An intelligent analysis system for cleaning the inner wall of refrigeration equipment

Abstract

The present invention relates to the field of intelligent cleaning technology, specifically to an intelligent analysis system for cleaning the inner wall of a refrigeration device, the system comprising an intelligent image capture module, a dirt identification and segmentation module, a dirt development prediction module, and a cleaning resource allocation module. In the present invention, the cleaning of the inner wall of the refrigeration device is optimized by intelligent image capture technology, image data and node relationships are accurately calculated, the accuracy of dirt area identification and boundary clarity are improved, and an optimized network segmentation cost strategy is applied to not only improve the recognition accuracy, but also enhance the pertinence and effect of cleaning, and use refined boundary map analysis to accurately segment the clean area and the dirt area, improve the efficiency and economy of cleaning resource allocation, introduce a hidden Markov model to predict the trend of dirt changes, improve the foresight and accuracy of resource allocation, enhance operation automation, optimize personnel and material utilization through dynamic scheduling, and significantly improve energy and environmental benefits.

Description

Technical Field

[0001] The present invention relates to the field of intelligent cleaning technology, and in particular to an intelligent analysis system for cleaning the inner wall of a refrigeration device. Background Art

[0002] The field of smart cleaning technology involves the use of automated and intelligent solutions to improve the efficiency and effectiveness of the cleaning process. Key technologies in the field include machine learning, artificial intelligence, sensor technology and automatic control systems. The application of technology can optimize the operating parameters of the equipment, predict cleaning needs, and automatically perform cleaning tasks when necessary. Smart cleaning technology also emphasizes energy efficiency and reducing environmental impact, such as reducing water and energy consumption by adjusting the cleaning cycle. The system can provide real-time feedback and performance monitoring to ensure that the equipment is maintained in optimal working condition.

[0003] Among them, the intelligent analysis of the freezing equipment inner wall cleaning system is designed to maintain and clean the internal surface of the freezing equipment. Its main purpose is to keep the inner wall of the equipment clean and sanitary, prevent the growth of bacteria and mold, and ensure the efficiency of the freezing process and the quality of the product. This system is crucial to the food processing and storage industry because the industry has strict requirements on hygiene standards and product safety. The freezing equipment inner wall cleaning system reduces the need for manual cleaning through automated cleaning, improves the consistency and repeatability of cleaning, and ensures that the freezing equipment can operate in the best condition.

[0004] The existing intelligent analysis system for cleaning the inner wall of refrigeration equipment has obvious limitations in terms of cleaning efficiency and environmental impact. Although the existing system can automatically perform cleaning tasks and provide real-time feedback, it lacks accuracy in dirt identification and segmentation, resulting in incomplete cleaning or excessive use of cleaning resources, which affects the long-term operating efficiency and maintenance costs of the equipment. The lack of an efficient prediction model makes it impossible to accurately predict the development trend of dirt. This deficiency is particularly prominent in the food processing and storage industry, which requires high hygiene standards. Due to insufficient optimization of resource allocation and scheduling, the existing technology wastes personnel and material utilization, which in turn affects the economic benefits and environmental sustainability of the entire system. Insufficient technology and operation result in the execution of cleaning tasks that are not economical or environmentally friendly. Summary of the invention

[0005] The present application provides a system for cleaning the inner wall of refrigeration equipment with intelligent analysis. The existing system for cleaning the inner wall of refrigeration equipment with intelligent analysis has obvious limitations in terms of cleaning efficiency and environmental impact. Although the existing system can automatically perform cleaning tasks and provide real-time feedback, it is insufficient in the accuracy of dirt identification and segmentation, resulting in incomplete cleaning or excessive use of cleaning resources, thus affecting the long-term operating efficiency and maintenance costs of the equipment. The lack of an efficient prediction model makes it impossible to accurately predict the development trend of dirt. This deficiency is particularly prominent in the food processing and storage industry, which requires high hygiene standards. Due to insufficient optimization of resource allocation and scheduling, the existing technology has waste in the utilization of personnel and materials, which in turn affects the economic benefits and environmental sustainability of the entire system. Insufficient technology and operation lead to problems in which the execution of cleaning tasks is not economical or environmentally friendly.

[0006] In view of the above problems, the present application provides a refrigeration equipment inner wall cleaning system with intelligent analysis.

[0007] The present application provides a system for cleaning the inner wall of a refrigeration equipment with intelligent analysis, wherein the system comprises:

[0008] The intelligent image capture module captures images of the inner wall of the refrigeration equipment, converts image data into pixel nodes, calculates weights between nodes based on color differences and distances, generates a node relationship graph, runs edge weight calculations based on the node relationship graph, extracts the edge weight set in the graph, and obtains initial edge weight data.

[0009] The dirt identification and segmentation module sets the energy function based on the initial edge weight data, iteratively adjusts the node classification and reduces the energy cost, generates a refined boundary graph, and based on the refined boundary graph, applies the optimized network segmentation cost strategy to identify and segment the inner wall clean area and the dirt area to obtain the segmentation marking result.

[0010] The fouling development prediction module summarizes the development data of differentiated fouling types based on the segmentation and labeling results, initializes the hidden Markov model, inputs equipment operation and environmental parameters, generates state transition data, runs a prediction algorithm based on the state transition data to predict fouling changes in future time periods, and obtains a fouling development trend graph.

[0011] The cleaning resource configuration module calculates the cleaning resource demand based on the dirt development trend diagram, adjusts the cleaning staff scheduling and cleaning agent distribution, generates a resource allocation plan, implements dynamic scheduling based on the resource allocation plan, optimizes the utilization of personnel and materials, and obtains an optimized scheduling plan.

[0012] Preferably, the step of obtaining the initial edge weight data is specifically as follows:

[0013] Based on the image capture of the inner wall of the freezing equipment, the image data is converted into a set of pixel nodes through pixel recognition technology, using the formula:

[0014]

[0015] Convert each image point into an independent pixel node to generate an image node set, where represents a pixel node, Represents the image at coordinates Col is a simplified function name for extracting pixel color information.

