Intelligent vending machine fault early warning system based on cloud platform

Abstract

The invention relates to the technical field of vending machine fault early warning, and particularly discloses an intelligent vending machine fault early warning system based on a cloud platform, which is characterized in that to-be-detected key fault maintenance items of a vending machine are determined, and abnormal data information is collected by using a sensor and other data acquisition means; and the data are uploaded to the cloud for analysis. An advanced deep learning algorithm and an artificial intelligence technology are adopted on the cloud platform to perform deep significant association aggregation analysis on the abnormal data information of the vending machine, so that predictive diagnosis on possible faults is realized, and an early warning result can be provided for operation and maintenance personnel. Therefore, the automation degree and the intelligent level of fault and early warning of the vending machine can be improved, the problems of unclear fault mode understanding and poorer generalization ability caused by a one-by-one matching mode with the fault maintenance items in the prior art are avoided, and the service quality of the vending machine is ensured.

Description

Technical Field

[0001] This application relates to the technical field of vending machine fault warning, and more specifically, to an intelligent vending machine fault warning system based on a cloud platform. Background Art

[0002] With the wide application of vending machines in urban life, such automated devices provide consumers with a convenient shopping experience and offer enterprises new sales channels. However, vending machines may encounter various types of faults during actual operation, such as blocked product channels, payment system problems, etc. If these problems are not addressed in a timely manner, they will not only affect the shopping experience of consumers but may also lead to the machine being out of service for a long time, increasing maintenance costs and reducing operational efficiency.

[0003] In response to the above technical problems, Chinese Patent CN117522382B provides a vending machine fault warning method and system, which can determine possible fault inspection items and set multiple fault phenomenon characteristics for each inspection item. Then, a fault identification queue is constructed, and the abnormal data information set is matched with each fault inspection item one by one. Based on a preset fault matching threshold, possible fault items are judged and recorded, and finally integrated into a set of faults to be inspected and input into a fault judgment model to output a fault warning result. This method aims to detect and warn of possible faults in advance to ensure the stable operation of vending machines.

[0004] However, in the above vending machine fault warning method and system, there are certain limitations: First, the method of directly matching the abnormal data information set with the fault inspection items one by one is relatively rigid and may not fully consider the complex relationships between individual abnormal data and fault phenomena; Second, for unseen or newly emerging fault patterns, the fault phenomenon characteristics established based on historical data may not have sufficient generalization ability, thus affecting the warning accuracy; Finally, this solution relies on a preset fault matching threshold for fault judgment, and this static threshold setting may be difficult to adapt to changes in different environments, resulting in false alarms or missed alarms.

[0005] Therefore, an optimized vending machine fault warning system is desired. Summary of the Invention

[0006] This application provides an intelligent vending machine fault warning system based on a cloud platform, which can improve the automation degree and intelligent level of vending machine fault and warning, avoid the problems of unclear understanding of fault patterns and poor generalization ability brought by the method of matching one by one with fault inspection items in the existing technical solutions, and is beneficial to ensuring the service quality of vending machines.

[0007] In a first aspect, an intelligent vending machine fault warning system based on a cloud platform is provided, including:

[0008] A fault repair item determination module, configured to determine the fault repair items to be detected for the vending machine;

[0009] An abnormal data information acquisition module, configured to detect and obtain the abnormal data information of the vending machine based on the fault repair items to obtain a set of abnormal data information;

[0010] A data transmission module, configured to upload the set of abnormal data information to the cloud platform;

[0011] An abnormal data information semantic encoding module, configured to perform semantic encoding on the set of abnormal data information on the cloud platform to obtain a set of abnormal data information semantic embedding encoding features;

[0012] A feature dynamic compensation aggregation analysis module, configured to perform feature dynamic compensation aggregation analysis on the set of abnormal data information semantic embedding encoding features to obtain abnormal data information semantic feature significantly aggregated encoding features, including: a hub feature extraction unit, configured to extract the hub features of the set of abnormal data information semantic embedding encoding features; a feature significant aggregation unit, configured to perform feature significant aggregation on the set of abnormal data information semantic embedding encoding features based on the hub features to obtain abnormal data information semantic feature significantly aggregated encoding features;

[0013] A fault warning result generation module, configured to generate a fault warning result based on the abnormal data information semantic feature significantly aggregated encoding features.

[0014] In the above intelligent vending machine fault warning system based on the cloud platform, the fault repair item determination module is configured to:

[0015] Click on the fault repair page to pop up a fault repair list;

[0016] Specify the fault repair items from the fault repair list.

[0017] In the above intelligent vending machine fault warning system based on the cloud platform, the abnormal data information semantic encoding module is configured to:

[0018] Input each abnormal data information in the set of abnormal data information into a semantic encoder based on the Word2Vec model to obtain a set of abnormal data information semantic embedding encoding vectors as the set of abnormal data information semantic embedding encoding features.

[0019] In the above intelligent vending machine fault warning system based on the cloud platform, the hub feature extraction unit is configured to:

[0020] Input the set of semantic-embedded encoding vectors of the abnormal data information into a sequence hub extraction network to obtain an abnormal data information hub feature vector as the hub feature.

