Bank intelligent printing equipment and consumable management platform thereof

Abstract

The embodiment of the invention relates to the technical field of artificial intelligence, and discloses bank intelligent printing equipment and a consumable management platform thereof, and the platform comprises an intelligent monitoring module which is used for detecting the state of each bank intelligent printing equipment in real time, and carrying out consumable allowance tracking and environment adaptation adjustment; the AI prediction module is used for performing prediction by adopting a multi-modal space-time fusion model according to the consumable use data in a preset time period before the current time period, and outputting and obtaining a consumable demand list and a fault risk in a future preset time period; and the dynamic inventory management module is used for automatically placing an order for a consumable list when the inventory reaches a critical value, and carrying out code scanning verification and intercepting mismatched consumables when the consumables are put in storage. Through the above mode, full-automatic intelligent management and control of the bank intelligent printing equipment are realized, and the fault rate and the inventory cost are effectively reduced.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the field of artificial intelligence technology, in particular to a bank intelligent printing device and a consumable management platform thereof. BACKGROUND

[0002] At present, when a bank clerk handles business, he needs to print vouchers, customers self-print flow, print reports in the background, etc. In these scenarios, the traditional pin printer and laser printer have the following problems: high device failure rate, paper jam, broken needle, etc. often occur, the clerk will pause the business processing failure, and the customer will be impatient. In terms of cost, the color band of the printer is consumed quickly, and the bank has a large amount of business every day, so the consumable expenditure is large. The maintenance cost of the device itself is also high, especially for old printers, spare parts are hard to find. Environmental protection is getting more and more attention. Traditional printing uses so many paper sheets, and also needs to dispose of waste color band boxes and toner cartridges, which is not environmentally friendly and requires additional disposal costs.

[0003] Therefore, the consumable management of the bank printing device directly affects the supply chain efficiency and cost control. SUMMARY

[0004] In view of the above problems, the embodiment of the present application provides a bank intelligent printing device and a consumable management platform thereof, which is used to solve the problems of low intelligence, low efficiency and high cost of the bank printing device in the prior art.

[0005] According to one aspect of the embodiment of the present application, a consumable management platform of a bank intelligent printing device is provided, which comprises:

[0006] An intelligent monitoring module is configured to detect the state of each bank intelligent printing device in real time, track the remaining amount of consumables, and adjust the environment adaptation;

[0007] An AI prediction module is configured to predict the consumable demand list and fault risk in a future preset time period by using a multi-modal spatio-temporal fusion model according to the consumable use data in a preset time period before the current time period, wherein the multi-modal spatio-temporal fusion model comprises a fault prediction sub-model, a spatio-temporal convolution network, a graph neural network, an attention mechanism, a feature embedding layer, a spatio-temporal fusion layer, and a multi-task prediction head; the multi-modal spatio-temporal fusion model is trained by using training data; the training data comprises historical printing amount time series data, historical business calendar data, branch topology relationship data, device state data, and external factor data; before training, the historical use rate, the business calendar, the branch topology relationship data, the device state data, and the external factor data are respectively subjected to differential encryption processing according to the sensitivity level of the data, and the sensitivity level is dynamically adjusted by feature contribution degree analysis at each training round; the external factor data comprises weather influence data;

[0008] A dynamic inventory management module is configured to automatically place an order for consumables when the inventory reaches a critical value, and to intercept mismatched consumables by scanning the codes when the consumables are stored in the warehouse.

[0009] In an optional manner, the platform further comprises a training module configured to:

[0010] The historical printing amount time series data, the historical business calendar data, the branch topology relationship data, the equipment state data and the external factor data are respectively encrypted by differential encryption processing to obtain encrypted historical printing amount time series data, encrypted historical business calendar data, encrypted branch topology relationship data, encrypted equipment state data and encrypted external factor data.

[0011] The encrypted historical printing amount time series data, the encrypted historical business calendar data, the encrypted branch topology relationship data, the encrypted equipment state data and the encrypted external factor data are subjected to feature engineering processing to obtain feature-enhanced historical printing amount time series data, feature-enhanced historical business calendar data, feature-enhanced branch topology relationship data, feature-enhanced equipment state data and feature-enhanced external factor data.

