Printing equipment fault detection method, equipment, medium and product

Abstract

The invention discloses a printing equipment fault detection method, equipment, a medium and a product, which are applied to the field of artificial intelligence, and comprise the following steps: collecting hardware parameters and software parameters of target printing equipment, and generating target equipment characteristics based on the hardware parameters and the software parameters; inputting the target equipment features into a pre-trained fault detection model to obtain a fault probability output by the fault detection model; and determining a target fault level according to the fault probability, and calling a target standard process in the maintenance strategy library based on the target fault level. By collecting parameters and generating target equipment features, equipment operation state information can be comprehensively captured, and multi-dimensional data support is provided for fault detection. The fault probability is determined through the fault detection model, quantitative evaluation of the fault probability can be realized by means of the advantage of double-model fusion, and the prediction accuracy is improved. By determining the target fault level and calling the corresponding standard process, hierarchical management and accurate operation and maintenance are realized, and the equipment maintenance efficiency and reliability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, and in particular to a printing device fault detection method, device, medium and product. BACKGROUND

[0002] As a key device in modern office and production processes, the stable operation of a printing device is crucial to work efficiency and business continuity. With the continuous expansion of the application scenarios of printing devices in various industries, the problems such as downtime loss and increased maintenance costs caused by device faults are increasingly prominent. Therefore, it is of great significance to efficiently and accurately detect and maintain the printing device.

[0003] Currently, the fault detection of a printing device mainly relies on manual inspection or threshold alarm mechanism. The manual inspection method requires technicians to periodically inspect the device by observing the device's running state and listening to the device's operating sound to determine whether there is a fault. The threshold alarm mechanism is to issue an alarm when the temperature exceeds a certain fixed value. The manual inspection method is not only inefficient, but also relies on the experience of technicians, making it difficult to detect potential early faults and easily causing sudden device failures to affect normal work. The simple threshold alarm mechanism lacks deep analysis and processing capabilities for data, and cannot adapt to complex situations under different devices and working conditions. SUMMARY

[0004] The present application provides a printing device fault detection method, device, medium and product, which predicts the fault probability and divides the risk level through a double model collaborative framework, realizes predictive maintenance of the printing device, solves the problems of device downtime, low operation and maintenance efficiency and resource waste caused by traditional passive maintenance, and improves the reliability and operation efficiency of bank self-service devices.

[0005] According to an aspect of the present application, a printing device fault detection method is provided, which comprises:

[0006] Collecting hardware parameters and software parameters of a target printing device, and generating target device features based on the hardware parameters and software parameters;

[0007] Inputting the target device features into a pre-trained fault detection model to obtain a fault probability output by the fault detection model, wherein the fault detection model is a double model collaborative framework using gradient boosting tree and long short-term memory network;

[0008] Determining a target fault level according to the fault probability, and calling a target standard process in a maintenance strategy library based on the target fault level.

[0009] Optionally, the training process of the fault detection model comprises: obtaining historical parameters, and obtaining fault labels based on the historical parameters to construct an original data set; extracting time sequence features and static features based on the original data set; using gradient boosting trees to process the static features and using long short-term memory networks to process the time sequence features, respectively, to independently train two basic models; calculating performance indicator values of the two basic models on a validation set, and calculating dynamic weights based on the performance indicator values to determine the contribution proportions of the two basic models; and fusing the two basic models according to the contribution proportions through a weighted fusion strategy to generate the fault detection model.

[0010] Optionally, the hardware parameters and software parameters of the target printing device are collected, comprising: deploying hardware sensors and consumable detection units at target positions of the target printing device to collect hardware parameters, wherein the hardware parameters include temperature data, vibration data, position data, consumable reserves, print head wear degree, power supply parameters and mechanical component states; determining target fields in a printing task log and system running data, and extracting software parameters from the target fields, wherein the software parameters include task settings, task statistics, error codes, abnormal responses, communication delays, firmware versions, historical maintenance records and user operation logs.

[0011] Optionally, the target device features are generated based on the hardware parameters and the software parameters, comprising: filtering transient abnormal values in the temperature data using a sliding window average method, and removing random interference values in the vibration data using Kalman filtering; normalizing each hardware parameter and software parameter to generate normalized parameters, and constructing fusion features according to each normalized parameter; selecting and reducing the dimensions of the fusion features to eliminate redundant features to generate the target device features.

[0012] Optionally, the target fault level is determined according to the fault probability, comprising: when the fault probability is greater than or equal to a first preset threshold, determining that the target fault level is high risk; when the fault probability is less than the first preset threshold and greater than or equal to a second preset threshold, determining that the target fault level is medium risk; and when the fault probability is less than the second preset threshold, determining that the target fault level is low risk.

