Defrosting control method based on spiral quick-freezing machine

Abstract

This application relates to the field of artificial intelligence technology and provides a defrosting control method based on a spiral quick-freezing machine. The method includes: configuring a monitoring unit to be connected to a defrosting unit, monitoring defrosting status data and performing feature analysis to construct a learning data set; obtaining control data to supplement and construct the learning data set; annotating and self-learning to generate a defrosting control analysis module; collecting information on the evaporator to be defrosted, parsing it and inputting it into the analysis module to determine a defrosting strategy; sending the strategy to a control unit to generate a control instruction to operate the defrosting unit. This application solves the technical problem that in defrosting control based on preset control parameters, it is impossible to adaptively adjust according to the real-time operating status and defrosting requirements of the quick-freezing machine, resulting in low efficiency and high energy consumption during the defrosting process, and achieves the technical effect of real-time monitoring of the defrosting status data of the defrosting unit, automatically adapting to different defrosting requirements, and improving the efficiency and accuracy of the defrosting process.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to the field of machine learning technology, and particularly to a defrost control method based on a spiral quick-freezing machine. Background Art

[0002] With the continuous progress of food processing technology and the increasing market demand, spiral quick-freezing machines are playing an increasingly important role in the field of food freezing and processing. Spiral quick-freezing machines, with their efficient and continuous production methods, are widely used in the quick-freezing treatment of various foods. However, with the expansion of production scale and the diversification of food types, during the long-term operation of the quick-freezing machine, frosting inevitably occurs on the surface of the evaporator, which directly affects the freezing effect and energy consumption level of the quick-freezing machine. Summary of the Invention

[0003] This application provides a defrost control method based on a spiral quick-freezing machine, aiming to solve the technical problems that in the defrost control based on preset control parameters, it is impossible to adaptively adjust according to the real-time operating state and defrosting requirements of the quick-freezing machine, resulting in low efficiency and high energy consumption during the defrosting process.

[0004] In view of the above problems, this application provides a defrost control method based on a spiral quick-freezing machine.

[0005] In an aspect of this application, a defrost control method based on a spiral quick-freezing machine is provided. The method includes: based on a humidity sensor, continuously monitoring the inner wall of the container of the target mixer in real time to obtain real-time monitoring data, where the real-time monitoring data includes humidity acquisition data at M key positions on the inner wall of the target mixer; obtaining the monitoring working characteristic indexes of the humidity sensor and the operating working characteristic indexes of the target mixer, and constructing a data twin model of the mixer; based on the change trend of the real-time monitoring data, evaluating the humidity level corresponding to the target mixer, judging the powder dryness degree according to the humidity level, and comparing it with a preset powder dryness threshold; if the powder dryness degree does not meet the preset powder dryness threshold, based on the historical monitoring data of the humidity sensor, combining the monitoring working characteristic indexes of the humidity sensor and the operating working characteristic indexes of the target mixer for efficiency compensation, determining the efficiency compensation parameter and the compensation credibility; based on the data twin model of the mixer, simulating the humidity level law under different working conditions, constructing a drying optimization space, and using the efficiency compensation parameter and the compensation credibility as the optimization auxiliary variables of the drying optimization space; performing parameter optimization in the drying optimization space to obtain the optimal process parameter combination, and performing optimization of the powder drying control of the target mixer.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The above defrosting control method based on a spiral quick-freezing machine collects the working state data of the defrosting unit in real time through a monitoring unit, and constructs a learning data set using this data. At the same time, the operation data of the control unit is obtained to enrich the learning data set, making the data more comprehensive and accurate. Subsequently, these learning data sets are labeled and self-learned to generate an intelligent defrosting control analysis module. This module is connected to the monitoring unit and the control unit to achieve real-time data exchange and instruction transmission. When the monitoring unit detects that the evaporator needs to be defrosted, it collects relevant defrosting information. This information is input into the defrosting control analysis module, and the module analyzes this information based on the learned knowledge to determine the optimal defrosting control strategy. Then, the control strategy is sent to the control unit, and the control unit generates control instructions according to the strategy to operate the defrosting unit. In this way, the entire defrosting process realizes intelligent and adaptive control, improving the defrosting efficiency and effect.

[0008] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the specific embodiments of this application are specifically given below. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0010] Figure 1 It is a schematic flowchart of the defrosting control method based on a spiral quick-freezing machine in an embodiment. Detailed Embodiments

[0011] The embodiments of this application solve the technical problems that in the defrosting control based on preset control parameters, it is impossible to adaptively adjust according to the real-time operating state and defrosting requirements of the quick-freezing machine, resulting in low defrosting efficiency and high energy consumption, by providing a defrosting control method based on a spiral quick-freezing machine.

[0012] The following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts belong to the scope of protection of this application.

[0013] It should be noted that the terms "including" and "having" and any of their variants are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0014] Embodiment 1

[0015] As Figure 1 shown, the present application provides a defrosting control method based on a spiral quick-freezing machine. The method includes:

[0016] Configure a monitoring unit to be connected to the defrosting unit, monitor the defrosting status data of the defrosting unit during the defrosting process, perform feature recognition and analysis on the defrosting status data, and construct a learning data set;

[0017] The defrosting control method based on the spiral quick-freezing machine is applied to a defrosting control system. This system consists of three parts: a control unit, a monitoring unit, and a defrosting unit. Among them, the monitoring unit is responsible for real-time monitoring of various status data of the defrosting unit during the defrosting process, such as temperature, humidity, etc., to ensure the accuracy and real-time nature of the data. The control unit is responsible for receiving the data transmitted by the monitoring unit and generating corresponding control instructions based on these data. The defrosting unit is the key component that performs the defrosting operation. It will work according to the instructions issued by the control unit to complete the defrosting task. Through the coordinated work of the defrosting control system, precise control of the defrosting process of the spiral quick-freezing machine can be achieved, improving the defrosting efficiency, reducing energy consumption, and ensuring the stable operation of the quick-freezing machine and food quality.