[0021] In the above intelligent vending machine fault warning system based on a cloud platform, the hub feature extraction unit is used for:

[0022] Calculate the essential extraction factors of each abnormal data information semantic-embedded encoding vector in the set of semantic-embedded encoding vectors of the abnormal data information to obtain a set of essential extraction factors of the abnormal data information;

[0023] Normalize the set of essential extraction factors of the abnormal data information to obtain a set of normalized essential extraction factors of the abnormal data information;

[0024] Use the set of normalized essential extraction factors of the abnormal data information as a set of weights to perform position-wise weighted aggregation on the set of semantic-embedded encoding vectors of the abnormal data information to obtain the abnormal data information hub feature vector.

[0025] In the above intelligent vending machine fault warning system based on a cloud platform, the feature significant aggregation unit includes:

[0026] A node-hub complementary information calculation sub-unit, which is used to calculate a set of abnormal data information node-hub complementary information embedded encoding vectors based on the abnormal data information hub feature vector;

[0027] A feature compensation aggregation sub-unit, which is used to perform attention-driven feature compensation aggregation on the set of abnormal data information node-hub complementary information embedded encoding vectors to obtain the significantly aggregated encoded feature of the abnormal data information semantic feature.

[0028] In the above intelligent vending machine fault warning system based on a cloud platform, the node-hub complementary information calculation sub-unit is used for:

[0029] Extract a first abnormal data information semantic-embedded encoding vector from the set of semantic-embedded encoding vectors of the abnormal data information;

[0030] Perform point convolution encoding on the first abnormal data information semantic-embedded encoding vector, multiply the result by a first weight matrix, and input the multiplied result into a Sigmoid function for activation to obtain a first abnormal data information semantic encoded embedded encoding modulation vector;

[0031] Perform dot convolution encoding on the abnormal data information hub feature vector, multiply the result by the second weight matrix, and input the multiplication result into the Sigmoid function for activation to obtain the abnormal data information hub feature modulation vector;

[0032] Calculate the position-wise difference between the first abnormal data information semantic encoding embedded encoding modulation vector and the abnormal data information hub feature modulation vector, and take the absolute value of the calculation result to obtain the first abnormal data information node-hub complementary information modulation weight vector;

[0033] Calculate the position-wise dot product of the first abnormal data information node-hub complementary information modulation weight vector with the first abnormal data information semantic encoding embedded encoding modulation vector and the abnormal data information hub feature vector respectively, perform position-wise addition on the calculation results, perform dot convolution encoding on the addition result, and input it into the Sigmoid function for activation to obtain the first node-hub complementary information embedded encoding vector.

[0034] In the above intelligent vending machine fault warning system based on the cloud platform, the feature compensation aggregation subunit includes:

[0035] The attention modulation secondary subunit is used to respectively perform complementary information significant marking on each abnormal data information node-hub complementary information embedded encoding vector in the set of abnormal data information node-hub complementary information embedded encoding vectors to obtain a set of abnormal data information node complementary information attention weights, and then perform attention modulation on the set of abnormal data information node-hub complementary information embedded encoding vectors with the set of abnormal data information node complementary information attention weights to obtain a set of abnormal data information significant node-hub complementary information embedded encoding vectors;

[0036] The fusion secondary subunit is used to fuse the abnormal data information hub feature vector and the set of abnormal data information significant node-hub complementary information embedded encoding vectors to obtain the abnormal data information semantic feature significant aggregation encoding feature vector as the abnormal data information semantic feature significant aggregation encoding feature.

[0037] In the above intelligent vending machine fault warning system based on the cloud platform, the fault warning result generation module is used to:

[0038] Input the abnormal data information semantic feature significant aggregation encoding feature vector into the fault judgment model based on the classifier to obtain the fault warning result.

[0039] An intelligent vending machine fault warning system based on a cloud platform provided by the present application determines key fault repair items to be detected for the vending machine, collects abnormal data information by using sensors and other data collection means, and uploads this data to the cloud for analysis. Advanced deep learning algorithms and artificial intelligence technologies are adopted on the cloud platform to perform in-depth and significant correlation aggregation analysis on this abnormal data information of the vending machine, so as to realize predictive diagnosis of possible faults and provide warning results to operation and maintenance personnel. In this way, the automation degree and intelligent level of vending machine fault and warning can be improved, avoiding the problems of unclear understanding of fault modes and poor generalization ability brought by the method of matching one by one with fault repair items in the existing technical solutions, which is beneficial to ensuring the service quality of the vending machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application and do not limit the present application.

[0041] Figure 1 It is a schematic block diagram of the intelligent vending machine fault warning system based on the cloud platform according to the embodiment of the present application.

[0042] Figure 2 It is a schematic diagram of data flow of the intelligent vending machine fault warning system based on the cloud platform according to the embodiment of the present application.

[0043] Figure 3 It is a schematic block diagram of the feature dynamic compensation aggregation analysis module in the intelligent vending machine fault warning system based on the cloud platform according to the embodiment of the present application.

[0044] Figure 4 It is a schematic block diagram of the feature significant aggregation unit in the intelligent vending machine fault warning system based on the cloud platform according to the embodiment of the present application.