[0013] Optionally, the target standard process in the maintenance strategy library is called based on the target fault level, comprising: when the target fault level is high risk, cutting off the power supply of the device and stopping the printing task, saving all fault data at the fault moment, generating prompt information according to the all fault data and the target fault level, and sending the prompt information to the target user; when the target fault level is medium risk, limiting the function of the target printing device based on the target fault level, and automatically performing a preset basic maintenance work; and when the target fault level is low risk, marking a risk point in the printing task log based on the target fault level.

[0014] Optionally, the method further comprises: acquiring a data query instruction input by the user, wherein the data query instruction comprises a display mode and a filtering range; filtering the collected software parameters and hardware parameters based on the filtering range to obtain target parameters; and processing the target parameters based on the display mode to generate a visual view and displaying the visual view to the user.

[0015] According to another aspect of the present application, there is provided an electronic device comprising:

[0016] at least one processor;

[0017] and a memory connected to the at least one processor in communication;

[0018] wherein the memory stores a computer program capable of being executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the method for detecting a fault of a printing device according to any one of the embodiments of the present application.

[0019] According to another aspect of the present application, there is provided a computer readable storage medium storing computer instructions for enabling a processor to perform the method for detecting a fault of a printing device according to any one of the embodiments of the present application when executed by the processor.

[0020] According to another aspect of the present application, there is provided a computer program product comprising a computer program for enabling a processor to perform the method for detecting a fault of a printing device according to any one of the embodiments of the present application when executed by the processor.

[0021] The technical solution of the embodiments of the present application can comprehensively capture device running state information by collecting parameters and generating target device features, provide multi-dimensional data support for fault detection, and ensure the representativeness and effectiveness of the features. By inputting the target device features into a pre-trained fault detection model to obtain a fault probability, the advantages of double model fusion can be utilized, and the static feature and time sequence feature analysis capabilities can be combined to realize quantitative evaluation of the fault probability and improve the prediction accuracy. By determining the target fault level and calling the corresponding standard process for processing, hierarchical management and accurate operation and maintenance are realized, and the device maintenance efficiency and reliability are improved.

[0022] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to make the technical solution in the embodiments of the present application clearer, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0024] Figure 1 is a flow chart of a printing device fault detection method according to an embodiment of the present application;

[0025] Figure 2 is a flow chart of a fault detection model training method according to an embodiment of the present application;

[0026] Figure 3 is a structural schematic diagram of a printing device fault detection apparatus according to an embodiment of the present application;

[0027] Figure 4 is a structural schematic diagram of an electronic device for implementing a printing device fault detection method according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the technical solution in the embodiments of the present application clearer, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0029] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to include only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0030] Embodiment one

[0031] Figure 1A flowchart of a printing device fault detection method is provided for Embodiment One of the present application. The embodiment can be applied to a printing device maintenance scenario. The method can be performed by a printing device fault detection device, which can be realized in the form of hardware and / or software, and can be configured in a computer controller. As shown in FIG. 10, the method comprises the following steps. Figure 1

[0032] In S110, hardware parameters and software parameters of the target printing device are collected, and target device features are generated based on the hardware parameters and the software parameters.

[0033] The hardware parameters refer to data collected by deploying hardware sensors and consumable detection units at target positions of the target printing device, including temperature data, vibration data, position data, consumable reserves, print head wear degree, power supply parameters, and mechanical component states, and are used to reflect the operating conditions of the hardware of the printing device. The software parameters refer to data extracted from target fields of printing task logs and system running data, including task settings, task statistics, error codes, abnormal responses, communication delays, firmware versions, historical maintenance records, and user operation logs, and can be used to understand the operating conditions of the software of the printing device. The target device refers to feature data obtained after a series of processing of the hardware parameters and the software parameters. The processing process includes filtering transient abnormal values in the temperature data by using a sliding window average method, removing random interference values in the vibration data by using Kalman filtering, normalizing the parameters to generate normalized parameters, constructing fusion features according to the normalized parameters, and finally selecting and reducing the fusion features to obtain the target device features, which are used to represent the overall operating features of the printing device.

[0034] Optionally, the hardware parameters and the software parameters of the target printing device are collected, including deploying hardware sensors and consumable detection units at target positions of the target printing device to collect the hardware parameters, wherein the hardware parameters include temperature data, vibration data, position data, consumable reserves, print head wear degree, power supply parameters, and mechanical component states; and determining target fields in printing task logs and system running data, and extracting software parameters from the target fields, wherein the software parameters include task settings, task statistics, error codes, abnormal responses, communication delays, firmware versions, historical maintenance records, and user operation logs.

[0035] ​Specifically, when collecting hardware parameters, different types of hardware sensors and consumable detection units can be deployed at key positions of the printing device. For example, temperature sensors are deployed on components prone to heat such as print heads, rubber rollers, and power modules to monitor temperature changes in real time; vibration sensors are installed on mechanical moving parts such as print head carriages and paper feeding mechanisms to capture vibrations during device operation; and position sensors are placed on the X / Y axis movement paths of the print head to record the actual position of the print head. In addition, consumable detection units are installed to detect the remaining toner amount of ink cartridges or toner cartridges, paper sensor status, and determine whether there is a paper shortage, paper jam, or other issues.

