Printer heating roller production quality monitoring method and system based on AI

Abstract

The invention discloses an AI-based printer heating roller production quality monitoring method and system, and relates to the technical field of product quality monitoring, and the method comprises the steps: collecting material parameters of a heating roller, predicting production parameters through employing an XGBoost regression device, and calculating the temperature rise duration of the heating roller through combining with temperature response; time windows are divided according to the heating duration, multi-modal data of the heating roller are collected to calculate multi-modal features, multi-modal fusion is carried out in combination with a multi-branch neural network to form fusion characterization, and the fusion characterization is analyzed through linear regression to output a quality classification result of the heating roller; and classifying the produced heating rollers according to the quality classification result of the heating rollers, counting the production qualification rate of the heating rollers, displaying, and forming a quality monitoring record for storage. Accurate evaluation is provided for the production quality of the heating roller, and real-time monitoring and feedback in the production process are achieved.

Description

Technical Field

[0001] This invention relates to the field of product quality monitoring technology, and in particular to an AI-based method and system for monitoring the production quality of printer heating rollers. Background Technology

[0002] With the widespread application of printers and printing technology, especially in the industrial manufacturing field, the heating roller, as a key component of printers, directly affects the printing accuracy, efficiency, and performance of the final product. In recent years, with the advancement of intelligent manufacturing technology, using machine learning and artificial intelligence to monitor and optimize the production process has become an important means to improve manufacturing quality. Especially in the production process of heating rollers, the reasonable control and optimization of parameters such as heating time, thermal uniformity, and temperature control accuracy have become a key technology to ensure product quality. At present, traditional heating roller quality inspection methods mostly rely on manual experience and physical measurement. Usually, manual inspection and classification are carried out after production. This method is not only inefficient and time-consuming, but also easily affected by human factors, making it difficult to guarantee stability and accuracy. In addition, existing technologies often lack real-time monitoring of various dynamic factors during the heating roller production process, and cannot identify quality problems in production in a timely manner, resulting in large quality fluctuations in the final product of the heated roller, which in turn affects the final printing effect and production pass rate. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] Therefore, this invention provides an AI-based method and system for monitoring the production quality of printer heating rollers, which solves the problem that existing technologies often lack real-time monitoring of various dynamic factors during the production process of heating rollers, and cannot identify quality problems in production in a timely manner, resulting in large quality fluctuations in the final product of the heated rollers, which in turn affects the final printing effect and production pass rate.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an AI-based method for monitoring the production quality of printer heating rollers, comprising, Collect heating roller material parameters, use XGBoost regressors to predict production parameters, and combine temperature response to calculate heating roller heating time; The time window is divided according to the heating time and multimodal data of the heating roller is collected to calculate multimodal features. Multimodal fusion is performed by combining multi-branch neural network to form fusion characterization. The fusion characterization is analyzed by linear regression to output the quality classification result of the heating roller. The produced heating rollers are classified according to the quality classification results, and the production qualification rate of heating rollers is statistically displayed and stored as a quality monitoring record.

[0006] As a preferred embodiment of the AI-based printer heating roller production quality monitoring method of the present invention, the method involves: collecting heating roller material parameters, using XGBoost regressors to predict production parameters, and combining temperature response to calculate the heating roller's heating time to obtain heating roller material parameters, forming feature rows x, and using four independent XGBoost regressors to construct a production parameter prediction model, which is then analyzed using historical material parameter feature rows. The production parameter prediction model is trained using these samples. Four production parameter prediction models were obtained through training. The predicted production parameters were then output by the trained production parameter prediction models and used as a reference for the production parameters of the heating roller. Calculate the equivalent thermal resistance of the heating roller based on its material parameters. and equivalent heat capacity ; The time constant is calculated and the temperature response is defined by using the equivalent thermal resistance and equivalent heat capacity of the heating roller; The heating time of the heating roller is calculated by back-calculating the temperature response. .

