data is divided into static reference datasets and dynamically changing datasets; Based on the spatial dimension of data collection, the multi-source sensing data is divided into fixed node datasets and mobile node datasets. Based on the modal dimension of data acquisition, the multi-source sensing data is divided into acoustic region dataset, optical region dataset and electromagnetic region dataset; A multi-level data hierarchy is formed among the static reference dataset, the dynamically changing dataset, the fixed node dataset, the moving node dataset, the acoustic region dataset, the optical region dataset, and the electromagnetic region dataset.

[0049] Based on the above, the principle of dividing the time dimension is as follows: Static reference data reflects the baseline state of the exhibition hall, such as building layout, equipment deployment location, and basic environmental parameters. These data change slowly and can serve as long-term reference benchmarks. Dynamic data reflects the real-time state of the exhibition hall, such as personnel flow, equipment operating status, and fluctuations in environmental parameters. These data change frequently and require real-time processing.

[0050] The technical principle behind spatial dimension division lies in the following: fixed node data comes from sensors deployed in fixed locations, such as temperature and humidity sensors installed on walls and light sensors installed on ceilings; this data possesses spatial stability. Mobile node data comes from mobile devices such as AI-controlled robot terminals; this data possesses spatial mobility and can cover areas that fixed nodes cannot reach.

[0051] The principle behind modal dimension segmentation is as follows: acoustic region data includes sound intensity, frequency distribution, noise level, etc.; optical region data includes light intensity, spectral distribution, color temperature, etc.; electromagnetic region data includes current, voltage, electromagnetic field strength, etc. Different modal data reflect different operating characteristics of the equipment, and separate processing can improve the targeting of the analysis.

[0052] Furthermore, the multi-level data hierarchy is constructed using a tree-like or graph-like data structure. The first level is divided by the time dimension, the second level by the spatial dimension, and the third level by the modal dimension. The levels are linked through indexes, supporting cross-level data querying and analysis.

[0053] Furthermore, the multi-level data hierarchy supports data retrieval and analysis, improving the efficiency and accuracy of data processing. Through cross-level data analysis, faults exhibit regular characteristics under specific time, space, and modality, which can be used for accurate fault location and prediction.

[0054] As an optional embodiment, the differential fusion process includes: Extract the first feature from the first dataset to generate the first feature vector set; Perform second feature extraction on the second dataset to generate a second feature vector set; Calculate the Euclidean distance between the first feature vector set and the second feature vector set, and generate a difference score; When the difference score exceeds the preset deviation threshold, the corresponding data segment is marked as the first difference set data; When the difference score is lower than the preset convergence threshold, the corresponding data segment is marked as the first intersection data; The first difference set data is amplified to enhance the discriminative power of abnormal features.

[0055] Based on the above, the first feature extraction employs principal component analysis (PCA). Assuming the first dataset contains n data samples, and each sample contains m feature dimensions, then the feature matrix X has dimensions n×m. By calculating the covariance matrix of X, principal component vectors are extracted to generate the first feature vector set V1. The feasibility of this method lies in the fact that PCA can effectively reduce dimensionality, retain the main changes in the data, and reduce the complexity of subsequent calculations.

[0056] The second feature extraction employs a multimodal fusion method. Let the second dataset contain a subset P of personnel features and a subset E of equipment operation features; then the second feature vector set V2 can be represented as: Where f(·) represents the feature extraction function, and α and β are weight coefficients that satisfy α + β = 1. The weight coefficients can adjust the relative importance of personnel features and equipment operation features to adapt to the needs of different scenarios.

[0057] The formula for calculating Euclidean distance is: in, and Let i and ii represent the i-th components of the first and second feature vector sets, respectively. Euclidean distance is the most commonly used distance metric in vector space. It is simple to calculate and has a clear physical meaning, making it suitable for evaluating the similarity of feature vectors.

[0058] The deviation threshold and convergence threshold are set using statistical methods. The deviation threshold is set as the upper quantile of the difference score distribution, and the convergence threshold is set as the lower quantile of the difference score distribution, to ensure that the division between the first difference set data and the first intersection set data is statistically significant.

[0059] Amplification can be achieved using either linear or nonlinear amplification methods. The linear amplification formula is: in, This is the original first difference set data. This is the magnification factor. For amplified data, nonlinear amplification can be achieved using exponential or logarithmic functions, with the appropriate amplification function selected based on the characteristics of the anomalous features.

[0060] Furthermore, through feature extraction and difference scoring, accurate data classification is achieved, providing reliable input for subsequent processing.

