An AI-algorithm-based dynamic allocation system for cloud printing order resources

Abstract

The present invention relates to the field of artificial intelligence technology, and particularly to a dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms, including an order analysis and clustering module, a cloud printer performance status perception module, an order matching and allocation strategy formulation module, and a real-time monitoring and dynamic adjustment module; the order analysis and clustering module uses composite technology to extract order features, conducts clustering analysis and determines its urgency; the cloud printer performance status perception module collects information through the Internet of Things and sensor networks, monitors the status of cloud printers in real time and predicts the probability of failure; the order matching and allocation strategy formulation module optimizes the order allocation strategy according to the order urgency and the status of cloud printers to improve efficiency; the real-time monitoring and dynamic adjustment module ensures the smooth completion of printing tasks by monitoring the working efficiency of cloud printers in real time, comparing with order requirements and making timely adjustments. The present invention is used to solve the technical problem of poor processing effect of current cloud printing orders.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly to a dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms. Background Art

[0002] In the current processing of cloud printing orders, the following problems still exist: the order processing efficiency is low, the whole process from order reception, allocation to completion of printing takes a long time, resulting in a long waiting time for users, a high degree of manual participation, a low automation level, and affecting the overall efficiency; the fault prediction of cloud printers is inaccurate, resulting in false alarms or missed alarms, affecting the equipment maintenance plan, untimely fault handling, prolonging the equipment downtime, and affecting the continuity of printing services; the order allocation is unreasonable, resulting in some printers being overloaded while other printers are idle, and the resource utilization is unbalanced; the abnormal handling during the printing process is not timely. When abnormalities such as paper jams and insufficient ink occur during the printing process, it is impossible to respond quickly; there is energy waste, and the printers remain in the operating state during non-working hours or at low loads, causing energy waste. Therefore, a series of targeted methods are urgently needed to address these problems to improve the efficiency and level of cloud printing order processing. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems in the background art, and to propose a dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] A dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms, including: an order analysis and clustering module, a cloud printer performance status perception module, an order matching and allocation strategy formulation module, and a real-time monitoring and dynamic adjustment module;

[0006] Order analysis and clustering module: Using composite technologies to extract order features, classifying orders based on the extracted features using clustering algorithms, determining the urgency of orders by comprehensively considering multi-dimensional processing factors of orders, calculating the urgency coefficient and marking it on the clustered order groups; among them, the composite technologies include image recognition technology and natural language processing technology;

[0007] Cloud printer performance status perception module: Collecting cloud printer status information through Internet of Things technology and sensor networks, embedding an intelligent diagnosis unit to monitor component operation parameters in real time, and at the same time, using historical fault records and machine learning algorithms to predict the future fault probability of cloud printers, providing a basis for the availability assessment of cloud printers;

[0008] Order Matching and Allocation Strategy Formulation Module: For orders with a high emergency coefficient, prioritize matching cloud printers with the best performance and low failure probability, and consider batch processing capabilities to avoid delays; for orders with a low emergency coefficient, predict the idle time and energy consumption of cloud printers and optimize allocation to maximize energy efficiency;

[0009] Real-time Monitoring and Dynamic Adjustment Module: Monitor the working efficiency of cloud printers in real time and conduct comparative analysis. If any anomalies are found, re-allocate orders to other cloud printers.

[0010] Furthermore, the process of the Order Analysis and Clustering Module using composite technologies to extract order features, classifying orders based on the extracted features using clustering algorithms, and determining the emergency level of orders by considering multi-dimensional processing factors of orders, calculating the emergency coefficient, and marking it on the clustered order groups includes:

[0011] Adopt a combination of image recognition and natural language processing technologies to extract features from the content of each order: For the order part containing images or graphics, use a convolutional neural network for image decomposition to identify features such as the complexity of the graphics, the color mode of the image, and the resolution of the image; for the text part of the order, use a word vector model to convert the text into a vector representation and analyze key text information such as the text layout format, the semantic features of the text content, and the number of words; integrate various feature information obtained through image recognition and natural language processing into a complete order feature vector;

[0012] Based on the extracted order features, use a clustering algorithm, with the distance between order feature vectors as the similarity measurement standard, to perform clustering operations on the feature vectors of all orders and group orders with similar distances into one category; among them, orders with similar distances are orders with similar features: When the multi-dimensional differences in graphic complexity, color mode, and key text information of the order feature vectors are all less than the set clustering threshold, these orders are classified into the same cluster; during the clustering process, continuously adjust the center and threshold of the cluster according to historical order data to adapt to the distribution changes of different types of orders;

[0013] According to the classification results of the order feature vectors by the clustering algorithm, further consider the multi-dimensional processing factors of the orders, calculate the emergency coefficient through weighted calculation, and mark this coefficient on the clustered order groups to reflect the emergency level of the orders, specifically including: Extract the timestamp when the order was submitted and calculate the interval from submission to the current time; identify the priority identifier set by the customer in the order, where the priority identifier set by the customer is divided into three levels: high, medium, and low;

[0014] Combined with multi-dimensional processing factors, a weighted calculation method is adopted to obtain an emergency coefficient, and the value range of the emergency coefficient is set; among them, the multi-dimensional processing factors of the order include the order submission time, the customer-set priority, the order attributes, and the order size.

[0015] Obtain the emergency coefficient and mark the obtained emergency coefficient on each clustered order group.