[0007] As a preferred embodiment of the AI-based printer heating roller production quality monitoring method of the present invention, wherein: the step of dividing the time window according to the heating time and collecting multimodal data of the heating roller to calculate multimodal features refers to dividing the time window according to the heating time. Based on the coordinate system defined by the heating roller, the axial direction is Zhou Xiangwei Multimodal sensors are time-aligned to collect multimodal data of the heating roller, including the thickness of the roller surface coating. Fangbu Infrared temperature And linear array vision I; Fixed slice size in space Slice Construct equal-width bins (B-type bins) from the scalar field in the statistical slices, and calculate the probability of each bin. ; Simultaneous use of Gaussian kernel Estimating continuous density ; Based on binning probability and continuous density Calculate Shannon entropy in the discrete domain And calculating differential entropy in the continuous domain ; The entropy features of the slices are formed by splicing Shannon entropy and differential entropy according to their corresponding slices. And obtain the roller coating thickness of the corresponding slice. Fangbu Infrared temperature In addition, linear visual I and entropy features are concatenated, and multimodal features are output by slice. .

[0008] As a preferred embodiment of the AI-based printer heating roller production quality monitoring method of the present invention, wherein: the multi-modal fusion of the combined multi-branch neural network to form a fusion characterization, and the output of the heating roller quality classification result by linear regression analysis of the fusion characterization, refers to the processing of multi-modal features using a multi-branch neural network, including visual branch, thickness branch and electrical branch; Output of branched neural networks Perform global average pooling to obtain the branch vector; Concatenate all branch vectors to form the global vector And generate modal weights using Softmax; The branch vectors are fused based on the modal weights to generate a fused representation. ; The heating roller quality classification probability is output through Softmax multi-class combined linear regression. .

[0009] As described in this invention, based on A preferred embodiment of a method for monitoring the production quality of printer heating rollers, wherein: classifying the produced heating rollers according to the heating roller quality classification results and displaying the production qualification rate of the heating rollers refers to selecting the quality classification result with the highest probability as the heating roller quality inspection result based on the calculated heating roller quality classification probability, selecting heating rollers of the same type, and simultaneously displaying the production qualification rate to production personnel based on the heating roller quality inspection results.

[0010] As a preferred embodiment of the AI-based printer heating roller production quality monitoring method of the present invention, the step of forming and storing quality monitoring records refers to generating quality monitoring records from the heating roller quality inspection results and production pass rate and storing them in a quality database.

[0011] As a preferred embodiment of the AI-based printer heating roller production quality monitoring method of the present invention, the multimodal sensor includes a linear array camera, a THz-TDS sensor, a four-probe array, and an infrared thermal imager.

[0012] Secondly, this invention provides an AI-based printer heating roller production quality monitoring system, including: The data collection module is used to collect the material parameters of the heating roller, use the XGBoost regressor to predict the production parameters, and combine the temperature response to calculate the heating time of the heating roller. The quality monitoring module is used to divide the time window according to the heating time and collect multimodal data of the heating roller to calculate multimodal features. It combines multi-branch neural network to perform multimodal fusion to form a fusion characterization. The fusion characterization is analyzed by linear regression to output the quality classification result of the heating roller. The classification and recording module is used to classify the produced heating rollers according to the quality classification results, display the production qualification rate of heating rollers, and store the quality monitoring records.

[0013] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the AI-based printer heating roller production quality monitoring method described in the first aspect of the present invention.

[0014] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the AI-based printer heating roller production quality monitoring method as described in the first aspect of the present invention.

[0015] The beneficial effects of this invention are as follows: This invention collects the material parameters of the heating roller, uses an XGBoost regressor to predict the production parameters, calculates and extracts features based on multimodal data, and then uses a multi-branch neural network to perform data fusion and quality classification, thereby providing an accurate assessment of the production quality of the heating roller and realizing real-time monitoring and feedback of the production process. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the AI-based printer heating roller production quality monitoring method in Example 1.