[0016] According to the image node set, the weight between nodes is calculated by the color difference and spatial distance of each node, and the weight calculation formula is used:

[0017]

[0018] Calculate the weight between node i and node j to generate a set of inter-node weights, where: is the weight between node i and node j, and are the color values ​​of nodes i and j, are the coordinates of the node, and is the weight adjustment parameter;

[0019] Utilizing the inter-node weight set, summarizing the input and output weights of each node, sorting the weight of each node, determining the priority of each node, adjusting the position of the node in the graph according to the priority, and generating a node relationship graph;

[0020] Extract the edge weights from the node relationship graph using the formula:

[0021]

[0022] Summarize the weights of edges in the graph , generate initial edge weight data, where represents the edge weight set, Represents a node relationship graph.

[0023] Preferably, the step of obtaining the refined boundary map is specifically as follows:

[0024] Based on the initial edge weight data, the weight sum between nodes is defined as the energy function, using the formula:

[0025]

[0026] Generate energy states where Indicates the energy state, Indicates the total number of nodes, represents the weight between node i and node j, is the normalization coefficient;

[0027] The energy state is used to adjust the node classification, and the accuracy of the classification effect is optimized by referring to the adjustment intensity and the energy adjustment amount of each node, through the formula:

[0028]

[0029] Get the adjusted energy, where is the adjusted energy, is the energy value of the previous iteration, is the energy adjustment of node i, To adjust the intensity, is the attenuation factor;

[0030] Based on the adjusted energy, after the energy optimization reaches the threshold, the formula is used:

[0031]

[0032] Generate a refined boundary map, where To improve the clarity of the boundary map, To adjust the parameters of the boundary sharpness, is the adjusted energy.

[0033] Preferably, the step of obtaining the segmentation labeling result is specifically as follows:

[0034] Using the refined boundary map, the pixel density of each region is calculated using the formula:

[0035]

[0036] Adjust the calculated bias of low-density areas to generate a pixel density matrix, where represents the average pixel density of the region, Represents the total number of pixels in the area. represents the intensity value of the i-th pixel, is the normalization coefficient, is the stabilization parameter;

[0037] Applying the pixel density matrix, density thresholds are calculated and dirt areas are identified using the formula:

[0038]

[0039] Get the dirt area identification, where represents the density threshold for distinguishing dirty and clean areas, represent The mean of represent The standard deviation of represents the adjustment factor, is the threshold adjustment factor;

[0040] Using the dirty region identification, a network segmentation cost strategy is applied using the formula:

[0041]

[0042] Generate segmentation labeling results, where Represents the confidence of the segmentation and labeling results of the clean area and the dirty area, is the scaling factor, Represents the dirt area logo, Represents the cost function.

[0043] Preferably, the step of acquiring the state transition data is specifically as follows:

[0044] Based on the segmentation and marking results, the frequency of dirt development of differentiated types is counted using the formula:

[0045]

[0046] Generate a frequency vector where Represents the average development trend indicator after taking into account the frequency of fouling development and regularization and adjustment factors, represents the sample size, represents the sample development frequency, represents the regularization parameter, represents the adjustment factor;

[0047] Using the frequency vector, the hidden Markov model is initialized to simulate nonlinear state changes using the formula:

[0048]

[0049] Get the state transfer matrix, where A represents the value of the state transfer matrix under given fouling frequency and environmental parameters, and is the state sensitivity adjustment parameter, is the smoothing offset, Represents the average development trend indicator after taking into account the frequency of fouling development and regularization and adjustment factors;

[0050] Based on the state transfer matrix, combined with equipment operation and environmental parameters, the state transfer data is calculated through the state transfer matrix, using the formula:

[0051]

[0052] in, Represents state transfer data, sin and cos are trigonometric function transformations of environmental parameters. and is the adjustment factor, is the normalization parameter, Represents the value of the state transfer matrix under given fouling frequency and environmental parameters, Represents environmental parameters.

[0053] Preferably, the steps for obtaining the dirt development trend graph are specifically as follows:

[0054] Based on the state transition data, the initial fouling data set is analyzed, and the initial fouling distribution is predicted by the exponential decay model using the formula:

[0055]

[0056] Generate initial dirt distribution, where represents the initial dirt distribution, is the time decay factor, is the stability constant, Represents state transfer data, represents the initial dirt dataset, and Represents the sensitivity adjustment parameter;

[0057] Using the initial fouling distribution, the fouling development is predicted through nonlinear transformation using the formula:

[0058]

[0059] Simulate the change of fouling under different conditions and obtain the predicted fouling distribution, where For the predicted fouling distribution, is the current dirt distribution, represents the nonlinear adjustment factor, represents the adjustment constant, Represents state transfer data;

[0060] Based on the predicted dirt distribution, the dirt distribution data at a time point is summarized, and the influence of the data at multiple time points is adjusted. The formula used is:

[0061]

[0062] Generate a fouling trend graph where: For the dirt development trend chart, represents the dirt distribution at a point in time, represents the weighting factor, Represents the total number of time points.

[0063] Preferably, the step of obtaining the resource allocation scheme is specifically as follows:

[0064] Based on the dirt development trend diagram, the cleaning demand is estimated using a nonlinear response function and calculated using a resource demand model using the formula:

[0065]

[0066] The predicted resource demand is obtained, where represents the predicted resource demand, , and is a dynamic coefficient adjusted based on historical data, and is the baseline demand adjustment parameter, Represents the dirt development trend graph;

[0067] According to the predicted resource demand, a resource allocation algorithm is used to optimize the allocation of manpower and materials, using the formula:

[0068]

[0069] Generate a human and material allocation plan, including: To provide human and material allocation plan, It is the amount of resource buffer reserved for dealing with emergencies. Indicates the cleaning area that can be covered by unit resources. Represents the predicted resource demand;

[0070] Based on the human resources and material allocation plan, the calculated data is integrated to form a resource allocation execution plan using the formula:

[0071]

[0072] Generate a resource allocation plan, where Represents a resource allocation plan, Indicates the number of time points within the planning cycle. Represents the resource allocation plan at each time point, is the time decay factor.