[0016] Furthermore, the process by which the cloud printer performance status perception module collects cloud printer status information through Internet of Things technology and sensor network and embeds an intelligent diagnosis unit to monitor the operating parameters of components in real time includes:

[0017] Establish connections with each cloud printer connected to the cloud printing service through Internet of Things technology for comprehensive detection;

[0018] Use the sensor network to collect the general status information of the cloud printer;

[0019] Embed an intelligent diagnosis unit inside the cloud printer to evaluate the health status of components, where the intelligent diagnosis unit is used to monitor the operating parameters of the internal components of the cloud printer in real time;

[0020] Integrate all the information obtained from the sensor network and the intelligent diagnosis unit to construct a performance vector of the cloud printer.

[0021] Furthermore, the process by which the cloud printer performance status perception module predicts the future failure probability of the cloud printer using historical failure records and machine learning algorithms includes:

[0022] Collect historical failure records from the maintenance logs or databases of each cloud printer;

[0023] Check and process missing values, outliers, or duplicate records in the data;

[0024] Extract features related to failure prediction from historical failure records;

[0025] Use machine learning algorithms to construct a failure prediction model and complete model training;

[0026] Obtain the performance vector of the current cloud printer; record the usage duration of the cloud printer and the running time since the last repair;

[0027] Input the performance vector of the current cloud printer, the usage duration of the cloud printer, and the running time since the last repair into the trained failure prediction model to calculate the probability of each cloud printer failing in the future for a period of time;

[0028] The probability value output by the fault prediction model is used as an important basis for evaluating the availability of the cloud printer. Among them, the value range of the output probability value is 0 - 1, where 0 indicates no fault risk and 1 indicates a very high fault risk.

[0029] Furthermore, the process of the order matching and allocation strategy formulation module for orders with a high emergency coefficient to preferentially match the cloud printer with the best performance and low fault probability and consider the batch processing ability includes:

[0030] Set the normal operation probability range , satisfying ; At the same time, set the urgency range , satisfying , which is used to quantify the urgency of the order;

[0031] For the order clustering with an emergency coefficient higher than the upper limit of the urgency range , screen out the list of cloud printers with the best performance vector and the probability value output by the fault prediction model lower than the lower limit of the normal operation probability range ;

[0032] For the screened list of cloud printers, further analyze the batch processing ability of each cloud printer for the current order clustering; among them, the batch processing ability refers to the ability of the cloud printer to process multiple orders or printing tasks within a unit time.

[0033] Furthermore, the process of the order matching and allocation strategy formulation module for orders with a low emergency coefficient to predict the idle time and energy consumption of the cloud printer and optimize the allocation includes:

[0034] For the order clustering with an emergency coefficient lower than the lower limit of the urgency range , obtain the predicted idle time information of all cloud printers; at the same time, consider the energy consumption situation of each cloud printer;

[0035] According to the predicted idle time and energy consumption situation, find out the cloud printers that are about to enter a long idle state and have a relatively high current energy consumption, and then allocate the low - emergency - coefficient orders to this cloud printer and adjust the working mode of the cloud printer accordingly.

[0036] Furthermore, the process of the real - time monitoring and dynamic adjustment module to continuously monitor the working efficiency of the cloud printer, conduct a comparative analysis, and re - allocate orders to other cloud printers if abnormalities are found includes:

[0037] After the order is allocated to the cloud printer and the printing process is started, the sensors on the cloud printer side continuously monitor the working efficiency of the cloud printer;

[0038] Compare and analyze the monitored working efficiency of the cloud printer with the original requirements of the order and the expected quality standards: If any abnormality is found in the printing process of any cloud printer during the comparison and analysis, immediately suspend the printing task of this order on this cloud printer;

[0039] Re-evaluate the status of all available cloud printers, and re-allocate the remaining unprinted part of the order to other cloud printers; at the same time, adjust the allocation plan for subsequent orders according to the abnormal situation.

[0040] Compared with the existing technologies, the advantages of the dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms provided by the present invention are as follows:

[0041] 1. By using composite technologies to extract the characteristics of cloud printing orders, classifying the orders based on the extracted characteristics using clustering algorithms, determining the urgency of the orders by comprehensively considering multi-dimensional processing factors of the orders, and calculating the urgency coefficient and marking it on the clustered order groups, the accuracy and efficiency of feature extraction are improved, which helps to quickly identify and prioritize the orders with high urgency coefficients, and ensures that key tasks are processed in a timely manner;

[0042] 2. By using Internet of Things technology and sensor networks to collect printer status information, embedding an intelligent diagnosis unit to monitor the operating parameters of components in real time, and using historical failure records and machine learning algorithms to predict the future failure probability of printers, providing a basis for printer availability assessment, ensuring that the operating status of cloud printers is always grasped, helping to detect and handle potential failures in advance, reducing the failure rate, and improving the reliability and stability of cloud printers; by predicting the failure probability, a more scientific and reasonable maintenance plan can be formulated, reducing unnecessary downtime and maintenance costs, and extending the service life of cloud printers;

[0043] 3. For orders with high urgency coefficients, preferentially match cloud printers with the best performance and low failure probability. At the same time, considering the batch processing ability, the printing efficiency is further improved. For orders with low urgency coefficients, predict the idle time and energy consumption of cloud printers, and optimize the allocation to maximize energy efficiency, which helps to reduce operating costs and achieve green printing. Through real-time monitoring and dynamic adjustment, the probability of similar abnormalities is reduced, ensuring that orders can be completed on time and with high quality, and enhancing customer satisfaction and trust. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a module diagram of a dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms proposed by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Referring to Figure 1 , a dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms, the system includes an order analysis and clustering module, a cloud printer performance status perception module, an order matching and allocation strategy formulation module, and a real-time monitoring and dynamic adjustment module;