[0018] Figure 2 This is a structural diagram of the AI-based printer heating roller production quality monitoring system in Example 1. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides an AI-based method for monitoring the production quality of printer heating rollers, including the following steps: S1. Collect the material parameters of the heating roller, use the XGBoost regressor to predict the production parameters, and calculate the heating time of the heating roller in combination with the temperature response; Specifically, the material parameters of the heating roller are collected, the XGBoost regressor is used to predict production parameters, and the heating time of the heating roller is calculated by combining the temperature response to obtain the material parameters of the heating roller, including thermal conductivity k and density. Specific heat c, coating resistivity , inner and outer radius and Roller length L, internal and external convective heat transfer coefficient and and heating target temperature The feature row x is generated, and a production parameter prediction model is constructed using four independent XGBoost regressors. The production parameters include heating power P and coating spraying speed. Coating spraying angle A and preheating temperature Through historical material parameter characteristics The production parameter prediction model was trained using these samples: Where N is the total number of training samples. For the first The actual value of each training sample. For the first The predicted values ​​for each training sample, y, represent the output of the production parameter prediction model, including heating power P and coating spraying speed. Coating spraying angle A and preheating temperature ; Four production parameter prediction models were obtained through training, including a power prediction model. Speed ​​prediction model Angle prediction model and temperature prediction models The trained production parameter prediction model outputs predicted production parameters and serves as a reference for the production parameters of the heating roller. Calculate the equivalent thermal resistance of the heating roller based on its material parameters. and equivalent heat capacity : in For the heating roller Layer material, , For the heating roller The inner and outer radii of the layer material, For the first The thermal conductivity of the layer material, For the first Density of layer material, For the first Specific heat of layer material For the first The volume of the layer material is calculated using the radius; The time constant is calculated and the temperature response is defined using the equivalent thermal resistance and equivalent heat capacity of the heating roller: in It is a time constant. This refers to the heating time; Taking a linear steady-state approximation under power P The steady-state temperature indicates the temperature at which the heating roller operates under power. The stable temperature after a long period of time, and the temperature response can be expressed as: The temperature response formula of this invention can be derived by reverse deduction. The heating time of the heating roller is calculated by back-calculating the temperature response. : in It is a logarithmic function.

[0023] In traditional heated roller production, setting production parameters often relies on experience or rough estimations, which can easily lead to inaccurate temperature control and quality issues. XGBoost regressors, with their precise predictions, allow for accurate parameter setting based on the material properties of the heated roller, optimizing the heating process, ensuring precise temperature control, and reducing quality problems caused by uneven or inaccurate temperatures. Multimodal data fusion captures more comprehensive details of the heated roller production process, improving the accuracy and stability of quality assessment. This fusion method ensures the complementarity of data from various sensors, avoiding errors from single-modal data, and enabling more accurate classification and prediction of heated roller quality. Continuous monitoring of heated roller quality and real-time feedback to the production system allows for timely detection and adjustment of quality fluctuations, reducing scrap rates and improving production efficiency. Furthermore, the generated quality monitoring records not only serve as a basis for quality traceability but also help further optimize the production process, providing data support for subsequent production improvements.

[0024] S2. Divide the time window according to the heating time and collect multimodal data of the heating roller to calculate multimodal features. Combine multimodal fusion with multi-branch neural network to form fusion characterization. Output the heating roller quality classification result through linear regression analysis of fusion characterization. Specifically, the time window is divided according to the heating time, and multimodal data of the heating roller is collected to calculate the multimodal characteristic index. Based on the coordinate system defined by the heating roller, the axial direction is Zhou Xiangwei The roller surface is naturally a cylindrical and flattened structure, which allows data to be projected onto the same flat surface when multimodal sensors collect data. Multimodal sensors are time-aligned to collect multimodal data of the heating roller, including the thickness of the roller surface coating. Fangbu Infrared temperature and linear array vision ; The thickness of the roller coating Obtained using the time difference of flight method: in At the speed of light, The time difference between the front and rear reflected echoes (obtained by terahertz time-domain spectroscopy). The refractive index of the coating; The sheet resistance Calculations based on the surface resistance of the coating on the conductive substrate: in External pin voltage, The current of the inner needle (obtained by a four-probe array); The infrared temperature Calculated by measuring radiation temperature and ambient temperature: in For radiation temperature, For ambient temperature (acquired by infrared thermal imaging). The emissivity of the roller coating is calibrated using a blackbody patch. The linear array vision I is acquired by a linear array camera as it rotates with the heating roller and falls onto the display plane. superior; In space (exhibition plane) (Use fixed slice size) Slice Construct equal-width bins (B-type) based on the scalar fields (e.g., thickness, sheet resistance, infrared temperature) in the statistical slices, and calculate the probability of each bin. : in For the first Data count for each bin; Simultaneous estimation of continuous density using Gaussian kernel KDE : in For the total number of data, For the first One data point, For bandwidth parameters, Silverman bandwidth is used for estimation. For Gaussian kernel; Based on binning probability and continuous density Calculate Shannon entropy in the discrete domain And calculating differential entropy in the continuous domain : The entropy features of the slices are formed by splicing Shannon entropy and differential entropy according to their corresponding slices. And obtain the roller coating thickness d and sheet resistance of the corresponding slice. Infrared temperature In addition, linear visual I and entropy features are concatenated, and multimodal features are output by slice. : in and This is the index of the slice in the flattened diagram.