[0045] Figure 5 It is a schematic block diagram of the feature compensation aggregation subunit in the intelligent vending machine fault warning system based on the cloud platform according to the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0047] Based on this, in the technical solution of this application, the technical concept of this application is to determine the key fault repair items to be detected in the vending machine, collect abnormal data information by using sensors and other data collection means, and upload this data to the cloud for analysis. Advanced deep learning algorithms and artificial intelligence technologies are adopted on the cloud platform to conduct in-depth and significant correlation aggregation analysis on these abnormal data information of the vending machine, so as to realize predictive diagnosis of possible faults and provide early warning results to the operation and maintenance personnel. In this way, the automation degree and intelligent level of vending machine fault and early warning can be improved, and the problems of unclear understanding of fault modes and poor generalization ability brought by the method of matching one by one with fault repair items in the existing technical solutions can be avoided, which is beneficial to ensuring the service quality of the vending machine.

[0048] Specifically, as Figure 1 and Figure 2 shown, the intelligent vending machine fault early warning system based on the cloud platform includes: a fault repair item determination module 10 for determining the fault repair items to be detected in the vending machine; an abnormal data information acquisition module 20 for detecting and obtaining the abnormal data information of the vending machine based on the fault repair items to obtain a set of abnormal data information; a data transmission module 30 for uploading the set of abnormal data information to the cloud platform; an abnormal data information semantic encoding module 40 for semantically encoding the set of abnormal data information on the cloud platform to obtain a set of abnormal data information semantic embedding encoding features; a feature dynamic compensation aggregation analysis module 50 for performing feature dynamic compensation aggregation analysis on the set of abnormal data information semantic embedding encoding features to obtain a set of abnormal data information semantic feature significant aggregation encoding features; and a fault early warning result generation module 60 for generating a fault early warning result based on the set of abnormal data information semantic feature significant aggregation encoding features.

[0049] Exemplarily, in the fault repair item determination module 10, the fault repair items to be detected in the vending machine are determined. Specifically, first, click on the fault repair page to pop up a fault repair list; wherein, the fault repair list is obtained from the collection of historical fault information. By collecting the historical fault information of the vending machine, it is possible to monitor key areas targeted, which can make subsequent analysis and judgment more accurate, and thus improve the accuracy of vending machine fault diagnosis. In one example, a special "fault repair" option can be set on the operation interface of the vending machine. When this option is selected, a fault repair list will pop up, and further specify the fault repair items from the fault repair list. That is, in one embodiment, the fault repair item determination module is used to: click on the fault repair page to pop up a fault repair list; and specify the fault repair items from the fault repair list.

[0050] Exemplarily, in the abnormal data information acquisition module 20, based on the fault repair item, the abnormal data information of the vending machine is detected and obtained to obtain a set of abnormal data information. It should be understood that when the vending machine is in the running state, the system will continuously monitor various parameters and save these real-time data together with the time stamp to the local or cloud log record. That is, the log record contains all the important events and operation details of the vending machine since it was started. These information not only reflect the normal working condition of the machine, but also hide the potential signals that may cause faults. On this basis, through the abnormal data information of the vending machine, the data points that are most likely to cause fault anomalies during the operation of the vending machine can be specifically collected, providing a solid data basis for subsequent fault warnings. It is worth mentioning that the set of abnormal data information constructed based on the log record has high reliability and traceability, which is crucial for deeply exploring the cause of the fault.

[0051] Exemplarily, in the data transmission module 30, the set of abnormal data information is uploaded to the cloud platform. It should be understood that due to the limited local data processing capacity and storage space, especially when facing a large number of vending machines deployed dispersedly, centralized data management and analysis become necessary. The cloud platform has powerful computing resources and a large amount of storage capacity, and can support complex data processing tasks, such as the application of deep learning algorithms; in addition, the cloud platform can provide a unified data access interface, enabling the data collected by vending machines from different geographical locations to be aggregated together, forming a larger-scale abnormal data information data set, thereby improving the training quality and generalization ability of the model in fault warning. Therefore, in the technical solution of this application, the set of abnormal data information is uploaded to the cloud platform. By uploading the abnormal data information to the cloud platform, advanced technologies can be used to conduct in-depth joint analysis of the data, capture the implicit patterns and trends within a single vending machine or across multiple vending machines, so as to further improve the accuracy of fault warning.

[0052] Exemplarily, in the abnormal data information semantic encoding module 40, in the cloud platform, semantic encoding is performed on the set of abnormal data information to obtain a set of abnormal data information semantic embedding encoding features. In one embodiment, the abnormal data information semantic encoding module is configured to: input each abnormal data information in the set of abnormal data information into a semantic encoder based on the Word2Vec model to obtain a set of abnormal data information semantic embedding encoding vectors as the set of abnormal data information semantic embedding encoding features. It should be understood that traditional numerical data analysis methods often can only reveal the superficial relationships between abnormal data, and it is difficult to deeply understand the fine-grained semantic associations behind each abnormal data. By inputting each abnormal data information in the set of abnormal data information into a semantic encoder based on the Word2Vec model, the deep semantic structure hidden in the abnormal data information can be mined. For example, when processing the log records of a vending machine, certain specific combinations of error codes or sensor reading patterns may indicate the type of upcoming faults. The Word2Vec model can recognize these combinations even if they have never appeared in the training data, thus greatly enhancing the generalization ability of the system.