[0012] The feature-enhanced historical printing amount time series data is input into the spatio-temporal convolution network, the feature-enhanced historical business calendar data is input into the attention mechanism, the feature-enhanced branch topology relationship data is input into the graph neural network, and the feature-enhanced external factor data is input into the feature embedding layer.

[0013] The feature-enhanced equipment state data is input into the fault prediction sub-model to obtain a predicted fault risk result.

[0014] A loss function is calculated according to the consumable prediction result and the predicted fault risk result, respectively, parameters are adjusted, and the feature contribution degrees of the historical printing amount time series data, the historical business calendar data, the branch topology relationship data, the equipment state data and the external factor data are calculated, and the sensitivity level is adjusted according to the feature contribution degrees.

[0015] According to the adjusted parameters and the adjusted sensitivity level, the iterative training is continued until a trained multi-modal spatio-temporal fusion model is obtained.

[0016] In an optional manner, the historical printing amount time series data, the historical business calendar data and the external factor data are respectively encrypted by differential encryption processing to obtain encrypted historical printing amount time series data, encrypted historical business calendar data, encrypted branch topology relationship data, encrypted device state data and encrypted external factor data, and further comprising:

[0017] The historical printing amount time series data and the historical business calendar data are homomorphically encrypted by a separate line key to obtain encrypted historical printing amount time series data;

[0018] The branch topology relationship data is graph structure obfuscated encrypted by a central key escrow to obtain encrypted branch topology relationship data;

[0019] The device state data is encrypted by a hardware level encryption using a device chip fuse key mechanism to obtain encrypted device state data;

[0020] The external factor data is calculated by differential privacy to obtain encrypted external factor data.

[0021] In an optional manner, the historical printing amount time series data with feature enhancement is input into the spatio-temporal convolution network, the historical business calendar data with feature enhancement is input into the attention mechanism, the branch topology relationship data with feature enhancement is input into the anonymous graph neural network, and the external factor data with feature enhancement is input into the feature embedding layer. The spatio-temporal fusion layer receives the outputs of the spatio-temporal convolution network, the attention mechanism and the feature embedding layer, and outputs consumable prediction results by the multi-task prediction head after feature fusion. The prediction results include predicted consumable demand, and further comprising:

[0022] When the historical printing amount time series data with feature enhancement is input into the spatio-temporal convolution network for training, a ciphertext state aggregation statistical training method is used for training;

[0023] After the historical business calendar data with feature enhancement is input into the attention mechanism, a ciphertext state aggregation statistical training method is used for training;

[0024] The branch topology relationship data with feature enhancement is input into the anonymous graph neural network for training;

[0025] Before the external factor data with feature enhancement is input into the feature embedding layer, the encrypted external factor data is decrypted locally at the edge node before training.

[0026] In an optional manner, the intelligent monitoring module is further used for:

[0027] Through real-time heartbeat detection, the state of each bank intelligent printing device is collected every preset time, and the abnormal bank intelligent printing device is marked;

[0028] For each bank intelligent printing device, the remaining percentage of ink cartridge or toner cartridge is updated every time a preset number of pages is printed, and a warning is issued when the carbon powder reserve is lower than the preset reserve value;

[0029] According to the temperature and humidity data, the printing speed and paper jam prevention mode are dynamically adjusted.

[0030] In an optional manner, the dynamic inventory management module is further used for:

[0031] When the consumables are placed in the bank intelligent printing device, the bank intelligent printing device scans the NFC encrypted tag of the consumables; wherein the bank intelligent printing device is provided with an NFC encrypted tag reading module;

[0032] The print head of the bank intelligent printing device verifies the NFC encrypted tag and matches the device model;

[0033] When the matching fails, an alarm is issued to avoid mismatch.

[0034] In an optional manner, the platform further includes a hidden fault detection detection module, which is used for:

[0035] Collecting the current running parameters of the bank intelligent printing device; the current running parameters include mechanical vibration parameters, thermal distribution parameters and circuit characteristic parameters;

[0036] Wavelet packet energy entropy analysis is used to determine whether the mechanical vibration parameters exceed the preset vibration threshold;

[0037] Convolution temperature field reconstruction is used to analyze and determine whether the local temperature difference of the bank intelligent printing device is abnormal;

[0038] The Fourier spectrum kurtosis is used to detect the current ripple in the circuit characteristic parameters to determine whether the circuit is abnormal.