[0047] Order analysis and clustering module: Use composite technologies to extract order features, classify orders based on the extracted features using clustering algorithms, determine the urgency of orders by comprehensively considering multi-dimensional processing factors of orders, calculate the urgency coefficient and mark it on the clustered order group; among them, the composite technologies include image recognition technology and natural language processing technology;

[0048] Cloud printer performance status perception module: Collect cloud printer status information through Internet of Things technology and sensor networks, embed an intelligent diagnosis unit to monitor component operation parameters in real time, and at the same time, use historical fault records and machine learning algorithms to predict the future fault probability of cloud printers, providing a basis for the availability assessment of cloud printers;

[0049] Order matching and allocation strategy formulation module: For orders with a high urgency coefficient, preferentially match cloud printers with the best performance and low failure probability, and consider batch processing capabilities to avoid delays; for orders with a low urgency coefficient, predict the idle time and energy consumption of cloud printers, and optimize the allocation to maximize energy efficiency;

[0050] Real-time monitoring and dynamic adjustment module: Real-time monitor the working efficiency of cloud printers, and conduct comparative analysis. If abnormalities are found, re-allocate orders to other cloud printers.

[0051] The steps of the order analysis and clustering module using composite technologies to extract order features, classifying orders based on the extracted features using clustering algorithms, determining the urgency of orders by comprehensively considering multi-dimensional processing factors of orders, calculating the urgency coefficient and marking it on the clustered order group include:

[0052] Step 101: Adopt a combination of image recognition and natural language processing technologies to extract features from the content of each order:

[0053] For the order part containing images or graphics, a convolutional neural network is used for image decomposition to identify the characteristics of the complexity of the graphics, the color mode of the images, and the resolution of the images. Among them, the complexity of the graphics includes the number of lines, the types of colors, the number of nested layers of the graphics, etc., and the color mode of the images includes RGB, CMYK, etc.;

[0054] For the text part of the order, a word vector model is used to convert the text into a vector representation, and key text information such as the text layout format, the semantic characteristics of the text content, and the number of words is analyzed. Among them, the text layout format covers typesetting details such as font type, font size, line spacing setting, and paragraph format. Through semantic analysis, it is judged whether the text content belongs to specific categories such as technical documents and business documents. At the same time, the total amount of text is counted to obtain the length information of the document;

[0055] The various feature information obtained through image recognition and natural language processing is integrated into a complete order feature vector, which comprehensively describes the printing requirement characteristics of the order;

[0056] Step 102: Based on the extracted order features, use a clustering algorithm. Taking the distance between order feature vectors as the similarity measurement standard, perform a clustering operation on the feature vectors of all orders, and group the orders with close distances into one category; among them, the orders with close distances are orders with similar features:

[0057] When the multi-dimensional differences in the graphic complexity, color mode, and key text information of the order feature vectors are all less than the set clustering threshold, these orders are classified into the same cluster; during the clustering process, the center and threshold of the clustering are continuously adjusted according to historical order data to adapt to the distribution changes of different types of orders;

[0058] Step 103: According to the classification results of the order feature vectors by the clustering algorithm, further comprehensively consider the multi-dimensional processing factors of the order, calculate the urgency coefficient through weighted calculation, and mark this coefficient on the clustered order group to reflect the urgency of the order. Among them, the multi-dimensional processing factors of the order include the order submission time, the customer-set priority, the order attributes, and the order scale:

[0059] Extract the timestamp of the order submission and calculate the interval from the submission to the current time;

[0060] Identify the priority identifier set by the customer in the order. Among them, the priority identifier set by the customer is divided into three levels: high, medium, and low, which are used to directly reflect the customer's expectations and requirements for the order processing speed. If the customer does not set an identifier, the priority is estimated based on the order attributes and order size. Order attributes include, but are not limited to, business emergency documents, ordinary documents, etc. These attributes can indirectly reflect the urgency of the order. At the same time, the order size is also an important consideration factor. Since the processing of large documents may be more complex and require more time, their urgency may be relatively low;

[0061] Combining multi-dimensional processing factors such as the order submission time, customer-set priority, order attributes, and order size, a weighted calculation method is used to obtain an emergency coefficient. Among them, the emergency coefficient will comprehensively consider multiple dimensions such as the time interval of the order, the customer-set priority (or estimated priority), order attributes, and order size to ensure its accuracy and rationality;

[0062] Set the value range of the emergency coefficient: 0 - 1 (0 represents the lowest emergency level, 1 represents the highest emergency level);

[0063] Obtain the emergency coefficient and mark the obtained emergency coefficient on each clustered order group so that subsequent processing and scheduling work can be carried out more efficiently and orderly, ensuring that each order can be processed promptly and appropriately according to its emergency level;

[0064] For example, (1) Order feature extraction: For the order part containing images or graphics:

[0065] Step Q1: Assume there is an order image, which is represented as a matrix , and the element in the matrix represents the pixel value of the image at the coordinate ;

[0066] Let the convolution kernel be , and its size is (assuming ). For each position of the image, the convolution layer outputs , where

[0067] ,

[0068] In the formula, is the coordinate index inside the convolution kernel; for example, when ,

[0069] ;

[0070] Through the convolution operation, the image can be decomposed into different feature layers;

[0071] Step Q2: Set the number of lines of the graphic to , the number of color types to , and the nesting level of the graphic to ; if the graphic is a circle nested within a rectangle, the number of lines (the outline of the circle can be approximated as being composed of many short straight lines, assumed here to be 10 approximate straight lines), the number of color types (assumed to be a single color), and the nesting level (assumed to be one level of nesting); define the graphic complexity function , that is,