[0073] Preferably, the step of obtaining the optimized scheduling plan is specifically as follows:

[0074] Based on the resource allocation scheme, the real-time utilization efficiency of resources is calculated using the formula:

[0075]

[0076] The dynamic adjustment factor introduced strengthens the adaptability of the model to differentiated conditions and calculates the weighted average efficiency, where: represents the weighted average efficiency, Indicates The amount of resources allocated, represents the efficiency weight of resources, and is the adjustment factor for each resource;

[0077] Using the weighted average efficiency, a dynamic scheduling strategy is formulated using the formula:

[0078]

[0079] Introduce adjustment factors to enhance the ability to respond to resource changes and generate scheduling strategies, among which, is the scheduling strategy, represents the weighted average efficiency, is the weight factor, and is a newly introduced regulatory factor;

[0080] Based on the scheduling strategy, resources and the scheduling strategy are integrated and the following formula is adopted:

[0081]

[0082] Strengthen the model's sensitivity to time and generate an optimized scheduling plan, where To optimize the scheduling plan, For time The scheduling strategy, is the adjusted time decay factor.

[0083] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0084] Through the innovative application of intelligent image capture technology, the cleaning efficiency of the inner wall of the refrigeration equipment is optimized. The conversion of image data and the precise calculation of node relationships make the identification of dirt areas more accurate and the boundaries clearer. The use of optimized network segmentation cost strategies not only improves the accuracy of recognition, but also enhances the pertinence and effectiveness of the cleaning process. Through refined boundary map analysis, the clean area and the dirt area are more accurately segmented, thereby improving the allocation efficiency and economic efficiency of cleaning resources. The prediction model of dirt development, especially the introduction of hidden Markov models, effectively predicts the changing trend of dirt, making resource allocation more forward-looking and accurate, and improving the level of automation of operations. It also optimizes the utilization of personnel and materials through dynamic scheduling, significantly improving the overall energy and environmental benefits.

[0085] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 The present invention proposes an overall flow chart of a refrigeration equipment inner wall cleaning system with intelligent analysis;

[0087] Figure 2 A flow chart of initial edge weight data of a refrigeration equipment inner wall cleaning system for intelligent analysis is proposed for the present invention;

[0088] Figure 3 A flow chart of a refined boundary map of a refrigeration equipment inner wall cleaning system for intelligent analysis is proposed for the present invention;

[0089] Figure 4 A flow chart of segmentation and labeling results of a refrigeration equipment inner wall cleaning system for intelligent analysis is proposed for the present invention;

[0090] Figure 5 A flow chart of state transition data of a refrigeration equipment inner wall cleaning system for intelligent analysis is proposed for the present invention;

[0091] Figure 6 A flow chart of a dirt development trend diagram of a refrigeration equipment inner wall cleaning system for intelligent analysis is proposed for the present invention;

[0092] Figure 7 A flow chart of a resource allocation scheme for a refrigeration equipment inner wall cleaning system with intelligent analysis is proposed for the present invention;

[0093] Figure 8 The present invention proposes a flow chart of an optimized scheduling plan for a refrigeration equipment inner wall cleaning system with intelligent analysis. DETAILED DESCRIPTION

[0094] The present application provides a refrigeration equipment inner wall cleaning system with intelligent analysis.

[0095] Application Overview

[0096] The existing intelligent analysis system for cleaning the inner wall of refrigeration equipment has obvious limitations in terms of cleaning efficiency and environmental impact. Although the existing system can automatically perform cleaning tasks and provide real-time feedback, it lacks accuracy in dirt identification and segmentation, resulting in incomplete cleaning or excessive use of cleaning resources, which affects the long-term operating efficiency and maintenance costs of the equipment. The lack of an efficient prediction model makes it impossible to accurately predict the development trend of dirt. This deficiency is particularly prominent in the food processing and storage industry, which requires high hygiene standards. Due to insufficient optimization of resource allocation and scheduling, the existing technology wastes personnel and material utilization, which in turn affects the economic benefits and environmental sustainability of the entire system. Insufficient technology and operation result in the execution of cleaning tasks that are not economical or environmentally friendly.

[0097] In response to the above technical problems, the overall idea of ​​the technical solution provided by this application is as follows: Embodiment 1

[0098] like Figure 1 As shown, the present application provides a system for cleaning the inner wall of a refrigeration equipment with intelligent analysis, wherein the system comprises:

[0099] The intelligent image capture module captures images of the inner wall of the refrigeration equipment, converts image data into pixel nodes, calculates weights between nodes based on color differences and distances, generates a node relationship graph, runs edge weight calculations based on the node relationship graph, extracts the edge weight set in the graph, and obtains initial edge weight data.

[0100] The dirt identification and segmentation module sets the energy function based on the initial edge weight data, iteratively adjusts the node classification and reduces the energy cost, generates a refined boundary graph, and based on the refined boundary graph, applies the optimized network segmentation cost strategy to identify and segment the clean area and the dirt area of ​​the inner wall to obtain the segmentation marking result.

[0101] The fouling development prediction module summarizes the development data of differentiated fouling types based on the segmentation and labeling results, initializes the hidden Markov model, inputs equipment operation and environmental parameters, generates state transition data, runs the prediction algorithm based on the state transition data, and predicts the fouling changes in the future time period to obtain a fouling development trend chart.