[0036] Among them, the temperature data is collected by the temperature sensor at a certain frequency, such as once per second, to collect the temperature of the print head, rubber roller, power module, etc. The measurement accuracy can reach ±0.5℃, which is used to determine whether the device has an overheating risk. The vibration data is collected by the vibration sensor at a high frequency, such as 1000Hz, to collect the vibration frequency, amplitude, and direction of the print head movement and paper feeding mechanism operation. By analyzing the vibration data, abnormal conditions such as mechanical component loosening and wear can be found. The position data refers to the actual position of the print head on the X / Y axis recorded by the position sensor in real time, and is compared with the instructed position. The accuracy can reach ±0.01mm. If the position deviation exceeds a certain threshold, such as 0.05mm, it may indicate problems such as guide rail deviation. The consumable remaining amount refers to the information detected by the consumable detection unit through capacitive sensors, infrared transmission sensors, weight sensors, etc., such as the remaining toner amount of ink cartridges / toner cartridges, the remaining number of paper, and the paper jam position, which provides a basis for consumable replacement and paper jam fault warning. The print head wear degree refers to the use of optical or electrical methods to detect print head nozzle blockage rate, ink output deviation, and other indicators to quantify the wear degree of the print head, such as representing the wear proportion with a value of 0-100%. The power parameter refers to monitoring the input voltage fluctuation range of the device, such as whether it is within the range of 220V±10%, as well as current consumption, power factor, and other power-related parameters to determine whether the power module is working normally. The mechanical component state can be indirectly reflected through vibration data and position data, such as judging whether the gear engagement is normal through vibration spectrum analysis, and judging whether the guide rail is worn through position deviation.

[0037] Specifically, when collecting software parameters, the target fields related to the device running state and faults can be determined by deeply analyzing the structure of the print task log and system running data. The print task log records various information during the device's execution of a print task, and the system running data contains running state parameters at the device software level. By filtering key fields, valuable software parameters can be extracted.

[0038] Among them, the task setting refers to extracting the setting parameters of the printing task from the printing task log, such as printing resolution, color mode, printing speed, etc., and the task setting will affect the workload and running state of the device. Task statistics refers to the statistics of related data of printing tasks, including the number of continuous printing pages, the total printing volume per day, the number of task interruptions, etc., which are used to analyze the use intensity and stability of the device. Error code refers to the error code generated when the device fails during operation, such as E03 indicating insufficient toner, E101 indicating paper jam, and extracting error code and its frequency from system running data can quickly locate the fault type. Abnormal response refers to recording the response information of the device to abnormal conditions, such as the number of automatic recovery attempts when paper jam, and the record of print head initialization failure, etc., reflecting the self-repairing ability and fault handling mechanism of the device. Communication delay refers to measuring the instruction response time of USB or network interface, which should be less than a certain threshold under normal circumstances, such as 5 milliseconds, and if the delay is too long, it may affect the normal work of the device. The firmware version refers to obtaining the firmware version number of the current device running, and different firmware versions may have known vulnerabilities or compatibility issues, which is convenient for subsequent version upgrade and troubleshooting. Historical maintenance records refer to synchronizing the historical maintenance records of the device from the central management system, including the last cleaning time, component replacement records, etc., which can understand the maintenance situation and component service life of the device. User operation log refers to recording the user's operation through the touch screen, device control panel or related software, such as user cancel printing, replacing consumables, adjusting printing settings, etc. Operation information can assist in analyzing the correlation between faults and user operations.

[0039] Optionally, the target device features are generated based on hardware parameters and software parameters, including: using a sliding window average method to filter transient abnormal values in temperature data, and using Kalman filtering to remove random interference values in vibration data; normalizing each hardware parameter and software parameter to generate normalized parameters, and constructing fusion features according to each normalized parameter; selecting and reducing the dimension of the fusion features, and eliminating redundant features to generate the target device features.

[0040] Specifically, the sliding window average method processing temperature data refers to replacing the value of the center data point in the window with the average value of the temperature in the window, for the real-time data collected by the temperature sensor, such as the temperature of the print head being 85.3°C, and setting a time window, such as the last 5 sampling points. For example, when the temperature suddenly jumps from 70°C to 100°C and then quickly falls, the sliding window average method can smooth such abnormal fluctuations, retain the true temperature change trend, and avoid misjudgment caused by sensor instantaneous failure or environmental interference. Kalman filtering processing vibration data refers to three-axis acceleration data collected by the vibration sensor, such as X-axis 0.32g and Y-axis 0.15g, which often contains random noise. Through Kalman filtering, a state space model can be established, and the optimal estimation value is recursively calculated based on the previous predicted value and the current measured value. For example, when the print head moves normally, the vibration amplitude should be within the range of 0.1-0.5g, and if the measured value at a certain time is 2.3g, it indicates that it may be caused by a short collision, and Kalman filtering will judge it as noise and eliminate it, retaining the vibration characteristics reflecting the true state of the mechanical parts.