[0025] By dividing the heating process into long-term windows and simultaneously collecting multimodal data (coating thickness, sheet resistance, infrared temperature, and linear array vision), comprehensive monitoring of the heated roller production process was achieved. This was accomplished by projecting the multimodal data onto a unified flattened coordinate system. This allows data from different sensors to be integrated within the same spatial framework. This process effectively solves the spatial inconsistency problem between different data sources, ensuring data alignment and consistency. Through time alignment, the system can accurately and synchronously analyze the state of the heating roller at various moments, ensuring data consistency along the time axis. This improves the real-time monitoring capability of the heating roller production process, providing real-time feedback for optimizing production processes and improving product quality. Gaussian kernel density estimation is used in this process. Entropy calculation improves the accuracy of heating roller quality assessment, unlike traditional simple statistical methods. The system can capture the continuous distribution characteristics of data, avoiding potential information loss caused by data discretization. By calculating Shannon entropy and differential entropy, the system can quantify the uncertainty and complexity of data in different slice regions, making the quality assessment of the heating roller more accurate. It can identify local fluctuations and non-uniformity of parameters such as coating thickness and sheet resistance in space, thus providing a reliable basis for subsequent quality classification and production optimization. By splicing data features of different modalities (coating thickness, sheet resistance, infrared temperature, linear array vision, and entropy features) to form a multimodal feature matrix, this multimodal feature fusion method can more comprehensively reflect the thermal performance, electrical performance, and surface condition of the heating roller during the production process compared to single-modal data analysis. By integrating data from multiple sensors, the system can overcome the information that may be missed by a single modality, improving the accuracy and reliability of heating roller quality assessment. The fused feature data provides a more accurate basis for subsequent quality classification and pass rate statistics, thereby making quality control in the production process more refined and reducing the generation of defective products.

[0026] It should be noted that multimodal sensors include linear scan cameras, (Terahertz time-domain spectroscopy), four-probe array, infrared thermal imaging.

[0027] Furthermore, multi-branch neural networks are combined to perform multi-modal fusion to form a fusion representation. The fusion representation is analyzed by linear regression to output the heating roller quality classification result. The multi-branch neural network is used to process the multi-modal features, including visual branches, thickness branches and electrical branches. The visual branch uses lightweight Networks, visual features Processing: in Output feature maps for the visual branch; The thickness branch uses a shallow layer. For thickness features Processing: in Output feature maps for the thickness branch; The electrical branch uses a 3-layer CNN to analyze electrical features. Processing: in Output feature maps for the electrical branch; The lightweight ResNet network, shallow UNet, and 3-layer CNN were all obtained through pre-training. Output of branched neural networks Global average pooling is used to obtain the branch vector: All branch vectors are concatenated to form a global vector g, and the modal weights are generated using Softmax. in and For the weight matrix and bias, , as well as These are the modal weights for the visual branch, thickness branch, and electrical branch, respectively. The branch vectors are fused based on the modal weights to generate a fused representation z: in This indicates that the components are spliced ​​together by channel 1. 1. Linear transformation to align dimensions; The heating roller quality classification probability is output through Softmax multi-class combined linear regression. : in For the first The log odds of the probability of a classification (e.g., excellent, good, average, poor). and For classification weights and biases.