[0053] Exemplarily, in the feature dynamic compensation aggregation analysis module 50, feature dynamic compensation aggregation analysis is performed on the set of semantic embedded encoding features of the abnormal data information to obtain significantly aggregated encoding features of the semantic features of the abnormal data information. It should be understood that although the semantic information of each vector in the set of semantic embedded encoding vectors of the abnormal data information has been captured through semantic embedded encoding by the Word2Vec model, these vectors still contain many features that are irrelevant to the understanding of the abnormal data information. In addition, during the semantic analysis process, noise may be introduced, interfering with the understanding of the essential features of the abnormal data information. Therefore, in order to more accurately describe the essential features of the set of semantic embedded encoding vectors of the abnormal data information, enhance the features in the non-essential features that are important for describing the abnormal state of the vending machine, and weaken the features with low importance, in the technical solution of this application, further feature dynamic compensation aggregation analysis is performed on the set of semantic embedded encoding features of the abnormal data information to obtain significantly aggregated encoding feature vectors of the semantic features of the abnormal data information. During this process, feature dynamic compensation aggregation analysis can effectively improve the accuracy of aggregation analysis. During this process, by adaptively adjusting the importance of the fine-grained semantic features of the abnormal data information, potential fault modes of the vending machine are accurately identified to provide reliable early warning results. In particular, the significantly aggregated encoding feature vectors of the semantic features of the abnormal data information obtained through feature dynamic compensation aggregation analysis, as highly concentrated and representative abnormal feature representations, not only contain the core information of the original abnormal data, but also integrate the fine-grained semantic features of the abnormal data in the entire dataset, forming a richer and more comprehensive data representation. In this way, a solid foundation is provided for the intelligent early warning of the vending machine, helping the operation and maintenance personnel to respond and solve problems faster, and ensuring the service quality and user experience of the vending machine.

[0054] In one embodiment, as Figure 3 shown, the feature dynamic compensation aggregation analysis module 50 includes: a hub feature extraction unit 51 for extracting the hub features of the set of semantic embedded encoding features of the abnormal data information; a feature significant aggregation unit 52 for performing feature significant aggregation on the set of semantic embedded encoding features of the abnormal data information based on the hub features to obtain significantly aggregated encoding features of the semantic features of the abnormal data information.

[0055] In one embodiment, the hub feature extraction unit 51 is configured to: input the set of semantic embedding coding vectors of the abnormal data information into a sequence hub extraction network to obtain a hub feature vector of the abnormal data information as the hub feature. Specifically, the hub feature extraction unit is configured to: calculate the essential extraction factors of the abnormal data information corresponding to each semantic embedding coding vector in the set of semantic embedding coding vectors of the abnormal data information to obtain a set of essential extraction factors of the abnormal data information. Specifically, this process can be expressed by the formula:

[0056] X = {x 1 , x 2 ,..., x i ,..., x n}

[0057]

[0058] where X represents the set of semantic embedding coding vectors of the abnormal data information, and x 1 , x 2 , x i , x n respectively represent the 1st, 2nd, ith, and nth semantic embedding coding vectors in the set of semantic embedding coding vectors of the abnormal data information. W 1i and b 1i respectively represent the linear modulation weight matrix and the linear modulation bias vector corresponding to x i . is matrix multiplication, represents the transposed vector of the hub feature correlation scoring conversion vector, and e i is the essential extraction factor of the abnormal data information corresponding to x i ;

[0059] Normalize the set of essential extraction factors of the abnormal data information to obtain a set of normalized essential extraction factors of the abnormal data information. Specifically, this process can be expressed by the formula:

[0060] a i = softmax(e i )

[0061] where softmax(·) is the normalized exponential function, and a i is the normalized essential extraction factor of the abnormal data information corresponding to x i ;

[0062] Using the set of the normalized abnormal data information essence extraction factors as the set of weights, perform position-wise weighted aggregation on the set of the abnormal data information semantic embedding encoded vectors to obtain the abnormal data information hub feature vector. Specifically, this process can be expressed by the formula:

[0063]

[0064] where n represents the number of vectors in the set of the abnormal data information semantic embedding encoded vectors, and v h represents the abnormal data information hub feature vector.

[0065] In this way, input the set of the abnormal data information semantic embedding encoded vectors into the sequence hub extraction network to identify and extract representative hub features from a large number of complex abnormal data information semantic features. These hub features can capture the essential attributes in the abnormal data information while ignoring the noise in the abnormal data information. In this way, the model can more clearly understand the most likely abnormal situations under the abnormal operating state of the vending machine.

[0066] In one embodiment, as Figure 4 shown, the feature significant aggregation unit 52 includes: a node-hub complementary information calculation sub-unit 521, configured to calculate a set of abnormal data information node-hub complementary information embedding encoded vectors based on the abnormal data information hub feature vector; and a feature compensation aggregation sub-unit 522, configured to perform attention-driven feature compensation aggregation on the set of the abnormal data information node-hub complementary information embedding encoded vectors to obtain the abnormal data information semantic feature significant aggregation encoded feature.