[0018] In the embodiment of the present application, the monitoring unit is connected to the defrosting unit to collect various status data during the defrosting process in real time. These data reflect the working conditions of the defrosting unit, such as the defrosting speed, temperature change, etc. Subsequently, convolutional channels are set to identify the defrosting status data, and fusion connection is performed according to the identification results to obtain a feature recognition result. After that, the identified feature recognition results are organized into a learning data set to provide data support for the subsequent optimization of the control strategy.

[0019] Furthermore, the present application provides a method for monitoring the defrosting status data of the defrosting unit during the defrosting process, performing feature recognition and analysis on the defrosting status data, and constructing a learning data set. The method further includes:

[0020] The monitoring unit includes a camera, and the defrosting status image of the defrosting unit is collected through the camera;

[0021] Set parallel first convolutional channels, second convolutional channels, and a subsequent fully connected layer. The first convolutional channel and the second convolutional channel include different image selection step sizes;

[0022] Preferably, the monitoring unit is equipped with multiple cameras for capturing real-time defrosting status images of the defrosting unit. To analyze these images more precisely, two parallel convolutional channels are set up: the first convolutional channel and the second convolutional channel. These two channels adopt different image selection strides during image processing, so as to be able to extract feature information of different scales. Specifically, for the task of defrosting effect recognition at the system terminal, the convolutional layer structure of the first convolutional channel is designed to extract features related to the defrosting effect. Since the defrosting effect involves the thickness, uniformity, coverage range, etc. of the frost layer, the convolutional layer needs to be able to capture these subtle features. Therefore, a smaller convolutional kernel and an appropriate stride are used to extract detailed features at different positions of the image. After the convolutional layer is determined, an appropriate activation function is used to increase non-linearity and obtain the required first convolutional channel. For example, if the size of the input images is 224x224 pixels, in order to capture the subtle features of the frost layer, a 3x3 convolutional kernel is selected because it can reduce the computational amount and increase the depth of the network while maintaining a certain receptive field. Subsequently, the stride is determined. The stride determines the step size at which the convolutional kernel slides on the input image. For the defrosting effect recognition task, relatively dense feature extraction is required, so the system terminal sets the stride to 1. In this way, the convolutional kernel slides pixel by pixel to ensure that the details of each position in the image are captured. After that, in order to keep the size of the output feature map the same as the input, the system terminal uses the same form of padding, so that the height and width of the feature map remain unchanged after the convolutional operation. Then, according to the task requirements, that is, the number of convolutional kernels to be extracted, the number of convolutional kernels is determined. After the convolutional layer is constructed, the system terminal adds an activation function to increase the non-linearity of the network and successfully constructs the first convolutional channel. For defrosting effect recognition of images with a size of 224x224 pixels, the system terminal uses ReLU (Rectified Linear Unit) as the activation function because ReLU can effectively help the network learn and extract meaningful features.

[0023] After the first convolution channel is constructed, the system terminal determines the convolution kernel and step size of the second convolution channel using the same method as the convolution layer of the first convolution channel. Subsequently, a recurrent neural network is added after the convolution layer of the second convolution channel to capture the temporal features for identifying the change of the defrost state over time. After that, the spatial features extracted by the convolution layer are aligned with the temporal features extracted by the recurrent neural network to ensure that the spatial features and the temporal features have the same spatial dimension. The aligned spatial features and temporal features are then added at the corresponding positions, and the second convolution channel is successfully constructed. This means that the feature value at each spatial position will be added to the corresponding temporal feature value to generate a new feature map. Specifically, the system terminal uses the convolution layer of the second convolution channel for feature extraction for each image frame. These convolution layers learn and extract spatial features related to the defrost state, such as the distribution, thickness, and texture of the frost layer. After each image frame is processed by the convolution layer, a feature map is output, which contains the representation of the important features in the frame image. Subsequently, each feature map is flattened into a one-dimensional feature vector. In this way, the entire image sequence is converted into a two-dimensional feature matrix, in which each row corresponds to a feature vector of an image frame. After that, a long short-term memory network (LSTM) is constructed to process the obtained feature matrix. The long short-term memory network has memory units that can capture long-term dependencies in sequence data. The long short-term memory network receives a feature vector as input, updates its internal state, and outputs a hidden state vector. This hidden state vector contains the accumulated information of the entire sequence so far. Then, through the processing of the long short-term memory network, a series of hidden state vectors are obtained, which capture the temporal features in the defrosting process image sequence. These hidden state vectors contain dynamic information about the defrosting state changing over time. Finally, the spatial features extracted by the convolution layer are fused with the temporal features extracted by the recurrent neural network by addition, and the fusion result is output. At this point, the second convolution channel is successfully constructed.