[0061] As an optional embodiment, generating a permission boundary vector includes: Construct a scenario risk assessment function, with input parameters including the weight coefficients of the first difference set data; Construct a personnel feature weight vector, with input parameters including the visitor's identity attributes, behavioral preferences, and physical characteristics; Construct a device health degradation curve, with input parameters including historical trends of device operating data; The Pareto optimal solution of the permission boundary vector is calculated using a multi-objective optimization algorithm; The Pareto optimal solution is mapped to the threshold for the effective control command, the tolerance for response delay, and the emergency takeover trigger condition.

[0062] Based on the above, the scenario risk assessment function is constructed using a weighted summation method. Let the first difference set contain k anomalous features, and the weight coefficient for each anomalous feature be... The degree of abnormality is Then the scenario risk score It can be represented as: The degree of anomaly reflects the deviation of each feature. Weighted summation can comprehensively assess the overall risk level of the scenario. The construction of the personnel feature weight vector adopts a classification mapping method. Identity attributes include categories such as ordinary audience, VIP guests, and maintenance personnel. Different categories correspond to different permission configurations. Behavioral preferences include dimensions such as visiting speed, dwell time preference, and willingness to interact, which are obtained through historical behavioral data analysis. Physical characteristics include height, body shape, and mobility, which are obtained through image recognition.

[0063] Furthermore, the equipment health degradation curve was constructed using time series analysis. Let the historical trend of the equipment operating data be the time series. Then health It can be expressed as a function of time: in, For initial health, The attenuation coefficient is... The exponential decay model, which takes the runtime as an example, conforms to the general rule that equipment performance gradually declines over time, and the parameters can be obtained by fitting historical data.

[0064] Multi-objective optimization algorithms employ genetic algorithms or particle swarm optimization algorithms, assuming an objective function vector. F1 is the security objective, F2 is the response efficiency objective, and F3 is the operation and maintenance cost objective. The algorithm iteratively searches for the Pareto optimal solution set to ensure that the best balance is achieved among the objectives.

[0065] The specific parameters of the Pareto optimal solution mapping to the permission boundary vector adopt linear or nonlinear mapping methods. The effective threshold is positively correlated with the security goal, the response delay tolerance is negatively correlated with the response efficiency goal, and the emergency takeover triggering condition is related to the operation and maintenance cost goal.

[0066] Furthermore, through multi-objective optimization, the dynamic adaptive adjustment of permission boundaries can be achieved, balancing security, efficiency, and cost.

[0067] As an optional implementation, the dialogue data is structured and extracted into fault quadruple knowledge, including: Semantic parsing technology is used to identify fault description fragments, diagnostic reasoning fragments, operation instruction fragments, and effect feedback fragments in dialogue data. By combining timestamp alignment and causal correlation analysis, fault description fragments, diagnostic reasoning fragments, operation instruction fragments, and effect feedback fragments into a complete fault handling chain; The fault handling chain is stored in a graph structure and labeled with confidence scores and applicable scenario tags; When the similarity between a newly generated fault handling chain and a historical fault handling chain is lower than a preset threshold, a manual intervention request is triggered.

[0068] Based on the above, semantic parsing technology employs a pre-trained language model. Key words for identifying fault description fragments include "fault", "abnormality", "error", and "failure". Key words for identifying diagnostic reasoning segments include "cause," "analysis," "judgment," and "speculation." Key words for identifying operation instruction fragments include "execute", "start", "shut down", and "adjust". Key words for identifying feedback segments include "completed", "success", "failure", and "verification".

[0069] Timestamp alignment uses a time window matching method. Let the timestamps of each segment be t1, t2, t3, and t4. If the following conditions are met... If so, then fragment i and fragment j are considered to be temporally related, where Using the time window threshold, segments of the same fault handling chain usually occur consecutively within a short period of time, and time window matching can effectively identify related segments.

[0070] Furthermore, the causal relationship analysis employs logical reasoning, with fault description segments as causes and diagnostic reasoning segments as effects; diagnostic reasoning segments as causes and operational instruction segments as effects; and operational instruction segments as causes and effect feedback segments as effects. By constructing causal chains, the logical integrity of the fault handling chain is ensured.

[0071] The graph structure storage uses a directed graph representation, where nodes represent segments and edges represent the relationships between segments. Confidence scores are calculated based on the reliability of the data source, logical consistency, and time series rationality. Applicable scenario tags indicate the environmental conditions, equipment types, and personnel configurations applicable to the fault handling chain.