[0067] Exemplarily, in the node-hub complementary information calculation sub-unit 521, calculate a set of abnormal data information node-hub complementary information embedding encoded vectors based on the abnormal data information hub feature vector. Specifically, this process can be expressed by the formula:

[0068]

[0069] where conv 1×1 (·) is point convolution processing, W 11 represents the first weight matrix, W 21 represents the second weight matrix, Sigmoid(·) is the Sigmoid function, x i ′ is the abnormal data information semantic encoding embedding encoded modulation vector corresponding to x i , v h ′ is the abnormal data information hub feature modulation vector, is position-wise subtraction, |·| is taking the absolute value, pi is x i and v h The complementary information modulation weight vector between xb i is the i-th abnormal data information-hub complementary information embedding coding vector in the set of abnormal data information-hub complementary information embedding coding vectors.

[0070] In one embodiment, the node-hub complementary information calculation sub-unit is configured to: extract a first abnormal data information semantic embedding coding vector from the set of abnormal data information semantic embedding coding vectors; perform point convolution coding on the first abnormal data information semantic embedding coding vector and multiply the result by a first weight matrix, and input the multiplied result into a Sigmoid function for activation to obtain a first abnormal data information semantic coding embedding coding modulation vector; perform point convolution coding on the abnormal data information hub feature vector and multiply the result by a second weight matrix, and input the multiplied result into a Sigmoid function for activation to obtain an abnormal data information hub feature modulation vector; calculate the position-wise difference between the first abnormal data information semantic coding embedding coding modulation vector and the abnormal data information hub feature modulation vector, and take the absolute value of the calculation result to obtain a first abnormal data information node-hub complementary information modulation weight vector; calculate the position-wise dot product of the first abnormal data information node-hub complementary information modulation weight vector with the first abnormal data information semantic coding embedding coding modulation vector and the abnormal data information hub feature vector respectively, add the calculation results position-wise, perform point convolution coding on the added result and input it into a Sigmoid function for activation to obtain a first node-hub complementary information embedding coding vector.

[0071] Specifically, calculate the complementary information of each abnormal data information semantic embedding coding vector in the set of abnormal data information semantic embedding coding vectors with respect to the abnormal data information hub feature vector to obtain a set of abnormal data information node-hub complementary information embedding coding vectors. Here, the complementary information refers to the detailed information of the abnormal state that may be lost during the hub feature extraction process, that is, the unique contribution of each abnormal data information semantic embedding coding vector with respect to the hub. By calculating this difference, the information unique to each abnormal data information semantic embedding feature but not fully expressed in the hub feature can be captured to obtain the set of abnormal data information node-hub complementary information embedding coding vectors. The role of this step is to retain the important details in the original data and avoid losing key information due to over-simplification.

[0072] In one embodiment, as Figure 5As shown, the feature compensation aggregation subunit 522 includes: an attention modulation secondary subunit 5221, which is used to respectively perform complementary information significant identification on each abnormal data information node-hub complementary information embedded coding vector in the set of abnormal data information node-hub complementary information embedded coding vectors to obtain a set of abnormal data information node complementary information attention weights, and then use the set of abnormal data information node complementary information attention weights to perform attention modulation on the set of abnormal data information node-hub complementary information embedded coding vectors to obtain a set of abnormal data information significant node-hub complementary information embedded coding vectors. Specifically, this process can be expressed by the formula:

[0073]

[0074]

[0075] Y = {xb 1 ·w 1 ,xb 2 ·w 1 ,...,xb i ·w i ,...,xb n ·w n}

[0076] where g(·, ·) is a complementary information calculation function, W 2i and b 2i are respectively the linear modulation weight matrix and the linear modulation bias vector corresponding to xb i , is the transposed vector of the complementary information significance scoring conversion vector, g i is the significance identification factor corresponding to xb i , exp(·) is the function to calculate the natural exponential function value with the natural constant e as the base, w i represents the i-th abnormal data information node complementary information attention weight in the set of abnormal data information node complementary information attention weights, and Y represents the set of abnormal data information significant node-hub complementary information embedded coding vectors.

[0077] A fusion secondary subunit 5222, which is used to fuse the abnormal data information hub feature vector and the set of abnormal data information significant node-hub complementary information embedded coding vectors to obtain an abnormal data information semantic feature significant aggregation coding feature vector as the abnormal data information semantic feature significant aggregation coding feature. Specifically, this process can be expressed by the formula:

[0078] V = Concat{v h ; Y}

[0079] Among them, Concat{·; ·} represents the vector fusion operation, and V represents the feature vector of the significantly aggregated encoding of the semantic features of the abnormal data information.

[0080] It should be understood that the feature importance of the complementary feature components of each abnormal data information semantic feature relative to the hub feature of the abnormal data information is different. That is, some complementary feature components may be supplementary descriptions of the essential features of the abnormal situation of the vending machine, while some complementary feature components may be non-essential feature interferences caused by changes in external conditions. In order to evaluate the importance of each abnormal data information node-hub complementary information in the set of abnormal data information node-hub complementary information for the detection of abnormal items in the vending machine, an attention mechanism is introduced to identify each abnormal data information node-hub complementary information. Specifically, each abnormal data information node-hub complementary information embedding encoding vector in the set of abnormal data information node-hub complementary information embedding encoding vectors is respectively subjected to significant complementary information identification to obtain a set of attention weights of the abnormal data information node complementary information.