[0024] After the first convolutional channel and the second convolutional channel are set up, the outputs of the two convolutional channels are respectively connected to their respective fully connected layers for further feature transformation and classification. Then, the labeled defrost status image dataset is used as training data for training, and the recognition performance of the two channels is optimized by adjusting the network parameters and structure. Specifically, the system terminal inputs the training data into the network, and performs forward propagation through the convolutional layer, pooling layer, and fully connected layer to obtain the output of the network. Subsequently, the output of the network is compared with the true label, and the value of the cross-entropy loss is calculated. The cross-entropy loss value measures the difference between the network prediction and the true label. Then, the backpropagation algorithm is used to calculate the gradient of the cross-entropy loss value with respect to the network parameters, and the network weights are updated by the optimizer to minimize the loss function. Repeat the above steps until the preset number of training epochs is reached. Finally, the first convolutional channel will output the recognition result of the defrost effect, while the second convolutional channel will output the recognition result of the defrost status change result. In summary, by collecting images through a camera and using two parallel convolutional channels to extract features at different scales, an in-depth analysis and understanding of the defrost status image are achieved.

[0025] The defrost effect of the defrost status image is recognized through the first convolutional channel, and the defrost status change result is recognized through the second convolutional channel;

[0026] Preferably, the system terminal recognizes the defrost effect of the input defrost status image through the first convolutional channel set. The first convolutional channel extracts features related to the defrost effect in the image, such as the distribution and thickness of the frost layer, through the built-in convolutional layer, activation function, and pooling layer. Subsequently, these features are sent to the fully connected layer for further transformation and classification, and finally the recognition result of the defrost effect is output. The second convolutional channel focuses on recognizing the result of the defrost status change. It also extracts features from the input defrost status image, but focuses more on capturing the dynamic features of the defrost status change over time in the image sequence. These features include the process of the frost layer gradually disappearing and the change in the defrost speed. Through the processing of the fully connected layer, the second convolutional channel can output the recognition of the defrost status change result, helping to understand the progress and effect of the defrost process. In summary, the first convolutional channel performs static defrost effect recognition, while the second convolutional channel performs dynamic defrost status change recognition. Combining the first convolutional channel and the second convolutional channel can more comprehensively analyze and understand the defrost process, providing strong support for practical applications.

[0027] The defrost effect recognition result output by the first convolutional channel is fused and connected with the defrost change timing feature output by the second convolutional channel to obtain the timing change feature data of the defrost status as the feature recognition result of the defrost status data, and the learning dataset is constructed.

[0028] Preferably, the system terminal fuses and connects the defrosting effect recognition result output by the first convolutional channel with the defrosting change timing feature output by the second convolutional channel, which can help the system terminal more comprehensively understand the timing change of the defrosting state. Specifically, after operations such as convolution and pooling, the first convolutional channel outputs a feature map of the defrosting effect recognition result. These features mainly reflect the static information of the defrosting effect in the image. The second convolutional channel also undergoes operations such as convolution and pooling, but it focuses more on extracting the timing features of the defrosting state change and outputs a series of feature maps, which capture the dynamic information changing over time during the defrosting process. Subsequently, the defrosting effect recognition result features output by the first convolutional channel are fused with the defrosting change timing features output by the second convolutional channel. The fusion method is to add the feature values at the corresponding positions to obtain new fused features, and the timing change feature data of the fused defrosting state is used as the feature recognition result of the defrosting state data. The fused features contain both the static information of the defrosting effect and the dynamic information of the defrosting state change, thus more comprehensively describing the timing change of the defrosting state. After that, the fused defrosting state timing change feature data is combined with the corresponding label (i.e., the true defrosting state) to form a learning data sample. Each sample contains the feature representation of the input image and the corresponding label. Then, all the samples are summarized to form a learning data set. The learning data set is crucial for training and optimizing the network. It contains rich defrosting state information, enabling the network to more accurately recognize and understand the timing change of the defrosting process.

[0029] Obtain the control data of the control unit, and according to the time mapping relationship between the control data and the learning data set, use the control data to supplement and construct the learning data set. Each group of data in the supplemented and constructed learning data set includes the recognition feature of the defrosting state data and its corresponding control data;

[0030] In one embodiment, to improve the learning data set, the system terminal obtains the control data of the control unit. There is a clear corresponding relationship in time between the control data and the data in the learning data set. Subsequently, the control data is used to supplement and construct the learning data set. In the supplemented learning data set, each group of data not only includes the recognition feature of the defrosting state data but also adds the corresponding control data. In this way, the system terminal obtains a more comprehensive and richer learning data set, which covers the defrosting state features and the corresponding control information, provides more powerful data support for subsequent model training and optimization, and improves the accuracy of defrosting state recognition.

[0031] Furthermore, the present application provides the method of obtaining the control data of the control unit, and according to the time mapping relationship between the control data and the learning data set, using the control data to supplement and construct the learning data set, the method further includes:

[0032] Obtain the control angle of the defrosting unit, the control parameters of the solenoid valve, and the corresponding output nozzle pressure, as well as the control sending time, as the control data;

[0033] Align the time of collecting data in the learning dataset with the control sending time, establish a mapping relationship between the control data and the learning dataset, add the control data to the learning dataset, and supplement and construct the learning dataset.

[0034] Optionally, to improve the learning dataset, the system terminal obtains the control angle of the defrosting unit, the control parameters of the solenoid valve, and their corresponding output nozzle pressure, as well as the sending time of these control data. These control data are crucial for understanding the defrosting process and optimizing the defrosting effect.