[0072] Similarity calculation uses graph edit distance or graph kernel method. When the similarity between the newly generated fault handling chain and the historical fault handling chain is lower than the preset threshold, it indicates that the fault is a new type, triggering a manual intervention request, and the operation and maintenance personnel confirm the repair strategy.

[0073] Furthermore, it enables structured storage and rapid retrieval of fault knowledge, improving fault repair efficiency. Through graph structure storage, the fault handling chain has reusable sub-chains, which can be independently applied to other fault scenarios, enhancing the flexibility of knowledge reuse.

[0074] As an optional embodiment, comparative analysis includes: The newly generated fault quadtuple knowledge forms a third data set; The historical fault knowledge base will form the fourth data set; Calculate the third difference between the third and fourth data sets to identify new fault types that are not recorded in the historical knowledge base; Calculate the third intersection of the third and fourth data sets to identify common fault types that already exist in the historical knowledge base; The newly added fault types in the third difference set are prioritized and corresponding warning instructions are generated.

[0075] Based on the above, the representations of the third and fourth data sets are performed using set theory. Let the third data set be... Where fᵢ represents the knowledge of the i-th fault quadruple; the fourth data set is , where gⱼ represents the j-th historical fault knowledge.

[0076] The formula for calculating the third difference is: Among them, set difference operation can accurately identify newly added fault types, ensuring that faults not recorded in the historical knowledge base are effectively identified.

[0077] The formula for calculating the third intersection is: Among them, the set intersection operation can identify common fault types and confirm that faults that already exist in the historical knowledge base reappear in the current scenario.

[0078] Priority ranking uses a multi-dimensional scoring method. Let the priority score P(f) of a newly added fault type f be expressed as: Where freq(f) is the frequency of failure occurrence, impact(f) is the scope of failure impact, difficulty(f) is the difficulty of failure repair, risk(f) is the potential risk of failure, and w1, w2, w3, w4 are weighting coefficients, satisfying... Multi-dimensional scoring can comprehensively assess the importance of new faults, ensuring that high-priority faults are handled first.

[0079] Furthermore, the generation of early warning instructions is based on the priority ranking results, and early warning notifications are triggered for fault types whose scores exceed preset thresholds. Notification methods include system pop-ups, SMS notifications, email notifications, etc.

[0080] Furthermore, it can accurately identify and prioritize new faults, ensuring that important faults are handled in a timely manner. By prioritizing faults, low-frequency faults with high potential risks can prompt the system to adjust its early warning strategy, incorporate risk dimensions into the priority assessment, and improve the accuracy of early warnings.

[0081] As an optional embodiment, a method for quickly resolving after-sales issues related to the audio-visual content of an exhibition hall based on artificial intelligence technology further includes step S6: The AI-controlled robot terminal collects electrical fingerprint data from acoustic, optical, and electronic devices within a preset range during movement. The electrical fingerprint data includes the current harmonic characteristics, voltage fluctuation spectrum, and acoustic response characteristics of the devices during operation. The electronic fingerprint data is divided into a fifth data set; The fifth dataset is merged and analyzed with the first dataset to generate the fourth difference dataset; Based on the fourth difference set data, early latent fault characteristics of the equipment are identified, and predictive maintenance instructions are generated when the latent fault characteristics exceed the preset health threshold.

[0082] Based on the above, the electronic fingerprint data acquisition utilizes high-precision sensors. Current harmonic characteristics are acquired using a current sensor, and the harmonic components of the current waveform are analyzed to identify the electrical characteristics of the device during operation. Voltage fluctuation spectra are acquired using a voltage sensor, and the frequency domain characteristics of the voltage signal are analyzed to identify the power supply stability of the device. Acoustic response characteristics are acquired using a microphone, and the sound characteristics of the device during operation are analyzed to identify the operating status of mechanical components.

[0083] The fifth dataset was divided using the same classification strategy as the first dataset to ensure data format consistency and facilitate subsequent fusion analysis.

[0084] Furthermore, the generation of the fourth difference set data adopts the same difference fusion method as in step S2. Let the fifth data set be S5 and the first data set be S1, then the fourth difference set data S_diff4 can be expressed as: S_diff4 = S5 - S1 By comparing mobile-collected electronic fingerprint data with fixed-collected environmental data, the unique operating characteristics of the equipment can be identified, and the influence of environmental interference factors can be eliminated.