[0102] The cleaning resource allocation module calculates the cleaning resource demand based on the dirt development trend chart, adjusts the cleaning staff scheduling and cleaning agent allocation, generates a resource allocation plan, implements dynamic scheduling based on the resource allocation plan, optimizes the utilization of personnel and materials, and obtains an optimized scheduling plan.

[0103] The node relationship diagram includes the node weight type and the characteristics of the associated data. The initial edge weight data includes the weight data set and the weight distribution characteristics. The refined boundary diagram includes the refined classification boundaries and the boundary clarity. The segmentation marking results include the area division identification and the area attribute marking. The state transition data includes the state change probability and the model parameters. The dirt development trend diagram includes the change prediction period and the trend judgment basis. The resource allocation plan includes the resource type and the resource quantity. The optimized scheduling plan includes the scheduling efficiency and the scheduling strategy.

[0104] Specifically, if Figure 2 As shown in Figure 2, the steps for obtaining the initial edge weight data are as follows:

[0105] Based on the image capture of the inner wall of the freezing equipment, the image data is converted into a set of pixel nodes through pixel recognition technology, using the formula:

[0106]

[0107] Convert each image point into an independent pixel node to generate an image node set, where represents a pixel node, Represents the image at coordinates Col is a simplified function name for extracting pixel color information.

[0108] According to the image node set, the weight between nodes is calculated by the color difference and spatial distance of each node, and the weight calculation formula is used:

[0109]

[0110] Calculate the weight between node i and node j to generate a set of inter-node weights, where: is the weight between node i and node j, and are the color values ​​of nodes i and j, are the coordinates of the node, and is the weight adjustment parameter;

[0111] Using the weight set between nodes, summarize the input and output weights of each node, sort the weight of each node, determine the priority of each node, adjust the position of the node in the graph according to the priority, and generate a node relationship graph;

[0112] Extract the edge weights from the node relationship graph using the formula:

[0113]

[0114] Summarize the weights of edges in the graph , generate initial edge weight data, where represents the edge weight set, Represents a node relationship graph.

[0115] Formula derivation of pixel nodes:

[0116]

[0117] : The image is located at coordinates Pixel values, including RGB color values;

[0118] Col: is a function used to Extract color information or process color data;

[0119] Suppose there is an image pixel at coordinates , the RGB color value of this point is , we will show how to extract color information from this RGB value;

[0120] The image processing function Col converts the RGB color into a single grayscale value or performs a form of processing, which is simplified here to calculate the average value:

[0121]

[0122] Calculated It is a grayscale value converted from the original RGB color value. This grayscale value can be used in subsequent image processing steps, such as distinguishing different color intensities when creating a node relationship graph of an image.

[0123] The formula for the weight between node i and node j is derived as follows:

[0124]

[0125] : The color values ​​of nodes i and j are extracted from the image data based on the RGB model;

[0126] : The coordinate position of nodes i and j, the coordinates represent the position of each pixel on the image;

[0127] and : Adjust parameters to control the impact of color difference and spatial distance on weights. The parameters are obtained through experiments and optimization;

[0128] Assume there are two pixel nodes, their positions are and , the color values ​​are and ,parameter and Set them to 1.5 and 0.5, and calculate the weight between these two nodes;

[0129] Calculate the square of the color difference: ;

[0130] Calculate the sum of squared spatial distances: ;

[0131] Applying tuning parameters : (Sum of squared spatial distances) ;

[0132] Calculate the total denominator: 100+250=350;

[0133] Apply the formula to calculate the weight: ;

[0134] Calculated weight It represents the connection strength between nodes i and j. A smaller weight value indicates that the similarity between the two nodes is low (larger color difference and farther distance). This weight value can be used to construct a node relationship graph of the image and identify different areas in image segmentation or pattern recognition.

[0135] The formula for edge weight set is derived as follows:

[0136]

[0137] : The weight between nodes i and j;

[0138] : Represents the node relationship of the entire graph, including the set of all node pairs.

[0139] Computational-based Weight, Hypothesis Graph Contains three pairs of node relationships: (1, 2), (1, 3) and (2, 3), assuming the corresponding weights of the nodes It is calculated that for ;

[0140] gather Will include all weights:

[0141] gather It is a picture The weight set of all edges in the image is used for further analysis of the image, such as structural segmentation of the image, or to evaluate the similarity and connection strength between nodes in image recognition algorithms.

[0142] Specifically, if Figure 3As shown, the steps for obtaining the refined boundary map are specifically as follows:

[0143] Based on the initial edge weight data, the weight sum between nodes is defined as the energy function, using the formula:

[0144]

[0145] Generate energy states where Indicates the energy state, Indicates the total number of nodes, represents the weight between node i and node j, is the normalization coefficient;

[0146] Use the energy state to adjust the node classification, refer to the adjustment intensity and the energy adjustment amount of each node, and optimize the accuracy of the classification effect. The formula is:

[0147]

[0148] Get the adjusted energy, where is the adjusted energy, is the energy value of the previous iteration, is the energy adjustment of node i, To adjust the intensity, is the attenuation factor;

[0149] Based on the adjusted energy, after the energy optimization reaches the threshold, use the formula:

[0150]

[0151] Generate a refined boundary map, where To improve the clarity of the boundary map, To adjust the parameters of the boundary sharpness, is the adjusted energy.

[0152] The formula and its derivation:

[0153]

[0154] : The weight between node i and node j, assuming that a set of data is given as: ;

[0155] : Normalization coefficient, used to adjust the scale of edge weights, assuming ;

[0156] : The total number of nodes. Here we consider a simple three-node graph, so ;

[0157] Calculate the normalized square of each edge weight:

[0158]

[0159] Then calculate the energy state :

[0160]

[0161] This energy value It represents the overall energy state of the entire graph under normalized edge weight conditions.