[0035] After obtaining the control data, the system terminal matches these control data with the data in the learning dataset. Specifically, the system terminal obtains the collection timestamp of each sample in the learning dataset. These timestamps record the collection moments of each defrost state characteristic data in the dataset. Then obtain the sending timestamp of the control data. These timestamps correspond to the moments when the control unit sends control instructions, including the sending times of control data such as the control angle of the defrosting unit, the control parameters of the solenoid valve, and the corresponding output nozzle pressure. Subsequently, compare the collection timestamp in the learning dataset with the sending timestamp of the control data, find the matching points between them, and correspond these timestamps one by one, indicating that the defrost state characteristic data is collected and the corresponding control instructions are sent at the same moment. When the timestamp alignment is completed, the system terminal establishes a mapping relationship between the control data and the learning dataset. The establishment of the mapping relationship needs to ensure that each sample in the learning dataset can find the corresponding control data, and vice versa. In this way, in subsequent data processing, the corresponding control data of each sample can be conveniently obtained according to the mapping relationship. After having the mapping relationship, the system terminal starts to add the control data to the learning dataset and supplement and construct the dataset. During the supplement and construction process, it is necessary to ensure the integrity and consistency of the data. That is, each sample in the dataset should have the corresponding defrost state characteristic data and control data, and these data are consistent in time. So far, the system terminal has completed the supplement and expansion of the learning dataset. The supplemented learning dataset not only contains the recognition features of the defrost state, but also contains the corresponding control data. Such a dataset provides more comprehensive and in-depth information for the system terminal, helps to better understand the defrosting process, improve the accuracy of defrost state recognition, and provide strong support for subsequent model training and optimization.

[0036] Annotate the learning data set according to the preset learning objective, perform self-learning on the learning data set, generate a defrost control analysis module, and connect the defrost control analysis module to the monitoring unit and the control unit;

[0037] In one embodiment, the system terminal performs detailed annotation work on the learning data set according to the preset learning objective. These annotated data provide clear guidance for the learning data set, enabling the defrost control analysis module to clearly know what analysis process and output result the current input data should correspond to. Subsequently, the system terminal uses these annotated learning data sets for self-learning to generate corresponding models. Through continuous iteration and optimization, these models gradually learn how to extract useful features from the input defrost state data and generate corresponding control strategies based on these features. After that, the system terminal combines these trained models to obtain a defrost control analysis module, which is connected to the monitoring unit and the control unit to achieve intelligent analysis and control of the defrost process.

[0038] Furthermore, the present application provides the method of annotating the learning data set according to the preset learning objective, performing self-learning on the learning data set, and generating a defrost control analysis module, and the method further includes:

[0039] The preset learning objective includes control angle, nozzle pressure, and control strategy optimization. When the preset learning objective is the control angle, obtain a sample set for adjusting the control angle based on the learning data set, establish a mapping relationship between the control angle and the defrost state data, construct an angle learning sample set by identifying the control angle, and obtain a control angle analysis model through training and learning of the angle learning sample set;

[0040] Optionally, the preset learning objectives include control angle, nozzle pressure, and optimization of the control strategy. When the system terminal focuses on the objective of control angle, the system terminal first filters out a sample set related to the adjustment of the control angle from the learning dataset. These sample sets contain control angle data under different defrosting states, providing rich training materials for the system terminal. Subsequently, a mapping relationship between the control angle and the defrosting state data is established. This means that the system terminal can clearly understand what control angle should be adopted under a specific defrosting state. Through the analysis of these mapping relationships, an angle learning sample set is further constructed, and the control angle is clearly identified so that the model can accurately learn. After obtaining the angle learning sample set, the system terminal initializes a basic model, which is a simple rule-based model. Then, the angle learning sample set is input into the self-learning algorithm, allowing the basic model to learn the mapping relationship from the defrosting state to the control angle by continuously adjusting its internal parameters and structure. During the self-learning process, the system terminal sets evaluation metrics to monitor the performance of the model and adjusts the learning algorithm or hyperparameters as needed. As the number of iterations increases, the model will gradually learn more useful information from the sample set, and its ability to predict the control angle will also continuously improve. When the self-learning process reaches the preset number of times, the iteration is stopped, and the currently trained basic model is saved as the final control angle analysis model. The control angle analysis model learns the complex relationship between the control angle and the defrosting state from a large number of samples through self-learning and can predict the appropriate control angle based on the input defrosting state data in practical applications.

[0041] When the preset learning objective is the nozzle pressure, a mapping relationship between the nozzle pressure and the defrosting state data is established, the nozzle pressure is identified to construct a pressure control learning sample set, and a control pressure analysis model is obtained through the training and learning of the pressure control learning sample set;

[0042] Optionally, when the preset learning objective is the nozzle pressure, the system terminal follows steps similar to those for obtaining the control angle analysis model described above. First, a mapping relationship between the nozzle pressure and the defrosting state data is established based on the learning data set. This means that the system terminal needs to identify how the nozzle pressure should be adjusted under different defrosting states to achieve the best defrosting effect. Subsequently, the nozzle pressure is labeled, and this data is combined with the corresponding defrosting state data to form a pressure control learning sample set. This sample set provides the system terminal with rich data resources for subsequent model training. Then, the same method is used to train and learn the basic model using the pressure control learning sample set. By continuously optimizing the parameters and structure of the basic model, the basic model can gradually learn from the sample set how to predict the appropriate nozzle pressure from the defrosting state data. Then, through training and learning, a control pressure analysis model is obtained. This model can intelligently predict and adjust the nozzle pressure based on real-time defrosting state data to achieve a more efficient defrosting effect.