[0085] The identification of early-stage latent fault characteristics of equipment adopts a pattern matching method. Let the electrical fingerprint characteristic pattern of normal equipment be M_normal, and the electrical fingerprint characteristic pattern of the current equipment be M_current. Then the latent fault characteristic F_hidden can be represented as: When F_hidden exceeds the preset health threshold, it indicates that the device has an early hidden fault, generating a predictive maintenance instruction to prompt maintenance personnel to perform preventive maintenance.

[0086] Furthermore, by using mobile edge diagnostics, we can accurately identify early-stage hidden faults in equipment, upgrading after-sales maintenance from reactive response to predictive maintenance. Through the fusion analysis of electronic fingerprint data, the characteristics of hidden faults in equipment can be identified several days before the fault occurs, significantly advancing the time window for predictive maintenance and reducing the risk of showroom operation interruption.

[0087] As an optional embodiment, a method for quickly resolving after-sales issues related to the audio-visual content of an exhibition hall based on artificial intelligence technology further includes step S7: Construct a coupled model of personnel characteristics, scene parameters, and equipment losses; Input the personnel characteristic data from the second dataset into the coupled model to generate an assessment value of the impact of scene control commands on equipment lifespan. The impact assessment value and the first difference set data are weighted and fused to generate a comprehensive optimization instruction; The comprehensive optimization command is used to dynamically balance user experience and operational costs.

[0088] Based on the above, the coupling model is constructed using multivariate regression or neural network methods. Let the personnel feature vector be P, the scene parameter vector be S, and the equipment loss vector be D, then the coupling model can be expressed as: Here, f(·) is the coupling function, which can be obtained through training with historical data. Personnel characteristics affect the selection of scene parameters, scene parameters affect the operating load of equipment, and equipment operating load affects equipment wear and tear. There is a clear causal chain among the three.

[0089] Furthermore, the generation of the impact assessment value adopts a model reasoning method, inputting the current personnel characteristic data into the coupled model and outputting the impact assessment value of the scenario control command on the equipment life. The impact assessment value can be expressed as the expected shortening ratio of the equipment life or the expected increase in maintenance costs.

[0090] The integrated optimization instruction is generated using a weighted fusion method. Let the impact assessment value be E, and the first difference set data be D_diff. Then, the integrated optimization instruction I_opt can be expressed as: Here, α and β are weighting coefficients, satisfying α + β = 1. The weighting coefficients can adjust the relative importance of user experience and operation and maintenance costs to adapt to the needs of different scenarios. The dynamic balancing adopts a feedback control method to monitor user experience indicators and operation and maintenance cost indicators in real time. When either indicator deviates from the target range, the weighting coefficients are adjusted and a comprehensive optimization instruction is regenerated to ensure that the two indicators are balanced within the target range.

[0091] Furthermore, it can achieve a dynamic balance between the user experience and maintenance costs, thereby improving the overall efficiency of exhibition hall operations. Through the coupling model, it can highlight the nonlinear relationship between personnel characteristics and equipment wear and tear, which can be used to optimize scene parameter configuration, thereby improving the user experience while reducing equipment wear and tear.

[0092] As an optional embodiment, in any of steps S1 to S7, the data set can be divided in at least one of the following ways: by time series, by spatial region, by data mode, and by processing level. In any of steps S2, S5, and S6, the generation methods of difference set data and intersection data include at least one of set operations, vector space mapping, and graph structure comparison. In any of steps S3, S4, and S7, the update methods for the permission boundary vector, fault quadruple knowledge, and coupling model include at least one of incremental learning, full reconstruction, and transfer learning.

[0093] Based on the above, the flexibility of the data set partitioning method ensures that the system can adapt to the needs of different scenarios. Time series partitioning is suitable for analyzing the changing patterns of data over time; spatial region partitioning is suitable for analyzing the spatial distribution characteristics of data; data modality partitioning is suitable for analyzing the characteristics of different types of data; and processing hierarchy partitioning is suitable for analyzing the characteristics of data at different processing stages.

[0094] The flexibility in generating difference and intersection data ensures that the system can adopt the most suitable data processing method. Set operations are suitable for processing structured data; vector space mapping is suitable for processing feature vectors; and graph structure comparison is suitable for processing relational data.

[0095] Furthermore, the update method ensures that the system can select the most suitable update strategy based on the amount of data and computing resources. Incremental learning is suitable for scenarios with a large amount of data that need to be continuously updated; full reconstruction is suitable for scenarios where the data distribution changes significantly; and transfer learning is suitable for scenarios where the new scenario is similar to the historical scenario.