[0162] The formula for the adjusted energy and its derivation:

[0163]

[0164] : The energy value of the previous iteration, assumed to be 1.5;

[0165] : The energy adjustment amount of each node. Assume that the node energy adjustment amount is ;

[0166] : Adjust the strength, assuming it is 2;

[0167] : Attenuation factor, used to balance the strength of the adjustment, assumed to be 3.

[0168] Calculate adjustments:

[0169] Then calculate the new energy value :

[0170]

[0171] This new energy value It represents the energy state of the node after adjustment.

[0172] The formula for refining the boundary map and its derivation:

[0173]

[0174] : The adjusted energy, i.e. 1.45167;

[0175] : Threshold adjustment parameter, assumed to be 1;

[0176] Compute the sharpness of the boundary map:

[0177]

[0178] This value Indicates the sharpness of the resulting refined boundary map.

[0179] Specifically, if Figure 4 As shown, the steps for obtaining the segmentation labeling results are as follows:

[0180] Using the refined boundary map, calculate the pixel density of each area using the formula:

[0181]

[0182] Adjust the calculated bias of low-density areas to generate a pixel density matrix, where represents the average pixel density of the region, Represents the total number of pixels in the area. represents the intensity value of the i-th pixel, is the normalization coefficient, is the stabilization parameter;

[0183] Apply the pixel density matrix, calculate the density threshold and identify the dirt area, using the formula:

[0184]

[0185] Get the dirt area identification, where represents the density threshold for distinguishing dirty and clean areas, represent The mean of represent The standard deviation of represents the adjustment factor, is the threshold adjustment factor;

[0186] Using the dirty region identification, a network segmentation cost strategy is applied, using the formula:

[0187]

[0188] Generate segmentation labeling results, where Represents the confidence of the segmentation and labeling results of the clean area and the dirty area, is the scaling factor, Represents the dirt area logo, Represents the cost function.

[0189] The formula for the pixel density matrix is:

[0190]

[0191] : The total number of pixels in the area, assumed to be 100;

[0192] : The intensity value of the i-th pixel, which is assumed to be an integer from 1 to 100;

[0193] : Normalization coefficient, assumed to be 10;

[0194] : Stabilization parameter, used to adjust the calculation deviation of low-density areas, assumed to be 5,

[0195] Calculate the normalized sum of squared pixel intensities: For each pixel, its intensity Divide by , then square;

[0196] Add all normalized square sums and add the stabilization parameter ;

[0197] Taking the square root of the above results gives the pixel density matrix ;

[0198] Suppose there are five pixels with intensity values ​​of 10, 20, 30, 40, and 50. Then:

[0199]

[0200] This result Represents the calculated average pixel density of the region.

[0201] Formula for dirt area identification:

[0202]

[0203] The mean of

[0204] Standard Deviation

[0205] : Adjustment factor, assumed to be 0.5;

[0206] : Threshold adjustment factor, assumed to be 0.1;

[0207] calculate , divide the result by ;

[0208] Assumptions (calculated from a set of pixel densities), then:

[0209]

[0210] This result Represents the density threshold used to distinguish dirty and clean areas.

[0211] The formula for segmentation labeling results is:

[0212]

[0213] : Dirt area identification;

[0214] : Cost function, representing the conversion cost, assumed to be 2;

[0215] : Scaling factor, assumed to be 1.5;

[0216] calculate ;

[0217] Applying the exponential function and fractional conversion yields ;

[0218] Assumptions (calculated from step 2), then:

[0219]

[0220] This result Represents the confidence of the segmentation and labeling results of the clean area and the dirty area.

[0221] Specifically, if Figure 5 As shown, the steps for obtaining state transition data are as follows:

[0222] Based on the segmentation and marking results, the frequency of dirt development of differentiated types is counted using the formula:

[0223]

[0224] Generate a frequency vector where Average development trend indicator taking into account the frequency of fouling development and regularization and adjustment factors, represents the sample size, represents the sample development frequency, represents the regularization parameter, represents the adjustment factor;

[0225] Use the frequency vector to initialize the hidden Markov model and simulate nonlinear state changes using the formula:

[0226]

[0227] Get the state transfer matrix, where A represents the value of the state transfer matrix under given fouling frequency and environmental parameters, and is the state sensitivity adjustment parameter, is the smoothing offset, Average development trend indicator taking into account the frequency of fouling development and regularization and adjustment factors;

[0228] Based on the state transfer matrix, combined with equipment operation and environmental parameters, the state transfer data is calculated through the state transfer matrix, using the formula:

[0229]

[0230] in, Represents state transfer data, sin and cos are trigonometric function transformations of environmental parameters. and is the adjustment factor, is the normalization parameter, Represents the value of the state transfer matrix under given fouling frequency and environmental parameters, Represents environmental parameters.

[0231] The calculation formula is:

[0232]

[0233] : Sample size, assumed to be 10;

[0234] : The development frequency of the i-th sample, the assumed values ​​are 1, 2, 3, ..., 10;

[0235] : Regularization parameter, assumed to be 0.5, to avoid the denominator being zero;

[0236] : Adjustment factor, assumed to be 2, used to smooth the denominator;

[0237] Compute the sum of the cubed development frequencies for each sample:

[0238]

[0239] Compute the regularization term and the adjustment factor:

[0240]

[0241]

[0242] Insert formula calculation :

[0243]

[0244] , which represents the average development trend indicator after taking into account the frequency of fouling development and the regularization and adjustment factors.

[0245] The calculation formula of A is:

[0246]

[0247] : The frequency of fouling development and the average development trend index after regularization and adjustment factors, 216.12;

[0248] : State sensitivity adjustment parameter, set to 0.1;

[0249] : Enhanced state sensitivity, set to 5;

[0250] : Smoothing offset, set to 0.01;

[0251] Compute the exponential term:

[0252]

[0253] Compute the logarithmic term:

[0254]

[0255] Example results: , represents the value of the state transfer matrix under given fouling frequency and environmental parameters, reflecting the sensitivity of state change.