[0043] When the preset learning objective is the optimization of the control strategy, the control angle analysis model and the control pressure analysis model are combined, and an optimization processing model is added at the back. The angle control analysis result and the pressure control analysis result are comprehensively optimized, including setting the optimization target value and the optimization iteration rule, to obtain a comprehensive optimization analysis model;

[0044] Optionally, when the preset learning objective is control strategy optimization, the goal of the system terminal is to further improve the defrosting effect by comprehensively optimizing the control angle and nozzle pressure. First, the system terminal combines the previously obtained control angle analysis model and control pressure analysis model. This means that the system terminal considers these two models as a whole, so that the impacts of the control angle and nozzle pressure can be taken into account simultaneously during the optimization process. Subsequently, based on the combined model, an optimization processing model is placed behind. The role of this model is to receive the outputs from the control angle analysis model and the control pressure analysis model, and comprehensively evaluate and adjust these two outputs according to the preset optimization target value and the current defrosting state data. Among them, the optimization target value is set according to the actual application scenarios and requirements. These target values include key performance indicators such as defrosting efficiency, energy consumption, and defrosting time. Setting clear optimization target values provides an optimization direction for the optimization processing model, enabling it to optimize for these performance indicators. After that, the system terminal formulates optimization iteration rules, including selecting optimization algorithms, setting the number of iterations, step sizes, and other parameters. These rules determine how the optimization processing model adjusts the values of the control angle and nozzle pressure according to the current optimization results, and gradually approaches the optimization target value through multiple iterations. For example, when the control strategy optimization is to minimize the defrosting time, the system terminal selects the gradient descent algorithm as the optimization algorithm because the problem of finding the combination of the control angle and nozzle pressure that can minimize the defrosting time can be transformed into a problem of minimizing an objective function. The gradient descent algorithm is an iterative optimization algorithm that gradually approaches the minimum value of the objective function by calculating the gradient of the objective function with respect to the control parameters and updating the parameter values in the opposite direction. Since the defrosting time is expressed as a continuous function of the control parameters and its gradient can be calculated, the gradient descent algorithm is an optimization algorithm suitable for this problem. Subsequently, considering the complexity of the defrosting system and the desired accuracy, the required number of iterations is set. After that, the step size is initially set, and the choice of the step size is a trade-off between the convergence speed and stability. If the step size is too large, the optimization process will oscillate near the minimum value and cannot converge; if the step size is too small, the convergence speed may be too slow. Finally, after the above steps, the system terminal obtains a comprehensive optimization analysis model. The comprehensive optimization analysis model can comprehensively consider the impacts of the control angle and nozzle pressure, and find an optimal control strategy through optimization processing, thereby realizing the intelligent control and optimization of the defrosting process. In summary, when the preset learning objective is control strategy optimization, the comprehensive optimization of the defrosting control strategy is achieved by combining the control angle analysis model and the control pressure analysis model and placing the optimization processing model behind. This process involves key steps such as setting optimization target values and optimization iteration rules, aiming to find the best control strategy to improve the defrosting effect.

[0045] Configure the target task, obtain the adapted model from the control angle analysis model, control pressure analysis model, and comprehensive optimization analysis model, and generate the defrost control analysis module.

[0046] Optionally, when configuring the target task, the system terminal first selects a suitable model for combination from the established control angle analysis model, control pressure analysis model, and comprehensive optimization analysis model. This process is determined according to the specific requirements and optimization objectives of the current defrost task. The selected adapted model will be able to more accurately reflect the influence of the control angle and nozzle pressure on the defrost effect, and can find the best parameter combination during the optimization process. Subsequently, these adapted models are used to generate the defrost control analysis module. This module will integrate the prediction and optimization functions of the model, and can calculate the optimal control angle and nozzle pressure values in real time according to the actual defrost state data and control objectives. In this way, we can use this module to guide the operation of the defrost system to ensure that while achieving the defrost effect, other performance index requirements can also be met. Generally speaking, configuring the target task and generating the defrost control analysis module by selecting the adapted model is to achieve the intelligent control of the defrost system, improve the defrost efficiency and optimize the system performance.

[0047] Collect the information of the evaporator to be defrosted through the monitoring unit, perform feature analysis on the information of the evaporator to be defrosted, obtain the frost area distribution and frost thickness, and input the frost area distribution and frost thickness into the defrost control analysis module for analysis and processing to determine the defrost control strategy;

[0048] In one embodiment, the system terminal collects the information of the evaporator to be defrosted in real time through the monitoring unit. These information reflect the frosting situation on the surface of the evaporator and are crucial for formulating an effective defrost control strategy. Subsequently, the collected information of the evaporator to be defrosted is subjected to feature analysis to obtain key data such as the area distribution and thickness of the frost. These data provide the system terminal with detailed information about the frosting situation, which helps to understand the severity and distribution of the frosting on the surface of the evaporator. After obtaining the frost area distribution and thickness, the system terminal inputs these data into the defrost control analysis module for processing. This module will analyze and calculate according to the input frost information and the internal model. Through this process, the optimal defrost control strategy can be determined, including the adjustment scheme of the control angle and nozzle pressure, as well as the timing and duration of defrosting. Finally, through data collection by the monitoring unit, feature analysis, and processing by the defrost control analysis module, an effective defrost control strategy can be determined to guide the operation of the defrost system to ensure that the frost on the surface of the evaporator is removed in time, thereby improving the performance and efficiency of the defrost system.