[0039] In an optional manner, the platform further includes a cost control module, which is used for:

[0040] Print behavior detection is performed to identify paper waste behavior and generate prompt information;

[0041] The cost information of each intelligent bank printing device is counted according to a preset period, a cost report of each intelligent bank printing device is generated, and information of the intelligent printing device exceeding the target cost is prompted.

[0042] In an optional manner, the platform further includes a blockchain audit token module, which is used for:

[0043] Recording the flow information of consumables in the blockchain, tracing the flow in the life cycle of the ink cartridge or the toner cartridge;

[0044] Recording the difference between the actual printing amount and the nominal value of the consumables in the blockchain, and initiating a claim when the difference is greater than a preset difference threshold;

[0045] Recording the consumable taking information in the blockchain.

[0046] According to another aspect of the embodiments of the present application, a bank intelligent printing device is provided, comprising the consumable management platform.

[0047] The embodiments of the present application realize full-automatic intelligent management and control of the bank intelligent printing device, effectively reduce the failure rate and inventory cost, and considering the privacy of printing data in the intelligent bank printing system, the data is classified and encrypted during data acquisition and training in the model training process, so that the problem of data leakage is effectively avoided.

[0048] The above description is only a summary of the technical solutions of the embodiments of the present application, in order to more clearly understand the technical means of the embodiments of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the embodiments of the present application more obvious and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0049] The accompanying drawings are only used to illustrate the embodiments, and are not considered as limiting the present application. Moreover, the same reference signs are used to represent the same parts throughout the drawings. In the drawings:

[0050] Figure 1 The structure of the consumable management platform of the bank intelligent printing device provided by the embodiments of the present application is shown;

[0051] Figure 2 The architecture of the multi-modal spatio-temporal fusion model in the consumable management platform of the bank intelligent printing device provided by the embodiments of the present application is shown;

[0052] Figure 3 The structure of the bank intelligent printing device provided by the embodiments of the present application is shown;

[0053] Figure 4 This diagram illustrates the interaction between a bank's intelligent printing device and a consumables management platform, according to another embodiment of the present invention. Detailed Implementation

[0054] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0055] Figure 1 This diagram illustrates the structure of a consumables management platform 100 for a bank's intelligent printing device according to an embodiment of the present invention. The consumables management platform 100 can be, for example... Figure 3 As shown, the software system set in a centrally controlled bank smart printing device 10 can also be, for example... Figure 4 The image shows a cloud-based consumables management platform that interacts with multiple bank smart printing devices. For example... Figure 1 As shown, the platform includes the following modules:

[0056] The intelligent monitoring module 110 is used to monitor the status of each bank's intelligent printing equipment in real time, track the remaining consumables, and adjust for environmental adaptation.

[0057] The intelligent monitoring module 110 is specifically used to: collect the status of each bank's smart printing equipment at preset intervals through real-time heartbeat detection, and mark abnormal smart printing equipment. Specifically, it can collect the equipment status of each bank's smart printing equipment every minute: online / offline, print queue, and error code. Abnormal smart printing equipment is automatically marked. For example, if a bank's smart printing equipment remains offline for more than 15 minutes, it indicates that the smart printing equipment is abnormal.

[0058] The intelligent monitoring module 110 is specifically used for: updating the remaining percentage of ink cartridges or toner cartridges for each bank's intelligent printing equipment every preset number of pages printed, and issuing an alert when the toner level is lower than the preset level. In this embodiment of the invention, the remaining percentage of the toner cartridge / ink drum is updated every 50 pages printed, and a yellow alert icon flashes when the toner level is lower than 20%. The printing speed and anti-paper jam mode are dynamically adjusted based on temperature and humidity data. Specifically, by installing chips or sensors on the toner cartridge / ink drum, the usage of consumables is monitored in real time, and the number of pages printed is recorded by a counter built into the printer.

[0059] The AI ​​prediction module 120 is used to make predictions based on the consumable usage data of a preset time period before the current time period using a multimodal spatiotemporal fusion model, and outputs a list of consumable demand and failure risks for the future preset time period.

[0060] Among them, such as Figure 2As shown, the multi-modal spatio-temporal fusion model includes a failure prediction sub-model, a spatio-temporal convolution network, a graph neural network, an attention mechanism, a feature embedding layer, a spatio-temporal fusion layer, and a multi-task prediction head. The multi-modal spatio-temporal fusion model is trained using training data. The training data includes historical printing volume time series data, historical business calendar data, branch topology relationship data, device state data, and external factor data. Before training, the historical usage rate, the business calendar, the branch topology relationship data, the device state data, and the external factor data are differentially encrypted according to the sensitivity level of the data, and the sensitivity level is dynamically adjusted through feature contribution analysis during each training.