[0072] ,

[0073] wherein, are the weight coefficients of the number of lines, the number of color types, and the nesting level respectively, used to measure the relative importance of the number of lines, the number of color types, and the nesting level to the graphic complexity; if , then

[0074] ;

[0075] Step Q3: Assume the image is in the RGB color mode, then each pixel point in the image , where represent the values of the red, green, and blue channels respectively, with a value range of 0 - 255; represent the color mode characteristics by statistically calculating features such as the average value and variance of different color channels in the entire image;

[0076] ] Step Q4: Set the number of horizontal pixels of the image to , and the number of vertical pixels to ; assume the number of horizontal pixels of the image , and the number of vertical pixels , then the resolution is expressed as pixels;

[0077] For the text part in the order, assume the text part in the order is "This is a business document, the font is Song typeface, the font size is 12, the line spacing is 1.5 times the line spacing, the paragraph format is first line indented by 2 characters, and the content is mainly about the detailed information of the product order, with 50 words", then the specific processing steps are as follows:

[0078] Step R1: Use the word vector model. Assume there is a pre-trained word vector model, where each word is mapped to a dimensional vector space; for the word "business document", assume the mapped vector is ; similarly, the mapped vector for the word "Song typeface" is ; Therefore, the whole sentence can be represented as a combination of these word vectors. For example, concatenating them results in a long vector ;

[0079] Step R2: Let the font type be represented by (here = "Song typeface"), the font size , the line spacing setting , and the paragraph format = "First line indented by 2 characters"; Define the typesetting feature function , that is

[0080] ,

[0081] In the formula, are the weight coefficients of the font type, font size, line spacing setting, and paragraph format respectively, used to measure the relative importance of the font type, font size, line spacing setting, and paragraph format to the typesetting features; among them,

[0082] ;

[0083] Step R3: Since the text content belongs to a business document, mark its semantic category as (assuming the business document category is 1, and other categories have different markings), where is obtained according to the predefined semantic category encoding;

[0084] Step R4: Define the number of words ; Among them, the number of words can be directly used as a feature. For example, in subsequent processing, a larger number of words may mean more detailed order content, which may affect the urgency or processing method of the order, etc.;

[0085] (2) Order clustering operation: Assume there are three feature vectors of orders , and the dimension of the feature vector is 3 (composed of image complexity, a certain feature of the color mode (such as the average value of the red channel), text typesetting features, semantic features, and the number of words);

[0086] Define the distance metric. Arbitrarily select two feature vectors and , then the Euclidean distance is

[0087] ;

[0088] Define the clustering threshold , calculate , and ; If , then classify Order 1 and Order 2 into the same category;

[0089] Step Z1: Set the order submission time as , and the current time as , with the time interval , assuming hours, and the time weight as , where the time weight represents the relative importance of the order submission time in calculating the emergency coefficient;

[0090] Step Z2: Given that the customer sets the priority as high priority , with the priority weight , where is the value mapped according to the priority level set by the customer, representing the importance of the customer-set priority in the calculation of the emergency coefficient;

[0091] Step Z3: If the order attribute is a business emergency document, set the value corresponding to the business emergency document as , with the order attribute weight , where is the value mapped according to the type of the order attribute, representing the relative importance of the order attribute in the calculation of the emergency coefficient;

[0092] Step Z4: If the order size is small, set the coefficient corresponding to the size as , with the order size weight , where is the value mapped according to the size of the order (measured by, for example, the number of words, file size, etc.), representing the relative importance of the order size in the calculation of the emergency coefficient;

[0093] Step Z5: Comprehensively consider the order submission time, the customer-set priority, the order attribute, and the order size, and calculate the emergency coefficient , that is

[0094] .

[0095] The steps for the cloud printer performance status perception module to collect cloud printer status information through the Internet of Things technology and the sensor network, embed the intelligent diagnosis unit to monitor the component operation parameters in real time, and use the historical failure records and machine learning algorithms to predict the future failure probability of the cloud printer include:

[0096] Step 201: Through the Internet of Things technology, establish a connection with each cloud printer connected to the cloud printing service for comprehensive detection; use the sensor network to collect the general status information of the cloud printer;

[0097] In step 201, the sensor network is a network composed of multiple sensor nodes with different sensing capabilities distributed at different locations (including ink volume sensors, paper margin sensors, temperature sensors, humidity sensors, etc.). These sensor nodes are interconnected through wireless communication, wired communication, or hybrid communication methods to form a collaborative system for real-time monitoring, sensing, and collecting various status information of the cloud printer; the conventional status information includes the ink volume, paper margin, temperature, humidity, etc. of the cloud printer;

[0098] Step 202: Embed an intelligent diagnosis unit inside the cloud printer to evaluate the health status of components;

[0099] In step 202, the intelligent diagnosis unit is used to monitor the operating parameters of the internal components of the cloud printer in real time. For example, for an inkjet cloud printer, the health status of the print head is evaluated by monitoring parameters such as the inkjet frequency of the print head and the consistency of the ink droplet size; for a laser cloud printer, indicators such as the stability of the heating temperature of its heating component and the heating speed are monitored to evaluate its health status;

[0100] Step 203: Integrate all the information obtained from the sensor network and the intelligent diagnosis unit (i.e., the conventional status information of the cloud printer and the operating parameters of the internal components of the cloud printer) to construct a performance vector of the cloud printer. Among them, the performance vector contains information about various aspects of the cloud printer, such as ink volume, paper margin, print head health, heating component efficiency, overall temperature of the cloud printer, overall humidity of the cloud printer, etc.;