[0049] Further, the present application provides a method of collecting the information of the evaporator to be defrosted by the monitoring unit, analyzing the characteristics of the information of the evaporator to be defrosted, and obtaining the frost formation area distribution and frost formation thickness. The method further includes:

[0050] Configure the camera of the monitoring unit to the coverage range of the evaporator, and collect the frost formation image on the surface of the evaporator;

[0051] Identify the coverage range, frost formation thickness, and frost formation area distribution of the frost formation image on the surface of the evaporator. Among them, the frost formation image on the surface of the evaporator includes multiple shooting angles. The coverage range and frost formation area distribution are identified by the frontal view angle through border recognition and color recognition, and the frost formation thickness is identified by the side view angle;

[0052] Optionally, in order to effectively monitor the frost formation situation of the evaporator, the system terminal configures the camera of the monitoring unit to ensure that its coverage range can fully cover the surface of the evaporator. In this way, the camera can collect the frost formation image on the surface of the evaporator in real time, providing intuitive and detailed data for the system terminal. Subsequently, a series of processing and analysis are performed on the collected frost formation image on the surface of the evaporator. In order to obtain more comprehensive frost formation information, the system terminal adopts the method of multiple shooting angles. By shooting from the frontal view angle, border recognition and color recognition can be performed on the coverage range and frost formation area distribution. For border recognition, the system terminal automatically detects the boundary of the frost formation area in the image, and based on pixel changes, color differences, or texture features, identifies the boundary line between the frost formation area and the non-frost formation area. Then, the coverage range of the frost formation is determined by a closed curve, and the corresponding border is generated. For color recognition, the system terminal sets appropriate color thresholds according to the color performance of the frost formation in the image, and performs segmentation through the color thresholds to divide the pixels in the image into frost formation pixels and non-frost formation pixels, and then analyzes the distribution of the frost formation pixels to obtain detailed information on the frost formation area distribution. At the same time, the system terminal also shoots from the side view angle to identify the frost formation thickness. The system terminal calculates the number of pixel points inside the contour according to the result of the border recognition to estimate the area of the frost formation. This area can indirectly reflect the thickness of the frost formation because thicker frost formation will occupy a larger area. Through shooting from this angle, the system terminal can more accurately judge the accumulation degree of the frost formation on the surface of the evaporator, providing an important basis for formulating subsequent defrosting control strategies. Through this series of image recognition and processing processes, key information such as the coverage range, frost formation thickness, and frost formation area distribution of the frost formation on the surface of the evaporator can be obtained. These information will provide strong data support for subsequent defrosting control analysis, helping to formulate more accurate and effective defrosting control strategies.

[0053] Perform position matching and fusion on the identification of the coverage range, frost formation thickness, and frost formation area distribution, and establish the corresponding relationship between the coverage range, frost formation distribution position, and frost formation thickness.

[0054] Optionally, by comprehensively applying the image recognition results from the front view and side view angles, the system terminal can perform position matching and fusion on the coverage range, frost thickness, and frost area distribution on the evaporator surface. This process is actually integrating the recognition information from different dimensions to establish a comprehensive correspondence. Specifically, to fuse the image recognition results from different angles, the system terminal calibrates the coordinates of the images so that the position information in the front view and side view images can correspond to each other, ensuring that these results are in the same coordinate system. Subsequently, based on the frost coverage range (border) and frost area distribution (color recognition area) obtained from the front view angle recognition, the specific distribution position of the frost on the evaporator surface is determined. These positions are represented in pixel coordinates. Then, the frost thickness data obtained from the side view angle recognition is associated with the frost distribution position recognized from the front view angle. For each frost area in the front view image, the thickness data at the corresponding position is found in the side view image. After determining the frost distribution position and the corresponding thickness, position matching and data fusion are performed. This means combining the position information of each frost area with its corresponding thickness information to form a comprehensive description of the frost situation. Then, the system terminal establishes the correspondence between the coverage range, frost distribution position, and frost thickness. In this way, the system terminal can clearly know which positions on the evaporator have frost and the frost thickness at these positions. The establishment of this correspondence provides important data support for formulating subsequent defrost control strategies.

[0055] Further, the present application provides a method of inputting the frost area distribution and frost thickness into the defrost control analysis module for analysis and processing to determine the defrost control strategy. The method further includes:

[0056] Obtain the defrost control parameters of the local spiral quick-freezing machine and determine the target task, where the target task is used to describe the adjustment target parameters of the defrost control parameters;

[0057] Optionally, to obtain the defrost control parameters of the local spiral freezer and determine the target task, the system terminal first deeply understands the operating principle of the spiral freezer and its defrost mechanism. The defrost control parameters include the temperature threshold for defrost start, defrost duration, defrost interval, etc. The setting of these parameters directly affects the operating efficiency of the freezer and the frosting situation. The target task is to determine the adjustment target parameters of the defrost control parameters based on the actual operating conditions and target performance of the freezer. For example, if the freezer frequently frosts during operation, resulting in increased energy consumption and affecting the freezing effect, then the target task is to reduce the frosting frequency and improve the operating efficiency of the freezer by adjusting the defrost control parameters. Specifically, the system terminal analyzes the advantages and disadvantages of the current defrost control parameters according to the historical operating data of the freezer, environmental temperature and humidity, etc., and then determines the adjustment target parameters, such as increasing the temperature threshold for defrost start to reduce unnecessary defrost operations, or increasing the defrost duration to ensure that each defrost can completely remove the frost. In summary, obtaining the defrost control parameters of the spiral freezer and determining the target task is to achieve the efficient and stable operation of the freezer by analyzing and adjusting these parameters, and reduce the impact of frosting on the freezing effect and energy consumption.