[0061] The platform also includes a training module.

[0062] The training module is configured to encrypt the historical printing volume time series data, the historical business calendar data, the branch topology relationship data, the device state data, and the external factor data using differential encryption processing to obtain encrypted historical printing volume time series data, encrypted historical business calendar data, encrypted branch topology relationship data, encrypted device state data, and encrypted external factor data.

[0063] Specifically, historical printing volume time series data is accumulated by collecting printer historical logs, such as collecting bank intelligent printing device operation data for at least 60 consecutive days, including daily total printing volume and color printing volume, consumable quantity change curve, and device power-on and power-off times. The collected time series data is standardized and linearly interpolated for missing values to construct a 30-day period feature sequence.

[0064] The historical business calendar data of each bank intelligent printing device is collected from the bank core system, including the end-of-month settlement day marker, the new branch opening countdown, and the marketing activity printing task reservation amount.

[0065] In the embodiments of the present application, the branch topology relationship data of all bank intelligent printing devices is constructed according to the location, device information, and quantity of the bank intelligent printing devices of each bank branch. The device state data includes historical sensor data of the device (including fuser temperature, gear set vibration spectrum, power module current harmonic, etc.), usage information of the selenium drum (such as the number of printed pages), etc. The external factor data includes environmental humidity and temperature information, etc.

[0066] The training module is further configured to: encrypt the historical printing amount time series data and the historical business calendar data by using homomorphic encryption, and manage the encryption keys independently by using branch independent keys, to obtain encrypted historical printing amount time series data and encrypted historical business calendar data. The data sensitivity level of the branch topology relationship data is high, and the branch topology relationship data is encrypted by using graph structure confusion (K-anonymization), and the encryption keys are managed by using a central key management, to obtain encrypted branch topology relationship data. The device state data is encrypted by using hardware level (SGX / TEE), and the data sensitivity level is extremely high, so the device chip fuse key mechanism is used for encryption, and the encryption keys are managed by using device chip fuse keys, to obtain encrypted device state data. The external factor data has a low data sensitivity level, and differential privacy calculation is used to obtain encrypted external factor data.

[0067] The training module is further configured to: perform feature engineering processing on the encrypted historical printing amount time series data, the encrypted historical business calendar data, the encrypted branch topology relationship data, the encrypted device state data, and the encrypted external factor data, to obtain feature-enhanced historical printing amount time series data, feature-enhanced historical business calendar data, feature-enhanced branch topology relationship data, feature-enhanced device state data, and feature-enhanced external factor data. The feature-enhanced branch topology relationship data is enhanced by using graph structure coding and adjacency matrix normalization, to obtain feature-enhanced branch topology relationship data. The encrypted historical business calendar data is processed by using one-hot encoding and time decay weighting. The encrypted device state data is processed by using wavelet transform and time-frequency domain feature extraction, to obtain feature-enhanced device state data.

[0068] The training module is further configured to: input the feature-enhanced historical printing amount time series data into the spatio-temporal convolution network, input the feature-enhanced historical business calendar data into the attention mechanism, input the feature-enhanced branch topology relationship data into the graph neural network, and input the feature-enhanced external factor data into the feature embedding layer. The spatio-temporal fusion layer receives outputs of the spatio-temporal convolution network, the attention mechanism, and the feature embedding layer, performs feature fusion, and outputs a consumable prediction result by using the multi-task prediction head. The prediction result includes a predicted consumable demand. In the embodiment of the present application, the encrypted data is trained by using a federated learning architecture.

[0069] The training module is further configured to: input the feature-enhanced device state data into a fault prediction sub-model, to obtain a predicted fault risk result.

[0070] The training module is further configured to: calculate a loss function according to the consumable prediction result and the predicted failure risk result respectively, adjust parameters, and calculate feature contribution degrees of historical printing volume time series data, historical business calendar data, branch topology relationship data, device state data and external factor data, and adjust a sensitive level according to the feature contribution degrees. In the embodiment of the application, the key feature contribution degrees are determined through SHAP value analysis. The main loss function can adopt a quantile loss function, and the auxiliary loss function can adopt a physical constraint loss to ensure that the consumable conservation law is met, and a fluctuation smoothing penalty to suppress the violent fluctuation of the predicted value, and the embodiment of the application does not make specific limitations.