[0041] Specifically, normalization processing refers to linear normalization formula for temperature data, vibration data, and position deviation parameters of different dimensions, which are scaled to the [0, 1] interval to ensure that different features have equal weights in model training and avoid model bias caused by dimensional differences.

[0042] Further, the controller will construct fusion features based on the normalized parameters. For example, from the temperature data, extract the temperature change rate of a specified period, the duration of exceeding the threshold, and other time series dynamic features; from the vibration data, calculate the vibration amplitude standard deviation, the main frequency feature, and other features. At the same time, calculate the static statistical features such as consumable consumption rate and print head wear degree daily growth rate, and extract the high-resolution task proportion, error code frequency, and other software parameters. Finally, select and reduce the fusion features. Analyze the correlation between features through Pearson correlation coefficient, eliminate redundant features with high correlation, calculate feature importance score, and select core features with high contribution to fault prediction. For high-dimensional time series features, principal component analysis is used for dimensionality reduction processing to reduce model calculation while retaining main information, and finally generate target device features, which are arranged as structured vectors and converted into corresponding input formats according to the needs of different models.

[0043] S120, input the target device features into the pre-trained fault detection model to obtain the fault probability output by the fault detection model, wherein the fault detection model is a dual-model collaborative framework using gradient boosting tree and long short-term memory network.

[0044] The fault detection model is a dual-model collaborative framework of gradient boosting tree and long short-term memory network, and is a trained model. The fault detection model can output a corresponding fault probability according to input target equipment features, to predict the possibility of a printing device failure, by learning the rules in historical data. The fault probability is calculated by the fault detection model according to input target equipment features, and is used to represent the possibility of a printing device failure. Gradient boosting tree is good at processing high-dimensional discrete features, and is used to learn static features in hardware parameters. Long short-term memory network can capture long-term dependencies of time series data, and is used to process time series features in hardware parameters and task time series information in software parameters.

[0045] Specifically, the generated target equipment features are input into the pre-trained fault detection model. The gradient boosting tree and the long short-term memory network in the model process the features respectively, output respective fault probability prediction results, and then perform weighted fusion according to dynamic weights to obtain a final fault probability, to evaluate the possibility of a printing device failure.

[0046] In S130, a target fault level is determined according to the fault probability, and a target standard process in the maintenance strategy library is called based on the target fault level.

[0047] The target fault level refers to a level divided according to the fault probability, and can intuitively reflect the severity of a printing device failure. The maintenance strategy library stores a database of various maintenance strategies and standard processes for different fault levels, so that corresponding maintenance measures can be called from the database after the target fault level is determined. The target standard process refers to a standardized maintenance operation process formulated in the maintenance strategy library for a specific target fault level, and clearly defines specific measures to be taken under the fault level.

[0048] Optionally, determining the target fault level according to the fault probability includes: when the fault probability is greater than or equal to a first preset threshold, determining that the target fault level is high risk; when the fault probability is less than the first preset threshold and greater than or equal to a second preset threshold, determining that the target fault level is medium risk; and when the fault probability is less than the second preset threshold, determining that the target fault level is low risk.

[0049] Specifically, the controller pre-sets two key probability thresholds, i.e., a first preset threshold and a second preset threshold. For example, the first preset threshold is set to 0.8, and the second preset threshold is set to 0.5. When the fault probability output by the fault detection model is greater than or equal to the first preset threshold, it indicates that the device has a very high possibility of failure, and the target fault level is determined as high risk, and emergency measures need to be taken immediately to deal with the possible failure. When the fault probability is less than the first preset threshold and greater than or equal to the second preset threshold, it indicates that the device has a certain risk of failure, but has not reached an emergency state, and the target fault level is determined as medium risk, and timely intervention and preventive maintenance are needed to prevent the failure from further developing. When the fault probability is less than the second preset threshold, it means that the device has a low possibility of failure and is in a relatively normal operating state, and the target fault level is determined as low risk, and the device needs to be continuously monitored to record potential risk points. By distinguishing the fault levels, different risk levels of faults can be processed more targetedly, and the efficiency and accuracy of device maintenance can be improved.

[0050] Optionally, a target standard process in the maintenance strategy library is called based on the target fault level, including: when the target fault level is high risk, cutting off the power supply of the device and stopping the printing task, saving all fault data at the fault moment, generating prompt information according to the all fault data and the target fault level, and sending the prompt information to the target user; when the target fault level is medium risk, limiting the function of the target printing device based on the target fault level, and automatically performing a preset basic maintenance work; when the target fault level is low risk, marking a risk point in a printing task log based on the target fault level.