[0256] State transition data The calculation formula is:

[0257]

[0258] : The value of the state transfer matrix under given fouling frequency and environmental parameters, 1.61;

[0259] : Environmental parameters, assumed to be 30;

[0260] : Adjustment factor, set to 1;

[0261] : Adjustment factor, set to 2;

[0262] : Normalization parameter, set to 10;

[0263] Compute trigonometric transformations:

[0264]

[0265] Insert formula calculation :

[0266]

[0267] Example results: , represents the state transition data under given environmental parameters and is used to predict future state changes of the system.

[0268] Specifically, if Figure 6 As shown, the steps for obtaining the dirt development trend diagram are as follows:

[0269] Based on the state transition data, the initial fouling data set is analyzed and the initial fouling distribution is predicted by the exponential decay model using the formula:

[0270]

[0271] Generate initial dirt distribution, where represents the initial dirt distribution, is the time decay factor, is the stability constant, Represents state transfer data, represents the initial dirt dataset, and Represents the sensitivity adjustment parameter;

[0272] Using the initial fouling distribution, the fouling development is predicted through nonlinear transformation, using the formula:

[0273]

[0274] Simulate the change of fouling under different conditions and obtain the predicted fouling distribution, where For the predicted fouling distribution, is the current dirt distribution, represents the nonlinear adjustment factor, represents the adjustment constant, Represents state transfer data;

[0275] Based on the predicted dirt distribution, the dirt distribution data at a certain time point is summarized and the influence of data at multiple time points is adjusted. The formula used is:

[0276]

[0277] Generate a fouling trend graph where: For the dirt development trend chart, represents the dirt distribution at a point in time, represents the weighting factor, Represents the total number of time points.

[0278] Derivation of the formula for initial dirt distribution:

[0279]

[0280] : Initial dirt distribution, assumed to be 1000 units;

[0281] : Time decay factor, assumed to be 0.5;

[0282] : State transfer data, assumed to be 2;

[0283] : Sensitivity adjustment parameter, assumed to be 0.1;

[0284] : Model sensitivity adjustment factor, assumed to be 1;

[0285] : stability constant, assumed to be 0.01.

[0286] Calculate the denominator of the attenuation term: ;

[0287] Calculate the attenuation term: ;

[0288] Calculate the overall attenuation coefficient: ;

[0289] Calculate the final dirt distribution: ;

[0290] The value of is about 333.58, which represents a reduction of about one-third of the predicted initial fouling distribution under the given initial fouling distribution and regime transition parameters.

[0291] Derivation of the formula for predicted fouling distribution:

[0292]

[0293] : Current dirt distribution, inherited from step 1, i.e. 333.58;

[0294] : Nonlinear adjustment factor, assumed to be 4;

[0295] : State transition data, again assumed to be 2 (inherited from step 1);

[0296] : Model adjustment constant, assumed to be 0.05.

[0297] Calculation adjustment part: ;

[0298] Compute the inverse function of an exponential: ;

[0299] Calculate the denominator: 1 + 0.000273 = 1.000273;

[0300] Calculate the square root of the nonlinear adjustment factor: ;

[0301] Calculate the final dirt distribution:

[0302] The value of is about 667.07, which shows the changing trend of dirt distribution under this model parameter.

[0303] Derivation of the formula for the dirt development trend chart:

[0304]

[0305] Assume there are 5 points in time for the distribution of dirt ;

[0306] : weighting factor, assumed to be 0.9;

[0307] : Total number of time points, assumed to be 5.

[0308] Assume that each Both are 667.07 (simplified calculation);

[0309] Calculate the weighted value at each time point: (for each i from 1 to 5);

[0310] Add up all weighted values: ;

[0311] The value of is about 4626.19, which represents a comprehensive assessment of the fouling development trend under the given weighting and time parameters.

[0312] Specifically, if Figure 7 As shown, the steps for obtaining the resource allocation plan are specifically as follows:

[0313] Based on the dirt development trend chart, the cleaning demand is estimated using a nonlinear response function and calculated using the resource demand model using the formula:

[0314]

[0315] The predicted resource demand is obtained, where represents the predicted resource demand, , and is a dynamic coefficient adjusted according to historical data, and is the baseline demand adjustment parameter, Represents the dirt development trend graph;

[0316] According to the predicted resource demand, the resource allocation algorithm is used to optimize the allocation of manpower and materials. The formula used is:

[0317]

[0318] Generate a human and material allocation plan, including: To provide human and material allocation plan, It is the amount of resource buffer reserved for dealing with emergencies. Indicates the cleaning area that can be covered by unit resources. Represents the predicted resource demand;

[0319] Based on the human resources and material allocation plan, the calculated data is integrated to form a resource allocation execution plan using the formula:

[0320]

[0321] Generate a resource allocation plan, where Represents a resource allocation plan, Indicates the number of time points within the planning cycle. Represents the resource allocation plan at each time point, is the time decay factor.

[0322] The formula for predicting resource demand is:

[0323]

[0324] : Dirt development trend;

[0325] , , : Dynamic coefficient adjusted according to historical data;

[0326] : Baseline demand adjustment parameters;

[0327] : Forecasted resource demand;

[0328] Data acquisition: Assume that through historical data analysis, we get (fouling index), which is an estimate of the average fouling development over a certain period in the future;

[0329] Parameter setting: Assumptions ,The parameters are determined through historical data and regression analysis, reflecting the relationship between resource demand and fouling development;

[0330] Formula calculation: insert numerical calculation :

[0331]

[0332]

[0333] This represents a requirement of 145 units of cleaning resources in a given period.

[0334] Formula for manpower and material allocation plan:

[0335]

[0336] : Resource demand;

[0337] : Resource buffer amount;

[0338] : Unit resource efficiency;

[0339] :Human resources and material allocation plan;

[0340] Parameter setting: Setting and ,The values ​​reflect the reserved resources and the cleaning effect per unit of resources.