[0058] When the adjustment target parameter is the nozzle angle, configure the control angle analysis model to generate a defrost control analysis module, and perform control angle analysis based on the frost area distribution and frost thickness to obtain the control angle and output it as the defrost control strategy;

[0059] Optionally, when the adjustment target parameter is the nozzle angle, the system terminal selects the constructed control angle analysis model and integrates the control angle analysis model into the defrost control analysis module. The defrost control analysis module can analyze based on the frost area distribution and frost thickness and output an appropriate control angle. This control angle is a key parameter in the defrost control strategy. Specifically, the control angle analysis model in the defrost control analysis module analyzes the frost area distribution and frost thickness to understand the specific frosting situation in the freezer, including which places have more serious frosting and which places have thinner frosting. Subsequently, based on this information, the control angle analysis model will calculate the optimal nozzle angle so that the airflow ejected by the defrost nozzle can more effectively cover the frosting area and improve the defrost efficiency. Then, this control angle will be output as the defrost control strategy to guide the defrost operation of the freezer. In summary, by constructing the control angle analysis model and analyzing according to the frosting situation, an appropriate nozzle angle can be obtained, thereby optimizing the defrost effect of the freezer.

[0060] When the adjustment target parameter is the nozzle pressure, configure the control pressure analysis model to generate a defrost control analysis module, and perform control pressure analysis based on the frost area distribution and frost thickness to obtain the pressure control parameter and output it as the defrost control strategy;

[0061] Optionally, when the adjustment target parameter is determined to be the nozzle pressure, the system terminal selects the constructed control pressure analysis model and integrates the control pressure analysis model into the defrost control analysis module. The function of this module is to perform precise pressure control analysis based on the frost formation area distribution and frost formation thickness. By deeply analyzing the specific conditions of frost formation, including the breadth of the frost formation area and the density of frost formation, the defrost control analysis module can calculate the optimal nozzle pressure parameters. These pressure parameters are key elements in the defrost control strategy and will directly guide the nozzle pressure adjustment of the quick-freezing machine during the defrosting process. Finally, these precisely calculated pressure parameters will be output as part of the defrost control strategy to optimize the defrosting effect of the quick-freezing machine, ensuring that it can effectively remove frost while maintaining efficient operation and improving the overall performance.

[0062] When the adjustment target parameters include the nozzle angle and the nozzle pressure, configure the comprehensive optimization analysis model to generate a defrost control analysis module, optimize the control angle and control pressure according to the frost formation area distribution and frost formation thickness, and perform evaluation and optimization according to the preset target value to find the comprehensive control strategy of the control angle and control pressure with the highest target value evaluation as the defrost control strategy for output.

[0063] Optionally, when the adjustment target parameters include both the nozzle angle and the nozzle pressure at the same time, the system terminal selects the constructed comprehensive optimization analysis model and generates a corresponding defrost control analysis module. The function of this module is to perform joint optimization analysis on the nozzle angle and the nozzle pressure according to the actual situation of the frost formation area distribution and frost formation thickness. Specifically, the system terminal passes through the defrost control analysis module and combines the specific conditions of frost formation to find the best combination of the nozzle angle and the nozzle pressure. In this process, the defrost control analysis module will perform evaluation and optimization according to the preset target values, such as maximizing the defrost efficiency and minimizing the time. The defrost control analysis module will try different combinations of angles and pressures and evaluate the effects of each combination through simulation tests. Finally, the defrost control analysis module finds the comprehensive control strategy with the highest target value evaluation, that is, the best combination of the nozzle angle and the nozzle pressure, and outputs it as the defrost control strategy. This strategy will guide the quick-freezing machine on how to adjust the angle and pressure of the nozzle during the defrosting process to achieve the optimal defrosting effect. In this way, the system terminal can comprehensively consider multiple adjustment parameters and find the best comprehensive control strategy, thereby further improving the defrosting efficiency and operating performance of the quick-freezing machine.

[0064] Furthermore, before inputting the frost formation area distribution and frost formation thickness into the defrost control analysis module for analysis and processing, the method further includes:

[0065] According to the frosting area distribution and frosting thickness, the area to be defrosted is segmented, and the segmented areas include different frosting thicknesses;

[0066] Optionally, the system terminal carefully segments the area to be defrosted according to the specific conditions of the frosting area distribution and frosting thickness. This segmentation is based on the frosting thickness of different areas, which means that areas with similar frosting thicknesses are divided into the same segmented area. The system terminal sets multiple different frosting thickness thresholds according to the actual situation of the quick-freezing machine and the defrosting requirements. These thresholds will be used to divide the area to be defrosted into sub-areas with different frosting thicknesses. Subsequently, according to the set frosting thickness thresholds, the area to be defrosted is segmented to obtain multiple frosting areas. In the above way, the system terminal can obtain multiple segmented areas, and each area represents a different range of frosting thicknesses. For example, primary segmented areas, secondary segmented areas, tertiary segmented areas, etc. Such segmentation helps to more accurately understand the frosting situation inside the quick-freezing machine and provides an important basis for formulating targeted defrosting control strategies in the follow-up.