[0101] Step 204: Collect historical fault records from the maintenance log or database of each cloud printer. Among them, the historical fault records include but are not limited to the fault type, the time of fault occurrence, and detailed repair records (such as repair time, repair content, etc.);

[0102] Step 205: Check and process missing values, outliers, or duplicate records in the data;

[0103] Step 206: Extract features related to fault prediction from the historical fault records, such as fault type, fault frequency, repair time interval, etc. It may also be necessary to transform time-related features, such as converting the timestamp into a specific time interval or time period;

[0104] Step 207: Construct a fault prediction model using machine learning algorithms and complete model training. Select the random forest algorithm. Based on the historical fault record data, divide the data set into a training set and a test set, usually in a ratio of 70% for the training set and 30% for the test set (or adjust according to specific circumstances). Use the training set data to train the selected random forest algorithm. During the training process, it may be necessary to adjust the model parameters to optimize its performance. Use the test set data to evaluate the trained model, calculate performance metrics such as accuracy, recall rate, and F1 score, and optimize the model according to the evaluation results to improve its prediction performance.

[0105] Step 208: Obtain the performance vector of the current cloud printer. Record the usage duration of the cloud printer and the running time since the last repair.

[0106] Step 209: Input the performance vector of the current cloud printer, as well as the usage duration of the cloud printer and the running time since the last repair, into the trained fault prediction model, and calculate the probability of each cloud printer having a fault within a future period (such as the next hour).

[0107] Step 210: Use the probability value output by the fault prediction model as an important basis for evaluating the availability of the cloud printer. The value range is 0 - 1, where 0 indicates no fault risk and 1 indicates a very high fault risk.

[0108] For example, Step P1: Collect historical fault records from the maintenance logs or databases of

[0109] cloud printers. Among them, the historical fault record of each cloud printer includes the fault type (assuming there are 5 types of fault types), the fault occurrence time , and the repair record (including the repair time and the repair content); among them, represents the number of cloud printers, represents the cloud printer index, represents the -th fault occurrence time. For each cloud printer , there are at most

[0110] fault records.

[0111] Step P2: For missing values in the data, if the fault occurrence time of a certain printer is missing, fill it according to the average value of the previous and subsequent fault occurrence times (assuming a reasonable time series pattern).

[0112] For duplicate records, directly delete the redundant identical records;

[0113] Step P3. For each printer , the failure frequency is , where is a fixed observation period, and days, the failure frequency is represents the proportion of the number of failures within this period;

[0114] Convert the failure occurrence time to the number of days relative to a fixed starting time, then the maintenance time interval is , where represents the th maintenance time interval relative to the failure occurrence;

[0115] Construct a performance vector , including features related to failure frequency and maintenance time interval for failure prediction;

[0116] Step P4. Divide the feature data (i.e., the constructed performance vectors) of the cloud printers collected into a training set and a test set according to the ratio of 70% training set and 30% test set. The training set contains samples, and the test set contains samples;

[0117] Let the number of decision trees in the random forest model be . During the training process, adjust other parameters of the random forest model according to the data in the training set, such as the maximum depth of each decision tree;

[0118] For the samples in the test set, let the number of samples predicted as failures be , the number of samples actually being failures be , and the number of samples correctly predicted as failures be ;

[0119] Calculate the accuracy rate , recall rate and F1 score , where the accuracy rate represents the proportion of correctly predicted samples among the samples predicted as failures, and the recall rate represents the proportion of correctly predicted failure samples among the actual failure samples, The F1 score comprehensively considers the performance indicators of accuracy rate and recall rate;

[0120] Based on the analysis of these evaluation results, if the accuracy rate is found to be low, the number of decision trees needs to be increased. , if the recall rate is low, the construction rules of the decision tree need to be adjusted, etc.

[0121] Step P5, obtain the performance vector and relevant time information of the current cloud printer: If there is currently one cloud printer , then its performance vector is ;

[0122] Obtain the usage duration , the running time after the most recent repair days;

[0123] Input the performance vector , the usage duration , and the running time after the most recent repair into the trained fault prediction model (random forest model);

[0124] The fault prediction model calculates the probability of this cloud printer having a fault within the next hour according to the internal decision tree structure and the trained parameters , where the calculation of is based on the prediction results of each decision tree in the random forest model for this sample (for example, the voting proportion of each decision tree judging this sample as a fault);

[0125] Suppose it is calculated by the fault prediction model that , indicating that this cloud printer has a 30% fault risk within the next hour. According to the set value range of 0 - 1, it is in a state of having a certain fault risk.

[0126] For orders with a high emergency coefficient, the order matching and allocation strategy formulation module preferentially matches the cloud printer with the best performance and low fault probability, and considers the batch processing ability. For orders with a low emergency coefficient, it predicts the idle time and energy consumption of the cloud printer. The steps for optimizing the allocation include:

[0127] Step 301, set the normal operation probability range , satisfying , indicating that the cloud printer is in a running state from a relatively low fault risk to a relatively high fault risk but still within an acceptable range; at the same time, set the emergency level range , satisfying , which is used to quantify the emergency level of the order;

[0128] Step 302, cluster the orders with an emergency coefficient higher than the upper limit of the emergency level range (such as 0.7), and screen out the ones with the best performance vector and the probability value output by the fault prediction model lower than the lower limit of the normal operation probability range A list of cloud printers with a failure risk level (e.g., 0.1 indicates a very low failure risk).

[0129] In step 302, during the screening process, for the evaluation of the performance vector, multiple dimensions are comprehensively considered, such as the printing speed of the cloud printer, the stability of printing quality (judged according to factors such as printhead health and heating component efficiency), and the paper handling ability (such as supported paper types, maximum paper thickness, etc.).