[0081] Further, use the set of complementary information attention weights of the abnormal data information nodes to perform attention modulation on the set of abnormal data information node-hub complementary information embedding encoding vectors to obtain a set of abnormal data information significant node-hub complementary information embedding encoding vectors. This process further enhances or suppresses certain abnormal data semantic node features, so that the finally output abnormal data information significant node-hub complementary information embedding feature representation focuses more on the most relevant information, better reflecting the actual importance of abnormal data semantic node features in the sequence, thereby improving the accuracy of the fault warning of the vending machine. In particular, the attention mechanism allows the model to automatically adjust the degree of attention to different parts of the abnormal data information. By evaluating the importance of each abnormal data information node-hub complementary information, it ensures that those parts crucial for the fault item identification and warning of the vending machine receive more "attention", thereby improving the comprehensiveness of the abnormal data information semantic features and further enhancing the accuracy of the fault warning. Finally, fuse the abnormal data information hub feature vector and the set of abnormal data information significant node-hub complementary information embedding encoding vectors to construct a semantic significant aggregation representation of the abnormal data information that can reflect both the global trend and local details of the abnormal data information. In a specific embodiment, feature concatenation can be used to achieve feature aggregation, that is, using the abnormal data information hub feature vector as the basic feature, and then sequentially adding the significant node-hub complementary information embedding encoding vectors modulated by the attention weights. This not only retains the important information in the original data but also enhances the model's understanding of local details. In this way, the model can accurately warn of the faults of the vending machine based on the rich and accurate feature representations it has learned, and thus provide reliable guidance for maintenance personnel.

[0082] Exemplarily, in the fault warning result generation module 60, based on the semantic feature significantly aggregated coding features of the abnormal data information, a fault warning result is generated. In one embodiment, the fault warning result generation module is configured to: input the semantic feature significantly aggregated coding feature vector of the abnormal data information into a fault judgment model based on a classifier to obtain the fault warning result. It should be understood that after the above data processing, the obtained semantic feature significantly aggregated coding feature vector of the abnormal data information contains the core information extracted from the original abnormal data that can highly represent the fault mode. These feature vectors not only condense the core information of the original abnormal data, but also fuse the fine-grained semantic features of the abnormal data in the entire dataset, forming a richer and more comprehensive data representation. By using a classifier model to analyze these feature vectors, it can be identified which feature combinations are most likely to indicate the upcoming fault type, and a specific fault warning result is generated accordingly. Specifically, a classifier is a supervised learning algorithm that can learn how to distinguish different classes or states based on known samples (historical data with labels). In this case, it is the normal operating state and different types of fault states. When new, unseen data (i.e., the semantic feature significantly aggregated coding feature vector of the abnormal data information from the current monitored vending machine) is input into the trained classifier, the model can make a prediction based on the previously learned knowledge, that is, judge which class the new data belongs to - that is, whether there is a risk of a specific type of fault.

[0083] In one embodiment, the fault judgment model based on a classifier uses a Support Vector Machine (SVM). SVM is a powerful binary classification model that maximizes the margin between two classes by finding an optimal hyperplane, thereby achieving effective classification of unknown data points. In the case of multi-class problems, the application of SVM can be extended through the one-vs-rest or one-vs-one method. When faced with a complex non-linear decision boundary, SVM can also utilize the kernel trick, such as the Radial Basis Function (RBF) kernel, to map the original feature space to a higher-dimensional space where it is easier to find a linear separation surface. This method is very suitable for tasks such as vending machine fault warning. In addition, SVM performs well on small sample datasets in high-dimensional spaces, which makes it an ideal choice for processing the semantic feature significantly aggregated coding feature vector of the abnormal data information after the above data processing. By training such an SVM classifier, it can effectively predict possible faults based on new abnormal data information, providing timely and reliable warning information for maintenance personnel to ensure the service quality and user experience of the vending machine.

[0084] In a preferred example, inputting the semantic feature significantly aggregated and encoded feature vector of the abnormal data information into a fault judgment model based on a classifier to obtain a fault warning result includes:

[0085] Determine the mean μ of the eigenvalues and the eigenvalue standard deviation corresponding to the semantic feature significantly aggregated and encoded feature vector of the abnormal data information σ ;

[0086] Subtract the dot product vector of the semantic feature significantly aggregated and encoded feature vector of the abnormal data information and the mean of the eigenvalues from the product of the eigenvalue standard deviation to obtain the first semantic feature significantly aggregated and encoded fairness target vector of the abnormal data information:

[0087]

[0088] where ⊙ represents dot product, represents subtraction, V represents the semantic feature significantly aggregated and encoded feature vector of the abnormal data information, V 1 represents the first semantic feature significantly aggregated and encoded fairness target vector of the abnormal data information;

[0089] Subtract the dot product vector of the semantic feature significantly aggregated and encoded feature vector of the abnormal data information and the eigenvalue standard deviation from the product of the eigenvalue mean to obtain the second semantic feature significantly aggregated and encoded fairness target vector of the abnormal data information:

[0090]

[0091] where V 2 represents the second semantic feature significantly aggregated and encoded fairness target vector of the abnormal data information;

[0092] After multiplying the reciprocal of each bit of the second semantic feature significantly aggregated and encoded fairness target vector by the first semantic feature significantly aggregated and encoded fairness target vector, take the logarithm to the base 2 of each bit to obtain the semantic feature significantly aggregated and encoded information correction vector of the abnormal data information:

[0093]

[0094] where, represents the reciprocal of each bit of the second semantic feature significantly aggregated and encoded fairness target vector, that is, calculate the reciprocal of each eigenvalue of the second semantic feature significantly aggregated and encoded fairness target vector, V in represents the semantic feature significantly aggregated and encoded information correction vector of the abnormal data information;

[0095] Take the square root of the quotient obtained by dividing the mean μ of the eigenvalue by the standard deviation σ of the eigenvalue, multiply it by the weight hyperparameter, and then add it to the corrected vector point of the significant aggregation coding information of the abnormal data information semantic feature to obtain an optimized significant aggregation coding feature vector of the abnormal data information semantic feature:

[0096]

[0097] Wherein, denotes point addition, α denotes the weight hyperparameter, and V′ denotes the optimized significant aggregation coding feature vector of the abnormal data information semantic feature;

[0098] Input the optimized significant aggregation coding feature vector of the abnormal data information semantic feature into the fault judgment model based on the classifier to obtain a fault warning result.

[0099] Here, since each abnormal data information semantic embedding coding vector in the set of abnormal data information semantic embedding coding vectors represents the semantic embedding coding feature of the abnormal data information, when performing feature dynamic compensation aggregation based on the sequence hub, the semantic feature population attributes corresponding to the modal heterogeneity of the abnormal data information will have a compensation aggregation fairness difference based on the sequence hub metric offset, thereby affecting the aggregation feature distribution inclusiveness of the significant aggregation coding feature vector of the abnormal data information semantic feature and reducing the accuracy of the fault warning result obtained by inputting into the fault judgment model based on the classifier.

[0100] Therefore, considering the fairness difference in the attribute level of the data population corresponding to the sequence fusion feature of the significant aggregation coding feature vector of the abnormal data information semantic feature, in order to improve the aggregation inclusiveness under the feature distribution diversity of the significant aggregation coding feature vector of the abnormal data information semantic feature, by using the cross probability value constraint based on the significant aggregation coding feature vector of the abnormal data information semantic feature as the interactive fairness target representation, to correct the group feature information interaction propagation of the significant aggregation coding feature vector of the abnormal data information semantic feature, and using the unified statistical feature response interaction based on the significant aggregation coding feature vector of the abnormal data information semantic feature as the feature distribution multi-level fairness target bias, to achieve a robust distribution fairness unified representation of the significant aggregation coding feature vector of the abnormal data information semantic feature, form a fair cooperation paradigm under the feature distribution framework of the significant aggregation coding feature vector of the abnormal data information semantic feature, and improve the accuracy of the fault warning result obtained by inputting it into the fault judgment model based on the classifier.

[0101] In summary, the intelligent vending machine fault warning system based on the cloud platform according to the embodiments of the present application is elucidated. It determines the key fault repair items to be detected for the vending machine, collects abnormal data information using sensors and other data collection means, and uploads this data to the cloud for analysis. Advanced deep learning algorithms and artificial intelligence technologies are adopted on the cloud platform to conduct in-depth significant correlation aggregation analysis on this abnormal data information of the vending machine, so as to realize predictive diagnosis of possible faults and provide warning results to the operation and maintenance personnel. In this way, the automation degree and intelligent level of vending machine fault and warning can be improved, avoiding the problems of unclear understanding of fault modes and poor generalization ability brought by the method of matching one by one with fault repair items in the existing technical solutions, which is beneficial to ensuring the service quality of the vending machine.

[0102] It should be noted that the block diagrams in the drawings illustrate the possible implementation architectures, functions, and operations of the systems according to various embodiments of the present application. In this regard, each block in the block diagram may represent a module, a program segment, or a part of code, and the module, the program segment, or the part of code contains at least one executable instruction for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram, as well as the combination of blocks in the block diagram, may be implemented by a dedicated hardware-based system for performing the specified functions or operations, or may be implemented by a combination of dedicated hardware and computer instructions.

[0103] The present application uses specific terms to describe the embodiments of the present application. Such as "the first / second embodiment", "an embodiment", and / or "some embodiments" mean a certain feature, structure, or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that the "an embodiment" or "one embodiment" or "an alternative embodiment" mentioned at different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of the present application can be appropriately combined.

[0104] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present application belongs. It should also be understood that terms such as those defined in a common dictionary should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense, unless explicitly defined as such herein.

[0105] The foregoing is a description of the present application and should not be construed as limiting thereof. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily appreciate that many modifications can be made to these exemplary embodiments without departing from the novel teachings and advantages of the present application. Accordingly, all such modifications are intended to be included within the scope of the present application. It should be understood that the foregoing is a description of the present application and should not be considered limited to the specific embodiments disclosed, and modifications to the disclosed embodiments as well as other embodiments are intended to be included within the scope of the definition of the present application. The present application is defined by its description and equivalents.