[0028] The multi-branch neural network employs three branches to process visual, thickness, and electrical features respectively. This allows it to handle the heterogeneity of multi-source data and uncover the relationships between different data sets. The visual branch extracts image features using a ResNet network, the thickness branch captures spatial information of the coating thickness using a UNet network, and the electrical branch uses a CNN to extract electrical features such as sheet resistance. This branching structure allows each network branch to focus on processing data specific to its domain, avoiding information confusion and error propagation during data fusion. It ensures in-depth analysis and processing of each modality's data, thereby improving the accuracy of heating roller quality prediction. The Softmax function is used to calculate the weights for each modality, enabling the system to adjust the quality prediction based on different modalities. The contribution of each modality is automatically adjusted to ensure the adaptability of different data sources (such as vision, thickness, and electrical) in the multimodal data fusion process. This avoids error bias caused by too much or too little information from a certain modality. In this way, the model can adaptively adjust the weights of different modalities under different production batches or conditions, thereby improving the robustness and accuracy of quality classification. By weighted fusion of the outputs of different branches, the final heating roller quality classification representation is generated. It is not just a simple splicing, but adjusts the contribution of each branch to the final result through modal weights, ensuring that the advantages of different modalities are brought into play. The fused representation is used for multi-class classification tasks, and the quality classification probability of the heating roller is output through a linear regression model.

[0029] S3. Classify the produced heating rollers according to the quality classification results, statistically analyze the qualified rate of heating roller production, display the results, and store the quality monitoring records. Specifically, the heating rollers produced are classified according to the quality classification results, and the production qualification rate of the heating rollers is statistically displayed. This means that the quality classification result with the highest probability is selected as the quality inspection result of the heating rollers based on the calculated quality classification probability, and heating rollers of the same type are selected. At the same time, the production qualification rate is statistically displayed to the production personnel based on the heating roller quality inspection results.

[0030] Furthermore, the formation of quality monitoring record storage refers to generating quality monitoring records based on the heating roller quality inspection results and production pass rate and storing them in the quality database.

[0031] This embodiment also provides an AI-based printer heating roller production quality monitoring system, including: The data collection module is used to collect the material parameters of the heating roller, use the XGBoost regressor to predict the production parameters, and combine the temperature response to calculate the heating time of the heating roller. The quality monitoring module is used to divide the time window according to the heating time and collect multimodal data of the heating roller to calculate multimodal features. It combines multi-branch neural network to perform multimodal fusion to form a fusion characterization. The fusion characterization is analyzed by linear regression to output the quality classification result of the heating roller. The classification and recording module is used to classify the produced heating rollers according to the quality classification results, display the production qualification rate of heating rollers, and store the quality monitoring records.

[0032] This embodiment also provides a computer device applicable to the AI-based printer heating roller production quality monitoring method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the AI-based printer heating roller production quality monitoring method proposed in the above embodiment.

[0033] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0034] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the AI-based method for monitoring the production quality of printer heating rollers as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0035] In summary, this invention collects the material parameters of the heating roller, uses an XGBoost regressor to predict production parameters, calculates and extracts features based on multimodal data, and then uses a multi-branch neural network to perform data fusion and quality classification, providing accurate assessment of the production quality of the heating roller and enabling real-time monitoring and feedback of the production process.

[0036] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An AI-based method for monitoring the production quality of printer heating rollers, characterized in that: include, Collect heating roller material parameters, use XGBoost regressors to predict production parameters, and combine temperature response to calculate heating roller heating time; The time window is divided according to the heating time and multimodal data of the heating roller is collected to calculate multimodal features. Multimodal fusion is performed by combining multi-branch neural network to form fusion characterization. The fusion characterization is analyzed by linear regression to output the quality classification result of the heating roller. The produced heating rollers are classified according to the quality classification results, and the production qualification rate of heating rollers is statistically displayed and stored as a quality monitoring record.