[0130] Step 303: For the screened list of cloud printers, further analyze the batch processing ability of each cloud printer for the current order clustering. If any cloud printer has special functions or algorithms, that is, it can process multiple orders simultaneously without affecting the printing quality (for example, some high-end cloud printers can optimize task scheduling to achieve interleaved printing of multiple images or documents and improve printing efficiency), then give priority to allocating such orders to this cloud printer. At the same time, when allocating orders, the current task queue situation of the cloud printer also needs to be considered to avoid printing delays caused by task backlogs.

[0131] Step 304: For order clusters with an emergency coefficient lower than the lower limit of the emergency level range (such as 0.3), obtain the predicted idle time information of all cloud printers. Among them, the predicted idle time is calculated comprehensively based on factors such as the current task queue of the cloud printer, the estimated completion time of tasks, and the strategy of the cloud printer entering the low-power mode. At the same time, consider the energy consumption situation of each cloud printer. Among them, the energy consumption data is obtained by long-term monitoring of the energy consumption of the cloud printer in different working modes (such as standby, printing, preheating, etc.).

[0132] Step 305: According to the predicted idle time and energy consumption situation, find the cloud printers that are about to enter a long idle state and have a relatively high current energy consumption, and then allocate the low-emergency coefficient orders to these cloud printers and adjust the working mode of the cloud printers accordingly.

[0133] In step 305, the working modes of the cloud printer include but are not limited to the standard mode and the energy-saving mode. The standard mode is the normal working mode of the cloud printer, which provides normal printing speed, printing quality, and energy consumption. In this mode, the cloud printer can complete the printing task according to the standard settings and is suitable for most regular orders. The energy-saving mode means that when the cloud printer is expected to be idle for a long time or the current energy consumption is high, it can be switched to this mode. In this mode, the cloud printer will reduce its energy consumption (such as by reducing the printing speed, adjusting the printing brightness or color mode, etc.). It is understandable that if a cloud printer is expected to be idle for more than 20 minutes, the working mode of the cloud printer will be adjusted from the standard mode to the energy-saving mode. At the same time, in order to ensure that the order can still be completed on time in the energy-saving mode, the printing parameters of the order are further adjusted (such as reducing the printing speed, adjusting the color mode, etc.) to adapt to the performance of the cloud printer in the energy-saving mode and maximize the energy efficiency.

[0134] The steps for the real-time monitoring and dynamic adjustment module to monitor the working efficiency of the cloud printer in real time, conduct comparative analysis, and reassign the order to other cloud printers if any abnormality is found include:

[0135] Step 401: After the order is assigned to the cloud printer and the printing process is started, sensors on the cloud printer side (such as speed sensors and image quality detection sensors) monitor the working efficiency of the cloud printer in real time. Among them, the speed sensor is responsible for monitoring the paper conveyance speed in real time to judge whether the printing speed is stable; the image quality detection sensor conducts quality detection on the printed image or text to check for problems such as blurring, ink shortage, and color deviation.

[0136] Step 402: Compare and analyze the monitored working efficiency of the cloud printer with the original requirements and expected quality standards of the order. If any abnormality is found in any cloud printer during the printing process in the comparative analysis, such as the printing speed suddenly drops by more than a certain proportion (such as 10%) or the printing quality shows an obvious deviation (such as the color deviation exceeds the set threshold), the printing task of this order on this cloud printer will be immediately suspended.

[0137] In step 402, the original requirements and expected quality standards of the order include but are not limited to the following aspects: The original requirements of the order specify the expected printing speed or the time frame for printing completion, which is to ensure that the order can be completed within the specified time to meet the customer's delivery date or other time-sensitive requirements; the expected quality standards detail the quality requirements of the printed output, such as the clarity of the image, the readability of the text, the accuracy and consistency of the color, etc.

[0138] Step 403: Re-evaluate the status of all available cloud printers (including performance vectors, failure probabilities, current task queues, etc.), and re-allocate the orders for the remaining unprinted parts to other cloud printers; when re-allocating, consider factors such as the warm-up time of the new cloud printer (if it is a laser cloud printer), the remaining amount of ink cartridges or toner cartridges, etc., to ensure that the orders can continue to be printed without affecting the overall progress and quality;

[0139] Step 404: Adjust the allocation plan for subsequent orders according to the abnormal situation. For example, for cloud printers that may cause similar abnormalities, reduce the number of orders allocated to it within a certain period of time until its problems are resolved or its performance is improved;

[0140] For example, assume there are 3 cloud printers, labeled as Printer , Printer and Printer ;

[0141] Let the total number of pages of the order be pages;

[0142] The order is allocated to 3 cloud printers in proportion. It is set that the number of pages allocated to Printer is pages, the number of pages allocated to Printer is , and the number of pages allocated to Printer is ;

[0143] Define the normal printing speed of Printer as pages per minute, the normal printing speed of Printer as pages per minute, and the normal printing speed of Printer as pages per minute; among them, the normal printing speed of the cloud printer refers to the average number of pages printed per minute under normal and stable working conditions;

[0144] Set the proportional threshold for the sudden drop in the set printing speed as . During the printing process, if the speed sensor of Printer detects that its printing speed has dropped to pages per minute, calculate the speed drop ratio as , then it is determined that Printer has an abnormality;

[0145] At this time, a part of the order has been printed. If Printer has printed pages, then Printer The remaining number of unprinted pages is ;

[0146] Re-evaluate the printer and the printer status, assuming the printer The tasks in the current task queue are expected to take minutes to complete (considering factors such as its performance vector and the current task queue), the printer The tasks in the current task queue are expected to take minutes to complete; and the printer The remaining amount of the ink cartridge can support printing pages, the printer The remaining amount of the ink cartridge can support printing pages. Since the remaining task completion time of the printer is relatively short, although the remaining amount of its ink cartridge is relatively small, considering comprehensively, re-allocate the remaining unprinted orders of the printer to the printer ;

[0147] In subsequent order allocations, due to this exception occurring in the printer reduce the number of orders allocated to it in the next order cycles; If the normal number of orders allocated to the printer in each order cycle is orders, then in order cycles, the number of orders allocated to the printer is reduced to orders.