Claims

1. An intelligent vending machine fault warning system based on a cloud platform, characterized in that: include: A fault repair item determination module, used to determine the fault repair items to be detected in the vending machine; an abnormal data information acquisition module, configured to detect and obtain abnormal data information of the automatic vending machine based on the fault repair item to obtain a set of abnormal data information; A data transmission module, used for uploading the collection of abnormal data information to the cloud platform; An abnormal data information semantic encoding module is used to semantically encode the set of abnormal data information on the cloud platform to obtain a set of abnormal data information semantic embedding encoding features; A feature dynamic compensation aggregation analysis module is used to perform feature dynamic compensation aggregation analysis on the set of abnormal data information semantic embedding coding features to obtain abnormal data information semantic feature significant aggregation coding features, including: a hub feature extraction unit, used to extract the hub features of the set of abnormal data information semantic embedding coding features; a feature significant aggregation unit, used to perform feature significant aggregation on the set of abnormal data information semantic embedding coding features based on the hub features to obtain abnormal data information semantic feature significant aggregation coding features; The fault warning result generation module is used to generate a fault warning result based on the semantic features of the abnormal data information and the significant aggregation coding features.

2. The cloud platform-based intelligent vending machine fault warning system according to claim 1, characterized in that: The fault repair item determination module is used to: Click the troubleshooting page to pop up the troubleshooting list; The troubleshooting item is specified from the troubleshooting list.

3. The cloud platform-based intelligent vending machine fault warning system according to claim 2, characterized in that: The abnormal data information semantic encoding module is used to: Each abnormal data information in the set of abnormal data information is input into a semantic encoder based on the Word2Vec model to obtain a set of abnormal data information semantic embedding coding vectors as the set of abnormal data information semantic embedding coding features.

4. The cloud platform-based intelligent vending machine fault warning system according to claim 3 is characterized in that: The hub feature extraction unit is used to: The set of the abnormal data information semantic embedding encoding vectors is input into a sequence hub extraction network to obtain an abnormal data information hub feature vector as the hub feature.

5. The cloud platform-based intelligent vending machine fault warning system according to claim 4 is characterized in that: The hub feature extraction unit is used to: Calculating the abnormal data information essential extraction factor corresponding to each abnormal data information semantic embedding coding vector in the set of abnormal data information semantic embedding coding vectors to obtain a set of abnormal data information essential extraction factors; Normalizing the set of abnormal data information essence extraction factors to obtain a set of normalized abnormal data information essence extraction factors; The set of normalized abnormal data information essential extraction factors is used as a set of weights, and the set of abnormal data information semantic embedding coding vectors is weightedly aggregated by position to obtain the abnormal data information hub feature vector.

6. The cloud platform-based intelligent vending machine fault warning system according to claim 5, characterized in that: The characteristic significant polymerization unit includes: A node-hub complementary information calculation subunit, used to calculate a set of abnormal data information node-hub complementary information embedding coding vectors based on the abnormal data information hub feature vector; The feature compensation aggregation subunit is used to perform attention-driven feature compensation aggregation on the set of abnormal data information node-hub complementary information embedded coding vectors to obtain the significant aggregation coding features of the abnormal data information semantic features.

7. The cloud platform-based intelligent vending machine fault warning system according to claim 6, characterized in that: The node-hub complementary information calculation subunit is used to: Extracting a first abnormal data information semantic embedding coding vector from the set of abnormal data information semantic embedding coding vectors; Performing point convolution encoding on the first abnormal data information semantic embedding coding vector and then multiplying it with the first weight matrix, and inputting the multiplication result into the Sigmoid function for activation to obtain the first abnormal data information semantic coding embedding coding modulation vector; The abnormal data information hub feature vector is point-convolution-encoded and then multiplied by a second weight matrix, and the multiplication result is input into a Sigmoid function for activation to obtain an abnormal data information hub feature modulation vector; Calculate the positional difference between the first abnormal data information semantic coding embedding coding modulation vector and the abnormal data information hub feature modulation vector, and take the absolute value of the calculation result to obtain the first abnormal data information node-hub complementary information modulation weight vector; The positional point multiplication of the first abnormal data information node-hub complementary information modulation weight vector, the first abnormal data information semantic coding embedding coding modulation vector and the abnormal data information hub feature vector is calculated respectively, and the calculated results are added by position, and the added result is point convolutionally encoded and input into the Sigmoid function for activation to obtain the first node-hub complementary information embedding coding vector.

8. The cloud platform-based intelligent vending machine fault warning system according to claim 7, characterized in that: The characteristic compensation aggregation subunit comprises: The attention modulation secondary subunit is used to respectively perform complementary information significant marking on each abnormal data information node-hub complementary information embedded coding vector in the set of abnormal data information node-hub complementary information embedded coding vectors to obtain a set of abnormal data information node complementary information attention weights, and then perform attention modulation on the set of abnormal data information node-hub complementary information embedded coding vectors with the set of abnormal data information node complementary information attention weights to obtain a set of abnormal data information significant node-hub complementary information embedded coding vectors; The secondary subunit is fused to fuse the abnormal data information hub feature vector and the set of abnormal data information significant node-hub complementary information embedding coding vectors to obtain the abnormal data information semantic feature significant aggregation coding feature vector as the abnormal data information semantic feature significant aggregation coding feature.

9. The cloud platform-based intelligent vending machine fault warning system according to claim 8, characterized in that: The fault warning result generating module is used for: The semantic feature significant aggregation encoding feature vector of the abnormal data information is input into a fault judgment model based on a classifier to obtain the fault warning result.

Images