[0341] Calculation formula:

[0342]

[0343] This means that approximately 3.54 units of resources need to be adjusted to meet cleaning needs.

[0344] Resource allocation plan formula:

[0345]

[0346] : Resource allocation at each time point;

[0347] : attenuation factor;

[0348] : time points;

[0349] A: Resource allocation plan;

[0350] Parameter setting: Setting , the resource allocation at each time point is the same, ;

[0351] Calculation formula:

[0352]

[0353]

[0354] This means that taking resource decay into account, a total of approximately 8.71 units of resources are required to meet demand over the entire planning period.

[0355] Specifically, if Figure 8 As shown in the figure, the steps for obtaining the optimized scheduling plan are as follows:

[0356] Based on the resource allocation plan, calculate the real-time utilization efficiency of resources using the formula:

[0357]

[0358] The dynamic adjustment factor introduced strengthens the adaptability of the model to differentiated conditions and calculates the weighted average efficiency, where: represents the weighted average efficiency, Indicates The amount of resources allocated, represents the efficiency weight of resources, and is the adjustment factor for each resource;

[0359] Using weighted average efficiency, a dynamic scheduling strategy is formulated using the formula:

[0360]

[0361] Introduce adjustment factors to enhance the ability to respond to resource changes and generate scheduling strategies, among which, is the scheduling strategy, represents the weighted average efficiency, is the weight factor, and is a newly introduced regulatory factor;

[0362] Based on the scheduling strategy, resources and scheduling strategies are integrated and the following formula is used:

[0363]

[0364] Strengthen the model's sensitivity to time and generate an optimized scheduling plan, where To optimize the scheduling plan, For time The scheduling strategy, is the adjusted time decay factor.

[0365] Weighted average efficiency formula:

[0366]

[0367] : No. The amount of resources allocated;

[0368] : No. The efficiency weight of each resource;

[0369] : For each resource The dynamic adjustment factors introduced;

[0370] Assume there are three resources with the following parameters:

[0371] , the amount of resource allocation;

[0372] , efficiency weight;

[0373] , resource efficiency adjustment factor;

[0374] , adjustment factor of the denominator;

[0375] Calculate weighted average efficiency :

[0376]

[0377]

[0378]

[0379]

[0380] The weighted average efficiency It is approximately 160.53, which indicates the overall resource efficiency under the adjusted resource allocation and efficiency weights.

[0381] Scheduling strategy formula:

[0382]

[0383] (weight factor);

[0384] (Adjustment Factor);

[0385] The calculation scheduling strategy uses the previous step :

[0386]

[0387]

[0388]

[0389] because is very large, the denominator approaches infinity, making It is close to 0, which indicates that there is almost no room for resource scheduling strategies to start under the given parameters.

[0390] Optimized scheduling formula:

[0391]

[0392] (the upper limit of the calculation time);

[0393] because Close to 0, it can be considered will also be close to 0, and no further integration calculation is required.

[0394] Final dispatch plan It is almost zero, which means that under this resource configuration and scheduling strategy, the actual execution effect is very limited, and it is necessary to reconsider the setting of resource allocation and efficiency weights.

[0395] Although the present application has been described in conjunction with specific features and embodiments thereof, it is obvious that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely exemplary illustrations of the present application as defined by the appended claims, and are deemed to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include these modifications and variations.

Claims

1. An intelligent analysis system for cleaning the inner wall of a refrigeration equipment, characterized in that: The system comprises: The intelligent image capture module is based on the image capture of the inner wall of the freezing equipment, converts the image data into pixel nodes, calculates the weights between the nodes according to the color difference and distance, generates a node relationship graph, runs the edge weight calculation based on the node relationship graph, extracts the edge weight set in the graph, and obtains the initial edge weight data; The dirt identification and segmentation module sets an energy function based on the initial edge weight data, iteratively adjusts node classification and reduces energy cost, generates a refined boundary graph, and based on the refined boundary graph, applies an optimized network segmentation cost strategy to identify and segment the inner wall clean area and the dirt area to obtain a segmentation labeling result; The fouling development prediction module summarizes the development data of differentiated fouling types based on the segmentation and labeling results, initializes a hidden Markov model, inputs equipment operation and environmental parameters, generates state transition data, runs a prediction algorithm based on the state transition data to predict fouling changes in a future time period, and obtains a fouling development trend diagram; The cleaning resource configuration module calculates the cleaning resource demand based on the dirt development trend diagram, adjusts the cleaning staff scheduling and cleaning agent distribution, generates a resource allocation plan, implements dynamic scheduling based on the resource allocation plan, optimizes the utilization of personnel and materials, and obtains an optimized scheduling plan.

2. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 1 is characterized in that: The steps for obtaining the initial edge weight data are specifically as follows: Based on the image capture of the inner wall of the freezing equipment, the image data is converted into a set of pixel nodes through pixel recognition technology, using the formula: ； Convert each image point into an independent pixel node to generate an image node set, where represents a pixel node, Represents the image at coordinates Col is a simplified function name for extracting pixel color information. According to the image node set, the weight between nodes is calculated by the color difference and spatial distance of each node, and the weight calculation formula is used: ； Calculate the weight between node i and node j to generate a set of inter-node weights, where: is the weight between node i and node j, and are the color values ​​of nodes i and j, are the coordinates of the node, and is the weight adjustment parameter; Utilizing the inter-node weight set, summarizing the input and output weights of each node, sorting the weight of each node, determining the priority of each node, adjusting the position of the node in the graph according to the priority, and generating a node relationship graph; Extract the edge weights from the node relationship graph using the formula: ； Summarize the weights of edges in the graph , generate initial edge weight data, where represents the edge weight set, Represents a node relationship graph.

3. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 2 is characterized in that: The steps of obtaining the refined boundary map are specifically as follows: Based on the initial edge weight data, the weight sum between nodes is defined as the energy function, using the formula: ； Generate energy states where Indicates the energy state, Indicates the total number of nodes, represents the weight between node i and node j, is the normalization coefficient; The energy state is used to adjust the node classification, and the accuracy of the classification effect is optimized by referring to the adjustment intensity and the energy adjustment amount of each node, through the formula: ； Get the adjusted energy, where is the adjusted energy, is the energy value of the previous iteration, is the energy adjustment of node i, To adjust the intensity, is the attenuation factor; Based on the adjusted energy, after the energy optimization reaches the threshold, the formula is used: ； Generate a refined boundary map, where To improve the clarity of the boundary map, To adjust the parameters of the boundary sharpness, is the adjusted energy.

4. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 3 is characterized in that: The steps for obtaining the segmentation labeling result are specifically as follows: Using the refined boundary map, the pixel density of each region is calculated using the formula: ； Adjust the calculated bias of low-density areas to generate a pixel density matrix, where represents the average pixel density of the region, Represents the total number of pixels in the area. represents the intensity value of the i-th pixel, is the normalization coefficient, is the stabilization parameter; Applying the pixel density matrix, density thresholds are calculated and dirt areas are identified using the formula: ； Get the dirt area identification, where represents the density threshold for distinguishing dirty and clean areas, represent The mean of represent The standard deviation of represents the adjustment factor, is the threshold adjustment factor; Using the dirty region identification, a network segmentation cost strategy is applied using the formula: ； Generate segmentation labeling results, where Represents the confidence of the segmentation and labeling results of the clean area and the dirty area, is the scaling factor, Represents the dirt area logo, Represents the cost function.

5. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 4 is characterized in that: The steps for obtaining the state transition data are specifically as follows: Based on the segmentation and marking results, the frequency of dirt development of differentiated types is counted using the formula: ； Generate a frequency vector where Represents the average development trend indicator after taking into account the frequency of fouling development and regularization and adjustment factors, represents the sample size, represents the sample development frequency, represents the regularization parameter, represents the adjustment factor; Using the frequency vector, the hidden Markov model is initialized to simulate nonlinear state changes using the formula: ； Get the state transfer matrix, where Represents the value of the state transfer matrix under given fouling frequency and environmental parameters, and is the state sensitivity adjustment parameter, is the smoothing offset, Represents the average development trend indicator after taking into account the frequency of fouling development and regularization and adjustment factors; Based on the state transfer matrix, combined with equipment operation and environmental parameters, the state transfer data is calculated through the state transfer matrix, using the formula: ； in, Represents state transfer data, sin and cos are trigonometric function transformations of environmental parameters. and is the adjustment factor, is the normalization parameter, Represents the value of the state transfer matrix under given fouling frequency and environmental parameters, Represents environmental parameters.

6. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 5 is characterized in that: The steps for obtaining the dirt development trend diagram are specifically as follows: Based on the state transition data, the initial fouling data set is analyzed, and the initial fouling distribution is predicted by the exponential decay model using the formula: ； Generate initial dirt distribution, where represents the initial dirt distribution, is the time decay factor, is the stability constant, Represents state transfer data, represents the initial dirt dataset, and Represents the sensitivity adjustment parameter; Using the initial fouling distribution, the fouling development is predicted through nonlinear transformation using the formula: ； Simulate the change of fouling under different conditions and obtain the predicted fouling distribution, where For the predicted fouling distribution, is the current dirt distribution, represents the nonlinear adjustment factor, represents the adjustment constant, Represents state transfer data; Based on the predicted dirt distribution, the dirt distribution data at a time point is summarized, and the influence of the data at multiple time points is adjusted. The formula used is: ； Generate a fouling trend graph where: For the dirt development trend chart, represents the dirt distribution at a point in time, represents the weighting factor, Represents the total number of time points.

7. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 6 is characterized in that: The steps for obtaining the resource allocation scheme are specifically as follows: Based on the dirt development trend diagram, the cleaning demand is estimated using a nonlinear response function and calculated using a resource demand model using the formula: ； The predicted resource demand is obtained, where represents the predicted resource demand, , and is a dynamic coefficient adjusted according to historical data, and is the baseline demand adjustment parameter, Represents the dirt development trend graph; According to the predicted resource demand, a resource allocation algorithm is used to optimize the allocation of manpower and materials, using the formula: ； Generate a human and material allocation plan, including: To provide human and material allocation plan, It is the amount of resource buffer reserved for dealing with emergencies. Indicates the cleaning area that can be covered by unit resources. Represents the predicted resource demand; Based on the human resources and material allocation plan, the calculated data is integrated to form a resource allocation execution plan using the formula: ； Generate a resource allocation plan, where Represents a resource allocation plan, Indicates the number of time points within the planning cycle. Represents the resource allocation plan at each time point, is the time decay factor.

8. The intelligent analysis refrigeration equipment inner wall cleaning system according to claim 7 is characterized in that: The steps for obtaining the optimized scheduling plan are specifically as follows: Based on the resource allocation scheme, the real-time utilization efficiency of resources is calculated using the formula: ； The dynamic adjustment factor introduced strengthens the adaptability of the model to differentiated conditions and calculates the weighted average efficiency, where: represents the weighted average efficiency, Indicates The amount of resources allocated, represents the efficiency weight of resources, and is the adjustment factor for each resource; Using the weighted average efficiency, a dynamic scheduling strategy is formulated using the formula: ； Introduce adjustment factors to enhance the ability to respond to resource changes and generate scheduling strategies, among which, is the scheduling strategy, represents the weighted average efficiency, is the weight factor, and is a newly introduced regulatory factor; Based on the scheduling strategy, resources and the scheduling strategy are integrated and the following formula is adopted: ； Strengthen the model's sensitivity to time and generate an optimized scheduling plan, where To optimize the scheduling plan, For time The scheduling strategy, is the adjusted time decay factor.

Images