[0071] The training module is further configured to: continue to perform iterative training according to the adjusted parameters and the adjusted sensitive level until a trained multi-modal spatio-temporal fusion model is obtained.

[0072] After the training module trains the multi-modal spatio-temporal fusion model, the AI prediction module 120 inputs the consumable use data of a preset time period before the current time period into the multi-modal spatio-temporal fusion model to perform prediction and obtain a consumable demand list and a failure risk in a future preset time period. For example, the consumable demand list of each bank intelligent printing device in the next 7 days, such as the number and model of required toner cartridges or ink cartridges, and the failure rate.

[0073] The dynamic inventory management module 130 is configured to automatically place an order for the consumable list when the inventory reaches a critical value, and to perform code scanning verification when the consumables are stored in the warehouse to intercept mismatched consumables.

[0074] When the consumable is placed in the bank intelligent printing device, the bank intelligent printing device scans the NFC encrypted tag of the consumable. The NFC encrypted tag reading module is arranged on the bank intelligent printing device. The print head of the bank intelligent printing device verifies the NFC encrypted tag and matches the device model. When the matching fails, an alarm is sent to avoid mismatch.

[0075] In the embodiment of the application, the platform further comprises an implicit failure detection module configured to: collect current operating parameters of the bank intelligent printing device; the current operating parameters include mechanical vibration parameters, thermal distribution parameters and circuit characteristic parameters. The wavelet packet energy entropy analysis is adopted to determine whether the mechanical vibration parameters exceed a preset vibration threshold. The convolution temperature field reconstruction is adopted to analyze and determine whether the local temperature difference of the bank intelligent printing device is abnormal. The Fourier spectrum kurtosis is adopted to detect the current ripple in the circuit characteristic parameters to determine whether the circuit is abnormal.

[0076] The platform further comprises a cost control module, which is configured to detect printing behaviors, identify paper waste behaviors, and generate prompt information; and to count cost information of each intelligent bank printing device according to a preset period, generate a cost report of each intelligent printing device, and prompt information of an intelligent printing device that exceeds a target cost.

[0077] In the embodiment of the present application, the platform further comprises a blockchain audit token module, which is configured to record the transfer information of the consumables in the blockchain, trace the transfer in the life cycle of the ink cartridge or the toner cartridge, record the difference between the actual printing amount and the nominal value of the consumables in the blockchain, and initiate a claim when the difference is greater than a preset difference threshold, and record the consumable taking information in the blockchain.

[0078] According to another aspect of the embodiment of the present application, a bank intelligent printing device is provided, comprising the consumable management platform described above.

[0079] In the embodiment of the present application, the intelligent monitoring module is configured to detect the state of each bank intelligent printing device in real time, track the remaining amount of consumables, and adjust the environment; the AI prediction module uses a multi-modal spatio-temporal fusion model to predict the consumable demand list and fault risk in a future preset time period according to the consumable usage data in a preset time period before the current time period; and the dynamic inventory management module automatically places an order for the consumable list when the inventory reaches a critical value, and verifies the code when the consumables are put into the warehouse, and intercepts mismatched consumables. The bank intelligent printing device is fully automatically and intelligently managed, the failure rate and inventory cost are effectively reduced, and the data is classified and encrypted during the data acquisition and training process of the model training, so that the problem of data leakage is effectively avoided.

[0080] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general purpose systems can be used with programs in accordance with the teachings herein, or it can prove convenient to construct more specialized apparatus to perform the required method steps. The required structure for a variety of these systems will be apparent from the description above. In addition, the present application is not intended to be limited to any particular programming language. It will be appreciated that a variety of programming languages can be used to implement the teachings of the application as described herein, and any references below to specific languages are provided for disclosure of enablement only.

[0081] In the specification provided herein, a large number of specific details are described. However, it can be understood that the embodiments of the present application can be practiced without these specific details. In some examples, well-known methods, structures and techniques are not shown in detail in order not to obscure the understanding of the present specification.