[0148] In the embodiments of the present invention, by using composite technologies (including image recognition technology and natural language processing technology) to extract order features, the order requirements can be quickly understood, laying a foundation for subsequent processing. Based on the extracted features, a clustering algorithm is used to classify the orders, and the urgency level of the orders is determined by comprehensively considering multi-dimensional processing factors, ensuring that orders with a high urgency coefficient are processed first, thereby improving customer satisfaction. The status information of the cloud printer is collected through the Internet of Things technology and the sensor network, and the operating parameters of the components are monitored in real time to ensure the stable operation of the cloud printer. By using historical fault records and machine learning algorithms to predict the future fault probability of the cloud printer, measures are taken in advance to reduce the downtime caused by faults, providing a basis for the availability assessment of the cloud printer and ensuring that there are sufficient available cloud printers during high-demand periods to meet the printing requirements. For orders with a high urgency coefficient, the cloud printer with the best performance and low fault probability is preferentially matched to ensure the rapid and high-quality completion of the orders. For orders with a low urgency coefficient, the idle time and energy consumption of the cloud printer are predicted, and the allocation is optimized to maximize the energy efficiency and reduce the operating costs. By monitoring the working efficiency of the cloud printer in real time and comparing it with the order requirements, any abnormalities are immediately suspended and the orders are reallocated to ensure that the printing quality meets the standards. By adjusting the subsequent order allocation plan according to the abnormal situation, the potential risks caused by equipment failures or quality problems are reduced. In summary, the embodiments of the present invention solve the problem of poor processing effect of current cloud printing orders. In actual situations, more data and context information may be required to make specific decisions and optimization plans.

[0149] In addition, the formulas involved above are all calculated by removing the dimension and taking their numerical values, which is a formula obtained by collecting a large amount of data for software simulation to be closest to the real situation. The proportionality coefficient in the formula and each preset threshold in the analysis process are set by those skilled in the art according to the actual situation or obtained through a large amount of data simulation; the size of the proportionality coefficient is a specific value obtained by quantifying each parameter for subsequent comparison. Regarding the size of the proportionality coefficient, it depends on the amount of sample data and the corresponding processing coefficients initially set by those skilled in the art for each group of sample data; as long as the proportional relationship between the parameters and the quantified values is not affected.

[0150] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiments, since they are basically based on the method embodiments, they are described relatively simply, and the relevant parts can refer to the partial description of the method embodiments.

[0151] For the convenience of description, when describing the above device, it is divided into various units according to functions for separate description. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0152] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0153] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0154] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0156] Second: In the accompanying drawings of the disclosed embodiments of the present invention, only the structures related to the disclosed embodiments are involved. For other structures, reference can be made to the general design. Without conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0157] Finally: The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered by the protection scope of the present invention.

Claims

1. A dynamic allocation system for cloud printing order resources based on artificial intelligence algorithms, characterized in that: It includes an order analysis and clustering module, a cloud printer performance status perception module, an order matching and allocation strategy formulation module, and a real-time monitoring and dynamic adjustment module; Order analysis and clustering module: It uses composite technology to extract order features. Specifically, for the image part in the order, it uses a convolutional neural network for image decomposition to identify the complexity of the graphics, the color pattern of the image, and the features of the image resolution. For the text part, it uses a word vector model to convert it into a vector representation, including the text layout format, the semantic features of the text content, and the key text information such as the number of words. Based on the extracted features, it uses a clustering algorithm to classify the orders, determines the urgency of the orders by considering multiple-dimensional processing factors of the orders, calculates the urgency coefficient, and marks it on the clustered order groups; among them, the composite technology includes image recognition technology and natural language processing technology; According to the classification results of the order feature vectors by the clustering algorithm, further considering multiple-dimensional processing factors of the orders, it obtains the urgency coefficient through weighted calculation and marks this coefficient on the clustered order groups to reflect the urgency of the orders. Specifically, it includes: Extract the timestamp when the order is submitted and calculate the interval from submission to the current time; identify the priority identifier set by the customer in the order. Among them, the priority identifier set by the customer is divided into three levels: high, medium, and low; Combined with multi-dimensional processing factors, where the multi-dimensional processing factors of an order include the order submission time, the priority set by the customer, the order attributes, and the order size, a weighted calculation method is adopted to obtain an emergency coefficient, and the value range of the emergency coefficient is set. Specifically: Set the order submission time as , the current time is , the time interval , the priority set by the customer , the order attributes , the order size ; Considering the order submission time, the priority set by the customer, the order attributes, and the order size comprehensively, calculate the emergency coefficient , that is , where are all weights; set the value range of the emergency coefficient: 0 - 1, where 0 represents the lowest emergency level and 1 represents the highest emergency level; Obtain the urgency coefficient and mark the obtained urgency coefficient on each clustered order group; Cloud printer performance status perception module: The process of collecting cloud printer status information through Internet of Things technology and sensor network, embedding an intelligent diagnosis unit to monitor component operation parameters in real time, and using historical fault records and machine learning algorithms to predict the future fault probability of cloud printers includes: Collect historical fault records from the maintenance logs or databases of each cloud printer; Check and process missing values, outliers, or duplicate records in the data; Extract features related to fault prediction from historical fault records; Use machine learning algorithms to build a fault prediction model and complete model training; Obtain the performance vector of the current cloud printer; record the usage duration of the cloud printer and the running time after the last repair; Input the performance vector of the current cloud printer, the usage duration of the cloud printer, and the running time after the last repair into the trained fault prediction model to calculate the probability of each cloud printer having a fault in a future period of time; Use the probability value output by the fault prediction model as an important basis for evaluating the availability of cloud printers. Among them, the value range of the output probability value is 0-1, where 0 indicates no fault risk and 1 indicates a very high fault risk; Order matching and allocation strategy formulation module: For orders with a high urgency coefficient, it preferentially matches cloud printers with the best performance and low fault probability, and considers batch processing capabilities; for orders with a low urgency coefficient, it predicts the idle time and energy consumption of cloud printers and optimizes the allocation; Real-time monitoring and dynamic adjustment module: It monitors the working efficiency of cloud printers in real time and conducts comparative analysis. If anomalies are found, it reallocates orders to other cloud printers.