[0051] Specifically, when the target failure level is high risk, the controller will immediately trigger the emergency response mechanism, first cut off the power supply of the device and stop all printing tasks to prevent the failure from expanding to cause more serious damage; save all hardware parameters and software parameters at the time of failure to form a complete fault snapshot; then generate prompt information containing fault code and predicted fault type based on fault data, which can be sent to maintenance personnel through SMS, system pop-up window, etc. In addition, if the printing device is configured with a backup module, it automatically switches to the backup module to maintain basic functions, and sends a spare parts demand order to the supply chain system, such as replacing the print head of model A380, and pushes a graphic guide containing specific maintenance steps to the maintenance terminal. When the target failure level is medium risk, the controller will execute the preventive maintenance strategy, first limit the functions of the device, such as reducing the printing speed and disabling the color mode to reduce the device load, then automatically execute the preset basic maintenance operations such as print head cleaning program, roller calibration, etc. At the same time, a prompt is popped up on the device control interface, and a work order with a priority of 24 hours is generated, which locates the possible failure components through feature weight analysis and attaches operation video tutorials; subsequently, the failure probability is continuously monitored every hour, and if the probability increases, it is automatically upgraded to high risk processing. When the target failure level is low risk, the controller will take mild intervention measures, only mark the potential risk point in the printing task log, without affecting the normal operation of the device, but will increase the sampling frequency of related sensors, such as increasing the temperature sampling from 1Hz to 5Hz, in order to more accurately capture subtle changes. At the same time, the controller will also set the current state as a new normal reference point for the model, and if the same type of potential risk frequently occurs, the corresponding preventive measures will be supplemented to the maintenance strategy library for optimizing the subsequent failure prediction model.

[0052] Optionally, the method further comprises: obtaining a data query instruction input by a user, wherein the data query instruction includes a display mode and a filtering range; filtering the collected software parameters and hardware parameters based on the filtering range to obtain target parameters; processing the target parameters based on the display mode to generate a visual view, and displaying the visual view to the user.

[0053] The display mode refers to the form of data presentation specified by the user, such as a ring-shaped progress bar that can display the proportion of consumable remaining amount, a line chart that can display temperature and vibration trends, and a column chart that can compare fault rates at different time periods. The filtering range refers to the data filtering conditions set by the user, such as time range, parameter type, and fault level. The filtering target parameter refers to filtering data from the historical collected hardware parameters and software parameters according to the user-specified filtering range. For example, if the user selects "records of print head temperature greater than 80 DEG C in the past 7 days", the controller will filter out all temperature sensor data within the corresponding time period and extract records exceeding the threshold. If the user selects "statistics of tasks with high-risk fault level", the controller will associate fault probability data and filter out the corresponding printing task parameters.

[0054] Specifically, the controller processes the filtered target parameters according to the user-specified display mode to generate a visual view. For example, the remaining amount of consumables, such as 30% of toner remaining, is converted into a ring chart to visually display the remaining proportion. Time series data such as temperature and vibration are plotted along the time axis to support the display of change trends and abnormal points. Further, the processed visual view is displayed to the user through the device control interface, web interface, or mobile application, and the user can view details, click on data points in the chart to display specific values and timestamps. Data export is also possible, and the controller supports exporting chart data in table format for further analysis.

[0055] The technical solution of the embodiment of the present application can capture device running state information comprehensively by collecting parameters and generating target device features, providing multi-dimensional data support for fault detection, and ensuring the representativeness and effectiveness of the features. By inputting the target device features into the pre-trained fault detection model to obtain the fault probability, the advantages of double model fusion can be utilized to combine static feature and time series feature analysis capabilities to realize quantitative evaluation of fault probability and improve prediction accuracy. By determining the target fault level and calling the corresponding standard process, hierarchical management and precise operation and maintenance are realized, improving the efficiency and reliability of device maintenance.

[0056] Embodiment Two

[0057] Figure 2 A flowchart of a fault detection model training method provided by Embodiment Two of the present application is provided. Based on Embodiment One described above, the training process of the fault detection model is specifically described. As shown in Figure 2 the method comprises:

[0058] S210, acquiring historical parameters, and obtaining fault labels based on the historical parameters to construct an original data set.

[0059] Specifically, the historical parameters refer to collecting a large amount of historical operation data from the hardware sensors and software logs of the printing device, covering normal and fault scenarios. Then, the user can combine the device maintenance records with expert experience to label the fault types or potential risks for the historical parameters. Then the controller can divide all the data into training set, validation set and test set in the ratio of 7:2:1, to ensure the uniform distribution of fault types in each set, and provide a reliable data basis for subsequent training.