[0082] Similarly, it is to be understood that the embodiments of the application can be adapted to any of the various aspects of the application described herein, and that the description of the exemplary embodiments of the application given above, does not limit the scope of the application, but that the scope of the application is to be given that broadest interpretation which will encompass all compatible embodiments including those now known or which will become known in the future.

[0083] Those skilled in the art will appreciate that modules in the apparatuses in the embodiments can be adapted and placed in one or more apparatuses different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and can be further split into more sub-modules or sub-units or sub-components. Any combination of all the features disclosed in the specification (including the accompanying claims, abstract and drawings), and any method or process or steps of any method or process so disclosed, can be made unless it is explicitly stated otherwise. Each feature disclosed in the description (including the accompanying claims, abstract and drawings) can be replaced by alternative features that serve the same, equivalent or similar purpose, unless expressly stated otherwise.

[0084] It is noted that the foregoing examples have been provided merely for the purposes of illustration and are not intended to limit the application in any way. As will be recognized by those skilled in the art that changes can be made to the embodiments described consisting of purely replacing like or equivalent elements with other like or equivalent elements while still remaining within the scope of the claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The application can be implemented by means of both hardware and software, and any combination thereof. In a unit claim, several devices or means can be listed, even though they are not in each other's depending category. The use of the terms first, second and third, and the like, does not imply any ordering but rather are used for naming purposes only. Features in the description and / or the claims that are expressed in the negative can be adapted to expressly include the positive features. Steps in the description and / or the claims that are expressed in a particular order can be adapted to include an order reverse to that particular order.

Claims

1. A consumables management platform for intelligent printing equipment in banks, characterized in that, The platform includes: The intelligent monitoring module is used to detect the status of each bank's intelligent printing equipment in real time, track the remaining consumables, and adjust for environmental adaptation. The AI ​​prediction module is used to predict consumable usage data from a preset time period prior to the current time period using a multimodal spatiotemporal fusion model, outputting a list of consumable demand and fault risks for the future preset time period. The multimodal spatiotemporal fusion model includes a fault prediction sub-model, a spatiotemporal convolutional network, a graph neural network, an attention mechanism, a feature embedding layer, a spatiotemporal fusion layer, and a multi-task prediction head. The multimodal spatiotemporal fusion model is trained using training data, including historical print volume time-series data, historical business calendar data, branch network topology data, equipment status data, and external factor data. Before training, the historical usage rate, business calendar data, branch network topology data, equipment status data, and external factor data are differentially encrypted according to their sensitivity levels. The sensitivity levels are dynamically adjusted in each training round through feature contribution analysis. The external factor data includes weather impact data. The dynamic inventory management module is used to automatically place an order for consumables when the inventory reaches a critical value, and to perform barcode verification when consumables are received into the warehouse to intercept mismatched consumables.

2. The platform according to claim 1, characterized in that, The platform also includes a training module for: Differential encryption is applied to the historical print volume time series data, historical business calendar data, branch topology data, equipment status data, and external factor data to obtain encrypted historical print volume time series data, encrypted historical business calendar data, encrypted branch topology data, encrypted equipment status data, and encrypted external factor data. Feature engineering is performed on the encrypted historical print volume time-series data, encrypted historical business calendar data, encrypted branch topology data, encrypted equipment status data, and encrypted external factor data to obtain feature-enhanced historical print volume time-series data, feature-enhanced historical business calendar data, feature-enhanced branch topology data, feature-enhanced equipment status data, and feature-enhanced external factor data. The feature-enhanced historical print volume time-series data is input into the spatiotemporal convolutional network, the feature-enhanced historical business calendar data is input into the attention mechanism, the feature-enhanced network topology relationship data is input into the graph neural network, and the feature-enhanced external factor data is input into the feature embedding layer. The spatiotemporal fusion layer receives the outputs of the spatiotemporal convolutional network, the attention mechanism, and the feature embedding layer, and performs feature fusion. Finally, the multi-task prediction head outputs the consumable prediction result, which includes the predicted consumable demand. The enhanced equipment status data is input into the fault prediction sub-model to obtain the predicted fault risk results; The loss function is calculated based on the consumables prediction results and the predicted failure risk results, the parameters are adjusted, and the feature contribution of historical print volume time series data, historical business calendar data, network topology relationship data, equipment status data and external factor data is calculated. The sensitivity level is adjusted based on the feature contribution. Based on the adjusted parameters and the adjusted sensitivity level, iterative training continues until a well-trained multimodal spatiotemporal fusion model is obtained.