2. The dynamic allocation system for cloud printing order resources based on artificial intelligence algorithm according to claim 1, wherein: The process of the order analysis and clustering module using composite technology to extract order features and classifying orders based on the extracted features using a clustering algorithm includes: Adopt a combination of image recognition and natural language processing technologies to extract features from the content of each order: For the order part containing images or graphics, use a convolutional neural network to decompose the images and identify features such as the complexity of the graphics, the color mode of the images, and the resolution of the images; For the text part of the order, use a word vector model to convert the text into a vector representation, and analyze key text information such as the text layout format, the semantic features of the text content, and the number of words; Integrate various feature information obtained through image recognition and natural language processing into a complete order feature vector; Based on the extracted order features, use a clustering algorithm, with the distance between order feature vectors as the similarity measurement criterion, to perform clustering operations on the feature vectors of all orders, and classify orders with close distances into one category; Among them, orders with close distances are orders with similar features: When the multi-dimensional differences in graphic complexity, color mode, and key text information of the order feature vectors are all less than the set clustering threshold, these orders are classified into the same cluster; During the clustering process, continuously adjust the center and threshold of the clustering according to historical order data to adapt to the distribution changes of different types of orders.

3. The dynamic allocation system for cloud printing order resources based on artificial intelligence algorithm according to claim 1, characterized in that: The process of the cloud printer performance status perception module collecting cloud printer status information through Internet of Things technology and sensor networks and embedding an intelligent diagnosis unit to monitor component operation parameters in real time includes: Establish connections with each cloud printer connected to the cloud printing service through Internet of Things technology for comprehensive detection; Use sensor networks to collect the general status information of cloud printers; Embed an intelligent diagnosis unit inside the cloud printer to evaluate the health status of components, where the intelligent diagnosis unit is used to monitor the operation parameters of internal components of the cloud printer in real time; Integrate all the information obtained from the sensor network and the intelligent diagnosis unit to construct a performance vector of the cloud printer.

4. The dynamic allocation system for cloud printing order resources based on artificial intelligence algorithm according to claim 1, characterized in that: The process of the order matching and allocation strategy formulation module preferentially matching the cloud printer with the best performance and low failure probability for orders with a high emergency coefficient and considering the batch processing ability includes: Set the normal operation probability range , to satisfy ; At the same time, set the urgency level range , to satisfy , which is used to quantify the urgency level of the order; For order clustering with an emergency coefficient higher than the upper limit of the emergency level range screen out a list of cloud printers with the best performance vector and a probability value output by the fault prediction model lower than the lower limit of the normal operation probability range ; For the screened list of cloud printers, further analyze the batch processing ability of each cloud printer for the current order clustering; Among them, the batch processing ability refers to the ability of the cloud printer to process multiple orders or printing tasks per unit time.

5. The dynamic allocation system for cloud printing order resources based on artificial intelligence algorithm according to claim 4, characterized in that: The process of the order matching and allocation strategy formulation module predicting the idle time and energy consumption of cloud printers and optimizing the allocation for orders with a low emergency coefficient includes: For order clustering with an urgency coefficient lower than the lower limit of the urgency range obtain the predicted idle time information of all cloud printers; meanwhile, consider the energy consumption of each cloud printer; Based on the prediction of idle time and energy consumption, identify the cloud printer that is about to enter a long idle state and has a high current energy consumption, and then allocate the order with a low emergency coefficient to this cloud printer and adjust the working mode of the cloud printer accordingly.

6. The dynamic allocation system for cloud printing order resources based on artificial intelligence algorithm according to claim 1, characterized in that: The process of the real-time monitoring and dynamic adjustment module monitoring the working efficiency of cloud printers in real time, conducting comparative analysis, and reallocating orders to other cloud printers if abnormalities are found includes: After the order is assigned to the cloud printer and the printing process is started, the sensors on the cloud printer side monitor the working efficiency of the cloud printer in real time; Compare and analyze the monitored working efficiency of the cloud printer with the original requirements and expected quality standards of the order: If any abnormality is found in the printing process of any cloud printer during the comparison and analysis, immediately pause the printing task of this order on this cloud printer; Re-evaluate the status of all available cloud printers, re-assign the remaining unprinted part of the order to other cloud printers; at the same time, adjust the distribution plan of subsequent orders according to the abnormal situation.

Images