[0096] Furthermore, data processing methods can be used to improve the system's adaptability and scalability. By selecting update methods, transfer learning can shorten the system training time when deployed in new exhibition hall scenarios, enabling the system to quickly adapt to new scenarios and reduce deployment costs.

[0097] As an optional embodiment, refer to Figure 2 A system based on artificial intelligence technology for quickly resolving after-sales issues related to the audio-visual content of exhibition halls, comprising: The global perception layer is used to perform data acquisition in steps S1 and S6. The AI ​​brain layer is used to perform data processing and model calculations in steps S2, S3, S5, and S7. The interaction service layer is used to perform dialogue data acquisition and instruction execution in step S4. The global perception layer, the AI ​​brain layer, and the interactive service layer are connected by a data bus to form a closed-loop control and self-learning feedback loop.

[0098] As described above, the hardware deployment of the global perception layer includes fixed sensor nodes and mobile perception nodes. Fixed sensor nodes are deployed in various functional areas of the exhibition hall, including temperature and humidity sensors, illuminance sensors, sound collectors, and current and voltage monitoring modules. Mobile perception nodes are integrated into the AI-controlled robot terminal, including camera modules, microphone arrays, infrared sensors, and electronic fingerprint acquisition modules.

[0099] The software architecture of the AI ​​brain layer includes a data processing module, a model calculation module, and a knowledge management module. The data processing module executes algorithms such as differential fusion processing and permission boundary calculation; the model calculation module executes inference calculations for fault prediction models, coupled models, etc.; and the knowledge management module executes the storage, retrieval, and updating of the fault knowledge base.

[0100] The terminals of the interactive service layer include a central control terminal, an AI-controlled robot terminal, and a human-computer interaction interface. The central control terminal is deployed in the exhibition hall control room and has the highest control authority; the AI-controlled robot terminal is a mobile device with limited control authority; and the human-computer interaction interface allows on-site personnel to interact with the system through voice or command input.

[0101] Furthermore, the data bus employs a high-speed communication protocol to ensure real-time and reliable data transmission between layers. Closed-loop control ensures the system can adjust its control strategy in real time based on feedback data; a self-learning feedback loop ensures the system can continuously learn and evolve from historical data.

[0102] Furthermore, through the collaborative work of the three-layer architecture, the exhibition hall's audio-visual after-sales system can achieve automation, intelligence, and self-evolution. By combining closed-loop control with self-learning feedback loops, the system can autonomously optimize control strategies without human intervention, enabling it to operate autonomously and significantly reducing maintenance manpower costs.

[0103] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0104] According to another aspect of the present invention, an electronic device is also provided, the electronic device including a memory and a processor; the memory is used to store a program; the processor executes the program to implement the method of any of the foregoing.

[0105] According to another aspect of the present invention, a computer-readable storage medium is also provided, the storage medium storing a computer program that, when executed by a processor, implements the method of any of the foregoing.

[0106] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described in any of the foregoing.

[0107] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for rapidly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, characterized in that: Includes the following steps: S1. Collect multi-source sensing data through the global sensing layer, and divide the multi-source sensing data into a first data set and a second data set. The first data set includes environmental status data, and the second data set includes personnel characteristic data and equipment operation data. S2. Perform differential fusion processing on the first data set and the second data set to generate a first difference set data and a first intersection set data, wherein the first difference set data is used to characterize abnormal deviation features, and the first intersection set data is used to characterize common stable features. S3. Based on the first difference set data and the first intersection data, a permission boundary vector between the central control terminal and the AI ​​control robot terminal is generated through the dynamic permission calculation module. The permission boundary vector includes the effective threshold of the control command, the response priority, and the emergency takeover triggering condition. S4. Execute scene control commands through the interactive service layer, and collect internal dialogue data between terminals and human-computer dialogue data during the execution process. Extract the dialogue data into a structured fault quadruple knowledge, which includes fault phenomena, diagnostic logic, repair actions and effect verification fields. S5. Compare and analyze the newly generated fault quadruple knowledge with the historical fault knowledge base to generate the second difference set data and the second intersection data. Identify the new fault types based on the second difference set data, confirm the common repair strategies based on the second intersection data, and update the fault prediction model.

2. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... The multi-source sensing data is divided into a first data set and a second data set, including: Based on the time dimension of data collection, the multi-source sensing data is divided into a static reference dataset and a dynamically changing dataset. Based on the spatial dimension of data collection, the multi-source sensing data is divided into fixed node datasets and mobile node datasets. Based on the modal dimension of data acquisition, the multi-source sensing data is divided into acoustic region dataset, optical region dataset and electromagnetic region dataset; The static reference dataset, dynamically changing dataset, fixed node dataset, moving node dataset, acoustic region dataset, optical region dataset, and electromagnetic region dataset form a multi-level data hierarchy structure.

3. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... The differential fusion process includes: Perform a first feature extraction on the first data set to generate a first feature vector set; Perform a second feature extraction on the second data set to generate a second feature vector set; Calculate the Euclidean distance between the first feature vector set and the second feature vector set, and generate a difference score; When the difference score exceeds a preset deviation threshold, the corresponding data segment is marked as the first difference set data; When the difference score is lower than the preset convergence threshold, the corresponding data segment is marked as the first intersection data; The first difference set data is amplified to enhance the discriminative power of abnormal features.

4. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... The generation of the permission boundary vector includes: Construct a scenario risk assessment function, with input parameters including the weight coefficients of the first difference set data; Construct a personnel feature weight vector, with input parameters including the visitor's identity attributes, behavioral preferences, and physical characteristics; Construct a device health degradation curve, with input parameters including the historical trend of the device's operating data; The Pareto optimal solution of the permission boundary vector is calculated using a multi-objective optimization algorithm; The Pareto optimal solution is mapped to the threshold for the activation of control commands, the tolerance for response delay, and the emergency takeover triggering condition.

5. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... The dialogue data is structured and extracted into fault quadruple knowledge, including: Semantic parsing technology is used to identify fault description fragments, diagnostic reasoning fragments, operation instruction fragments, and effect feedback fragments in dialogue data. By combining the fault description fragment, diagnostic reasoning fragment, operation instruction fragment, and effect feedback fragment into a complete fault handling chain through timestamp alignment and causal correlation analysis; The fault handling chain is stored in a graph structure and labeled with confidence scores and applicable scenario tags; When the similarity between a newly generated fault handling chain and a historical fault handling chain is lower than a preset threshold, a manual intervention request is triggered.

6. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that: The comparative analysis includes: The newly generated fault quadtuple knowledge forms a third data set; The historical fault knowledge base will form the fourth data set; Calculate the third difference between the third data set and the fourth data set to identify new fault types that are not recorded in the historical knowledge base; Calculate the third intersection of the third data set and the fourth data set to identify common fault types that already exist in the historical knowledge base; The newly added fault types in the third difference set are prioritized and corresponding warning instructions are generated.

7. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... It also includes step S6: The AI-controlled robot terminal collects electrical fingerprint data from acoustic, optical, and electronic devices within a preset range during movement. The electrical fingerprint data includes the current harmonic characteristics, voltage fluctuation spectrum, and acoustic response characteristics of the devices during operation. The electronic fingerprint data is divided into a fifth data set; The fifth data set is fused and analyzed with the first data set to generate the fourth difference set data; Based on the fourth difference set data, early latent fault characteristics of the device are identified, and predictive maintenance instructions are generated when the latent fault characteristics exceed a preset health threshold.

8. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that... It also includes step S7: Construct a coupled model of personnel characteristics, scene parameters, and equipment losses; The personnel characteristic data in the second dataset is input into the coupling model to generate an assessment value of the impact of scene control commands on equipment lifespan. The impact assessment value is weighted and fused with the first difference set data to generate a comprehensive optimization instruction; The comprehensive optimization instructions are used to dynamically balance user experience and operational costs.

9. The method for quickly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, as described in claim 1, is characterized in that: In any of steps S1 to S7, the data set can be divided in at least one of the following ways: by time series, by spatial region, by data mode, and by processing level. In any of steps S2, S5, and S6, the generation methods of the difference set data and intersection data include at least one of set operations, vector space mapping, and graph structure comparison. In any of steps S3, S4, and S7, the update methods for the permission boundary vector, fault quadruple knowledge, and coupling model include at least one of incremental learning, full reconstruction, and transfer learning.

10. A system for rapidly resolving after-sales issues related to audio-visual content in exhibition halls based on artificial intelligence technology, used to implement the method described in any one of claims 1 to 9, characterized in that, include: The global perception layer is used to perform data acquisition in steps S1 and S6. The AI ​​brain layer is used to perform data processing and model calculations in steps S2, S3, S5, and S7. The interaction service layer is used to perform dialogue data acquisition and instruction execution in step S4. The global perception layer, AI brain layer, and interactive service layer are connected by a data bus to form a closed-loop control and self-learning feedback loop.

Images