[0060] It should be noted that when the fault data sample is insufficient, a generative adversarial network can be used to generate synthetic fault data, or a reversible transformation of time series data can be used to expand the diversity of the training set and improve the model generalization ability.

[0061] S220, extract time sequence features and static features based on the original data set.

[0062] Specifically, the controller will perform feature extraction, time sequence feature extraction including statistical calculation, periodic feature mining and lag feature construction. Statistical calculation refers to using sliding window technology for time series data such as temperature and vibration that changes over time. Periodic feature mining refers to analyzing the periodic variation of data over time. Lag feature construction refers to using data at the previous time or previous times as part of the current time feature. Static feature extraction includes discretization processing, ratio feature calculation and historical maintenance time interval extraction. Discretization processing refers to encoding software parameters such as firmware version and consumable type, for example, encoding firmware version into different categories, assuming there are three firmware versions V1.0, V2.0 and V3.0, encoded as 0, 1 and 2 respectively, so that gradient boosting tree can be processed. Ratio feature calculation refers to calculating consumable consumption rate, task completion rate and other ratio features. For example, by calculating the amount of ink or toner consumed per hour to get the consumable consumption rate, and by calculating the ratio of completed printing tasks to the total number of tasks to get the task completion rate, to reflect the running state of the device from different angles. Historical maintenance time interval extraction refers to extracting the interval days from the last maintenance to the current time from the historical maintenance records, to judge the possibility of device failure under long-term maintenance.

[0063] S230, respectively use gradient boosting tree to process static features and long short-term memory network to process time sequence features, and independently train two basic models.

[0064] Specifically, when training the gradient boosting tree base model, the normalized static features can be first sorted into a feature matrix form. For example, each row of the feature matrix represents a data sample, and each column corresponds to a static feature, such as firmware version, consumable type, daily average print volume, etc. Then, the classification error of the fault type is minimized as the optimization objective of the training. During the training process, the gradient boosting tree base model adjusts the internal decision tree structure and parameters to accurately classify the input static features into the corresponding fault type as much as possible. Finally, the grid search method can be used to optimize the hyperparameters of the gradient boosting tree base model. For example, the tree depth, learning rate, and other hyperparameters are tested in combination, and the model performance is evaluated on the validation set to find the hyperparameter combination that minimizes the model classification error.

[0065] Specifically, when training the long short-term memory network base model, the time series feature data can be first sorted into a three-dimensional tensor form, i.e., [sample number, time step length, feature dimension]. Assuming there are 1000 samples, each sample contains 60 time steps, such as collecting data every minute for 60 minutes, and each time step has 5 features, then the shape of the input data is [1000, 60, 5]. Then the network structure is built, i.e., a network structure containing 2 layers of long short-term memory network layers and a fully connected layer is built. Each layer of long short-term memory network layer is set to 128 neurons, and the long short-term memory network layer can effectively handle the long-term dependence relationship in the time series data through the gating mechanism to selectively remember and forget information. The fully connected layer maps the feature vector output by the long short-term memory network layer to the final fault prediction dimension. Finally, an optimizer can be used to adjust the parameters of the model, and the optimizer can adaptively adjust the learning rate. During training, the batch size can be set to 64, i.e., 64 samples are taken from the training set each time for parameter update. At the same time, the early stopping mechanism is adopted to prevent model overfitting, and when the performance of the model on the validation set, such as accuracy, loss function value, etc., no longer improves within a certain number of rounds, the training is stopped.

[0066] S240, calculate the performance indicator value of the two base models on the validation set, and calculate the dynamic weight based on the performance indicator value to determine the contribution proportion of the two base models.

[0067] Specifically, the performance index value can be the area under curve (AUC). The AUC value is an important index for measuring the performance of a binary classification model, and the value ranges from 0 to 1. The closer the value is to 1, the better the performance of the model on the validation set. The AUC values corresponding to the gradient boosting tree and the long short-term memory network base models are calculated respectively. For example, the AUC of the gradient boosting tree base model on the validation set is 0.85, and the AUC of the long short-term memory network base model is 0.92. According to the calculated AUC values, the dynamic weights of the two models can be further calculated. For example, the dynamic weight = 0.85 / (0.85+0.92)≈0.48, at this time, the contribution ratio of the two base models can be determined, that is, the gradient boosting tree base model contributes 48% in the final prediction result, and the long short-term memory network base model contributes 52%.

[0068] S250, fusing the two base models according to the contribution ratio through a weighted fusion strategy to generate a fault detection model.

[0069] Specifically, when performing the final fault prediction, the prediction values of the two models are weighted and fused according to the calculated weights. For example, for a certain data sample, the gradient boosting tree base model predicts a fault probability of 0.6, and the long short-term memory network base model predicts a fault probability of 0.7, then the final prediction probability = 0.48x0.6+(1-0.48)x0.7≈0.652.