2. The AI-based printer heating roller production quality monitoring method as described in claim 1, characterized in that: The process involves collecting heating roller material parameters, using XGBoost regressors to predict production parameters, and combining this with temperature response calculations to determine the heating roller's heating time. This process generates a feature row x, and four independent XGBoost regressors are used to construct a production parameter prediction model. The model is then analyzed using historical material parameter feature rows. The production parameter prediction model is trained using these samples. Four production parameter prediction models were obtained through training. The predicted production parameters were then output by the trained production parameter prediction models and used as a reference for the production parameters of the heating roller. Calculate the equivalent thermal resistance of the heating roller based on its material parameters. and equivalent heat capacity ; The time constant is calculated and the temperature response is defined by using the equivalent thermal resistance and equivalent heat capacity of the heating roller; The heating time of the heating roller is calculated by back-calculating the temperature response. .

3. The AI-based method for monitoring the production quality of printer heating rollers as described in claim 2, characterized in that: The process of dividing the time window based on the heating duration and collecting multimodal data from the heating roller to calculate multimodal features refers to dividing the time window based on the heating duration. Based on the coordinate system defined by the heating roller with the axial direction as X and the circumferential direction as S, multimodal sensors are time-aligned to collect multimodal data of the heating roller, including the roller surface coating thickness d and sheet resistance. Infrared temperature And linear array vision I; Fixed slice size in space Slice Construct equal-width bins (B-type bins) from the scalar field in the statistical slices, and calculate the probability of each bin. ; Simultaneous use of Gaussian kernel Estimating continuous density ; Based on binning probability and continuous density Calculate Shannon entropy in the discrete domain And calculating differential entropy in the continuous domain ; The entropy features of the slices are formed by splicing Shannon entropy and differential entropy according to their corresponding slices. And obtain the roller coating thickness d and sheet resistance of the corresponding slice. Infrared temperature In addition, linear visual I and entropy features are concatenated, and multimodal features are output by slice. .

4. The AI-based printer heating roller production quality monitoring method as described in claim 3, characterized in that: The method of combining multi-branch neural networks to perform multimodal fusion to form a fusion representation, and outputting the heating roller quality classification result through linear regression analysis, refers to the processing of multimodal features using multi-branch neural networks, including visual branches, thickness branches, and electrical branches. Output of branched neural networks Perform global average pooling to obtain the branch vector; Concatenate all branch vectors to form the global vector And generate modal weights using Softmax; The branch vectors are fused according to the modal weights to generate a fused representation z; The heating roller quality classification probability is output through Softmax multi-class combined linear regression. .

5. The AI-based printer heating roller production quality monitoring method as described in claim 4, characterized in that: The process of classifying the produced heating rollers according to their quality classification results and displaying the production qualification rate of the heating rollers refers to selecting the quality classification result with the highest probability based on the calculated quality classification probability as the quality inspection result of the heating rollers, selecting heating rollers of the same type, and displaying the production qualification rate to the production personnel based on the production qualification results of the heating roller quality inspection results.

6. As claimed The AI-based method for monitoring the production quality of printer heating rollers is characterized by: The term "forming quality monitoring record storage" refers to generating quality monitoring records based on the heating roller quality inspection results and production pass rate and storing them in a quality database.

7. The AI-based printer heating roller production quality monitoring method as described in claim 3, characterized in that: The multimodal sensor includes a linear array camera, a THz-TDS sensor, a four-probe array, and an infrared thermal imager.

8. An AI-based printer heating roller production quality monitoring system, based on the AI-based printer heating roller production quality monitoring method according to any one of claims 1 to 7, characterized in that: include, The data collection module is used to collect the material parameters of the heating roller, use the XGBoost regressor to predict the production parameters, and combine the temperature response to calculate the heating time of the heating roller. The quality monitoring module is used to divide the time window according to the heating time and collect multimodal data of the heating roller to calculate multimodal features. It combines multi-branch neural network to perform multimodal fusion to form a fusion characterization. The fusion characterization is analyzed by linear regression to output the quality classification result of the heating roller. The classification and recording module is used to classify the produced heating rollers according to the quality classification results, display the production qualification rate of heating rollers, and store the quality monitoring records.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the AI-based printer heating roller production quality monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the AI-based printer heating roller production quality monitoring method according to any one of claims 1 to 7.

Images