3. The platform according to claim 2, characterized in that, The historical print volume time-series data, historical business calendar data, and external factor data are encrypted using differentiated encryption processing to obtain encrypted historical print volume time-series data, encrypted historical business calendar data, encrypted branch topology data, encrypted equipment status data, and encrypted external factor data, further including: The historical print volume time-series data and historical business calendar data are homomorphically encrypted using separate keys for each line, resulting in encrypted historical print volume time-series data and encrypted historical business calendar data, respectively. The network topology data is encrypted using graph-structure obfuscation and central key custody to obtain encrypted network topology data. The device status data is encrypted at the hardware level using a device chip fuse key mechanism to obtain the encrypted device status data. Differential privacy computation is applied to the external factor data to obtain encrypted external factor data.

4. The platform according to claim 3, characterized in that, The process involves inputting the enhanced historical print volume time-series data into the spatiotemporal convolutional network, the enhanced historical business calendar data into the attention mechanism, the enhanced network topology data into the anonymous graph neural network, and the enhanced external factor data into the feature embedding layer. The spatiotemporal fusion layer receives the outputs from the spatiotemporal convolutional network, the attention mechanism, and the feature embedding layer, performs feature fusion, and then outputs the consumables prediction result from the multi-task prediction head. The prediction result includes predicted consumables demand, and further includes: When the feature-enhanced historical printing volume time-series data is input into the spatiotemporal convolutional network for training, a training method using encrypted state aggregation statistics is adopted. After inputting the feature-enhanced historical business calendar data into the attention mechanism, it is trained using a training method that aggregates and statistically analyzes encrypted states. Feature-enhanced network topology data is input into an anonymous graph neural network for training; Before inputting the feature-enhanced external factor data into the feature embedding layer, the encrypted external factor data is locally decrypted at the edge nodes before training.

5. The platform according to claim 1, characterized in that, The intelligent monitoring module is further used for: By detecting heartbeats in real time, the status of each bank's smart printing device is collected at preset intervals, and abnormal bank smart printing devices are marked. For the smart printing devices of various banks, the remaining percentage of ink cartridges or toner cartridges is updated every preset number of pages printed, and an alarm is issued when the toner level is lower than the preset level. The printing speed and anti-paper jam mode are dynamically adjusted based on temperature and humidity data.

6. The platform according to claim 1, characterized in that, The dynamic inventory management module is further used for: When consumables are placed into the bank's smart printing device, the bank's smart printing device scans the NFC encryption tag of the consumables; wherein, the bank's smart printing device is equipped with an NFC encryption tag reading module; The printhead of the bank's smart printing equipment verifies the NFC encrypted tag and matches the device model. An alarm is issued when there is a mismatch to prevent mismatches.

7. The platform according to any one of claims 1-6, characterized in that, The platform also includes a latent fault detection module, used for: Collect the current operating parameters of the bank's smart printing equipment; the current operating parameters include mechanical vibration parameters, thermal distribution parameters, and circuit characteristic parameters. Wavelet packet energy entropy analysis is used to determine whether the mechanical vibration parameters exceed a preset vibration threshold. Convolutional temperature field reconstruction was used to analyze and determine whether the local temperature difference of the bank's smart printing equipment was abnormal. Fourier spectrum kurtosis is used to detect current ripple in the circuit characteristic parameters to determine whether the circuit is abnormal.

8. The platform according to any one of claims 1-7, characterized in that, The platform also includes a cost control module for: Perform printing behavior detection, identify paper waste, and generate prompt messages; The system calculates the cost information of each smart bank printing device according to a preset cycle, generates cost reports for each smart printing device, and alerts users of smart printing devices that exceed the target cost.

9. The platform according to any one of claims 1-7, characterized in that, The platform also includes a blockchain audit token module, used for: Record the circulation information of consumables in the blockchain to trace the circulation of ink cartridges or toner cartridges throughout their life cycle; The difference between the actual printing quantity and the nominal value of consumables is recorded in the blockchain. When the difference exceeds a preset difference threshold, a claim is initiated. Record consumable requisition information in the blockchain.

10. A smart printing device for banks, characterized in that, The bank's smart printing equipment includes a consumables management platform as described in any one of claims 1-9.

Images