[0070] The technical scheme of the embodiment of the present application can extract time sequence features and static features based on the original data set, can deeply mine data value from two dimensions of time sequence change rule and device static attribute, and can provide adaptive effective features for subsequent model training. By independently training two base models using gradient boosting tree and long short-term memory network respectively, the unique advantages of the two in processing static data classification and time sequence data dependency can be utilized to learn device fault features from different angles. By calculating the performance index value of the base model on the validation set and determining the dynamic weight, the contribution ratio of the model in the fusion can be flexibly allocated according to the actual performance of the model, and the advantages can be complementary. By fusing the two base models through a weighted fusion strategy, the respective advantages can be integrated to generate a more comprehensive and accurate detection model for predicting the fault of a printing device.

[0071] Embodiment three

[0072] Figure 3 A structure schematic diagram of a printing device fault detection device provided by the embodiment three of the present application is shown in FIG. 3. Figure 3 As shown in the figure, the device comprises a parameter acquisition and feature generation module 310, which is configured to acquire hardware parameters and software parameters of a target printing device, and generate target device features based on the hardware parameters and software parameters.

[0073] The fault probability determination module 320 is configured to input the target equipment feature into a pre-trained fault detection model to obtain a fault probability output by the fault detection model, wherein the fault detection model is a dual-model collaborative framework adopting a gradient boosting tree and a long short-term memory network.

[0074] The fault level determination module 330 is configured to determine a target fault level according to the fault probability, and call a target standard process in a maintenance strategy library based on the target fault level.

[0075] Optionally, the apparatus further comprises a fault detection model training module configured to: obtain historical parameters, and obtain fault labels based on the historical parameters to construct an original data set; extract time sequence features and static features based on the original data set; independently train two basic models by using a gradient boosting tree to process the static features and a long short-term memory network to process the time sequence features; calculate performance index values of the two basic models on a verification set, and calculate dynamic weights based on the performance index values to determine contribution ratios of the two basic models; and fuse the two basic models according to the contribution ratios by a weighted fusion strategy to generate the fault detection model.

[0076] Optionally, the parameter acquisition and feature generation module 310 specifically comprises: a parameter acquisition unit configured to: deploy hardware sensors and consumable detection units at target positions of the target printing equipment to acquire hardware parameters, wherein the hardware parameters include temperature data, vibration data, position data, consumable reserves, print head wear degree, power supply parameters and mechanical component states; and determine target fields in printing task logs and system running data, and extract software parameters from the target fields, wherein the software parameters include task settings, task statistics, error codes, abnormal responses, communication delays, firmware versions, historical maintenance records and user operation logs.

[0077] Optionally, the parameter acquisition and feature generation module 310 specifically comprises: a target feature generation unit configured to: filter transient abnormal values in the temperature data by using a sliding window average method, and remove random interference values in the vibration data by using Kalman filtering; normalize each hardware parameter and software parameter to generate normalized parameters, and construct fusion features according to the normalized parameters; select and reduce dimensions of the fusion features, and eliminate redundant features to generate the target equipment feature.

[0078] Optionally, the fault level determination module 330 specifically comprises: a fault level determination unit configured to: determine that the target fault level is high risk when the fault probability is greater than or equal to a first preset threshold; determine that the target fault level is medium risk when the fault probability is less than the first preset threshold and greater than or equal to a second preset threshold; and determine that the target fault level is low risk when the fault probability is less than the second preset threshold.

[0079] Optionally, the fault level determination module 330 specifically comprises a standard process calling unit, configured to: when the target fault level is high risk, cut off the power supply of the device and stop the printing task, save all fault data at the fault moment, generate prompt information according to the all fault data and the target fault level, and send the prompt information to the target user; when the target fault level is medium risk, limit the function of the target printing device based on the target fault level, and automatically execute the preset basic maintenance work; when the target fault level is low risk, mark the risk point in the printing task log based on the target fault level.

[0080] Optionally, the apparatus further comprises a visualization module configured to: acquire a data query instruction input by a user, wherein the data query instruction comprises a display mode and a filtering range; filter the collected software parameters and hardware parameters based on the filtering range to obtain target parameters; process the target parameters based on the display mode to generate a visualization view, and display the visualization view to the user.

[0081] The technical scheme of the embodiment of the present application can comprehensively capture device running state information by collecting parameters and generating target device features, provide multi-dimensional data support for fault detection, and ensure the representativeness and effectiveness of the features. By inputting the target device features into a pre-trained fault detection model to obtain a fault probability, the advantages of double model fusion can be used, and the static feature and time sequence feature analysis capabilities can be combined to realize quantitative evaluation of the fault probability and improve the prediction accuracy. By determining the target fault level and calling the corresponding standard process, hierarchical management and accurate operation and maintenance are realized, and the device maintenance efficiency and reliability are improved.

[0082] The printing device fault detection apparatus provided in the embodiment of the present application can execute the printing device fault detection method provided in any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0083] Embodiment four

[0084] Figure 4 A structural schematic diagram of an electronic device 10 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the applications described and / or claimed in this document.

[0085] As Figure 4As shown, the electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., connected to the at least one processor 11 in communication. The memory stores a computer program executable by the at least one processor 11, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0086] Various components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc., an output unit 17, such as various types of displays, a speaker, etc., a storage unit 18, such as a magnetic disk, an optical disk, etc., and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0087] The processor 11 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as a printing device fault detection method.

[0088] In some embodiments, a printing device fault detection method can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of a printing device fault detection method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform a printing device fault detection method by any other appropriate means, such as by means of firmware.

[0089] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0090] Computer programs used to implement the processes of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program, when executed, can cause instructions defined in the flow charts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package and partially on a remote machine or entirely on a remote machine or server.

[0091] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0092] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0093] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0094] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0095] It should be understood that various forms of flow shown above can be used, with steps reordered, added, or removed. For example, the steps recited in the present invention can be performed in parallel, in series, or in a different order, without limitation herein, as long as the desired results of the technical solutions of the present invention can be achieved.

[0096] The specific embodiments described above are not intended to be limiting, but rather to illustrate the principles of the present invention. Variations, combinations, sub-combinations, and modifications can be made to the described embodiments within the spirit and scope of the present invention. Accordingly, other implementations are within the scope of the following claims.

Claims

1. A method for detecting faults in a printing device, characterized in that, include: Collect the hardware and software parameters of the target printing device, and generate target device characteristics based on the hardware and software parameters; The target device features are input into a pre-trained fault detection model to obtain the fault probability output by the fault detection model, wherein the fault detection model is a dual-model collaborative framework using gradient boosting tree and long short-term memory network. The target fault level is determined based on the fault probability, and the target standard process in the maintenance strategy library is invoked based on the target fault level.

2. The method according to claim 1, characterized in that, The training process of the fault detection model includes: Obtain historical parameters and fault labels based on the historical parameters to construct the original dataset; Extract temporal and static features based on the original dataset; Two base models were trained independently: one using gradient boosting trees to process static features and the other using long short-term memory networks to process temporal features. Calculate the performance metrics of the two base models on the validation set, and calculate the dynamic weights based on the performance metrics to determine the contribution ratio of the two base models. The two base models are fused according to their contribution ratios using a weighted fusion strategy to generate a fault detection model.

3. The method according to claim 1, characterized in that, The hardware and software parameters of the target printing device are collected, including: Hardware sensors and consumable detection units are deployed at the target location of the target printing device to collect hardware parameters, including temperature data, vibration data, position data, consumable balance, printhead wear, power parameters, and mechanical component status. Identify target fields in the print task logs and system operation data, and extract software parameters from the target fields. The software parameters include task settings, task statistics, error codes, exception responses, communication delays, firmware versions, historical maintenance records, and user operation logs.

4. The method according to claim 3, characterized in that, The process of generating target device features based on the hardware parameters and the software parameters includes: The sliding window averaging method is used to filter out instantaneous outliers in the temperature data, and Kalman filtering is used to remove random interference values ​​in the vibration data. The hardware and software parameters are normalized to generate normalized parameters, and fusion features are constructed based on the normalized parameters. The fused features are selected and dimensionality reduced to eliminate redundant features in order to generate target device features.

5. The method according to claim 1, characterized in that, The determination of the target fault level based on the fault probability includes: When the failure probability is greater than or equal to a first preset threshold, the target failure level is determined to be high risk; When the failure probability is less than a first preset threshold and greater than or equal to a second preset threshold, the target failure level is determined to be medium risk. When the failure probability is less than the second preset threshold, the target failure level is determined to be low risk.

6. The method according to claim 5, characterized in that, The process of invoking the target standard procedure in the maintenance strategy library based on the target fault level includes: When the target fault level is high risk, the device power is cut off and the printing task is stopped. All fault data at the time of the fault is saved. A prompt message is generated based on the all fault data and the target fault level and sent to the target user. When the target fault level is medium risk, the target printing device is functionally restricted based on the target fault level, and preset basic maintenance work is automatically performed. When the target fault level is low risk, the risk point is marked in the print task log based on the target fault level.

7. The method according to claim 1, characterized in that, The method further includes: Obtain the data query instruction input by the user, wherein the data query instruction includes the display method and the filtering range; The collected software and hardware parameters are filtered based on the filtering range to obtain the target parameters; The target parameters are processed based on the aforementioned display method to generate a visual view, which is then displayed to the user.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

9. A computer storage medium, characterized in that, The computer storage medium stores computer instructions that are used to cause a processor to execute the method of any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images