[0067] According to the frosting thickness of the segmented areas, the segmented areas are sorted in ascending order. The minimum frosting thickness is used as the base thickness and set as the base identifier 1. The thickness difference between adjacent arranged areas is used as the grade compensation value to generate the defrosting grade identifiers for each segmented area. The grade identifier of the segmented area is equal to the base identifier plus the grade compensation value;

[0068] Optionally, the system terminal counts the number of frosting areas included in each segmented area, adds up the thicknesses of each frosting area, and then divides by the number of frosting areas to obtain the average frosting thickness of each segmented area. Subsequently, according to the average frosting thickness of the segmented areas, these areas are sorted in ascending order according to the thickness. Then, the minimum frosting thickness is selected as the base thickness and set as the base identifier 1. After that, the thickness difference between adjacent arranged areas is calculated and this difference is defined as the grade compensation value. Then, the base identifier is added to the grade compensation value to generate the defrosting grade identifier for each segmented area. For example, if the frosting thickness of a certain area is 2, the difference between this area and the base thickness is calculated as 1, and this difference of 1 is added to the base identifier in the form of the grade compensation value to obtain the corresponding defrosting grade identifier 2 for this area, and so on. These grade identifiers help the system terminal clearly identify the severity of frosting in different areas and provide a basis for formulating targeted defrosting strategies in the follow-up. In this way, each segmented area is assigned a unique defrosting grade identifier. In subsequent defrosting operations, more accurate and efficient defrosting strategies can be formulated according to these grade identifiers.

[0069] Multiply the numerical value of the defrost level identifier by the area of the divided area, and then perform a proportion calculation with the total amount to convert it into the control adjustment coefficient of the divided area. The control adjustment coefficient is used to describe the weight proportion value of the control parameter in the defrost control analysis.

[0070] The operation expression of the control adjustment coefficient is: , where is the control adjustment coefficient of the i-th divided area, The level compensation value of the i-th divided area, representing the difference between the frost thickness and the base thickness of this area, is the area of the i-th divided area, j is the total number of divided areas, and B is the base identifier, B = 1.

[0071] Optionally, the system terminal multiplies the defrost level identifier of each divided area by the corresponding area size, aiming to combine the frost thickness and the area size, and comprehensively consider the importance of each area in the overall defrosting process. Subsequently, perform a proportion calculation on the obtained result with the total amount (total defrost area), calculate the proportion of each divided area in the overall defrosting, and use the calculated proportion as the control adjustment coefficient. The control adjustment coefficient is a weight value, which describes the influence degree of each divided area on the control parameter during the defrost control analysis. The system terminal sets the corresponding defrost control parameter for the one with a larger control adjustment coefficient according to the size of the control adjustment coefficient, so as to give priority to solving the area with serious frosting. Through this method, the defrost control strategy can be adjusted more precisely according to the frosting situation and area size of each area, ensuring that the defrost operation is both efficient and accurate.

[0072] Furthermore, the present application provides the method for identifying the frost thickness from the side view angle, and the method further includes:

[0073] Perform border recognition on the captured image of the side view angle and perform binarization processing;

[0074] Based on the binarization of the image, perform pixel distribution recognition, and perform quantity superposition according to the distribution recognition to determine the frost area;

[0075] Set a reference cross-section, starting from the reference cross-section, and estimate according to the frost area to determine the frost thickness distribution.

[0076] Optionally, the system terminal uses the same method as described above to identify the border of the image taken from the side view angle. This step is to accurately define the frosting area that needs to be analyzed in the image. Subsequently, the captured image is converted into a grayscale image, and a binarization threshold is set. The selection of the binarization threshold determines which pixels will be regarded as white and which pixels will be regarded as black. Among them, the binarization threshold is selected according to the simple binarization method, that is, a fixed threshold is selected, such as 127. After the binarization threshold is determined, the system terminal compares each pixel of the image with the threshold. If the grayscale value of the pixel is greater than or equal to the threshold, it is set to white; otherwise, it is set to black. This process will convert the image into a binary image, that is, an image that only contains two colors, black and white. Then, based on the binarized image, the system terminal performs pixel distribution recognition. By counting the distribution of black pixels (representing the frosted part), the frosting area is estimated. Specifically, the system terminal counts the number of black pixel points in each black pixel aggregation area (frosting area), so that the approximate area of the frosting area can be obtained. Subsequently, in order to determine the thickness distribution of the frost, the system terminal sets a reference section. This reference section is actually a benchmark for comparing with the area of each frosting area. By comparing the relationship between the area of each frosting area and the reference section, the frost thickness of each area can be estimated. In summary, this process uses image processing technology to identify and analyze the frosting situation inside the quick-freezing machine. By setting the comparison between the reference section and the frosting area, the thickness distribution of the frost is estimated, providing accurate data support for subsequent defrosting operations.

[0077] Send the defrosting control strategy to the control unit, and the control unit generates a control instruction by performing parameter identification according to the defrosting control strategy to operate and control the defrosting unit.

[0078] In one embodiment, when the system terminal obtains the defrosting control strategy, it sends this policy information to the control unit. The control unit is the core component responsible for executing various operations inside the quick-freezing machine. Once it receives the defrosting control strategy, the control unit will immediately start working. It will perform parameter identification according to the instructions in the strategy and generate corresponding control instructions. These instructions will directly guide how the defrosting unit operates, including how to adjust the angle and pressure of the nozzle, and when to start and end the defrosting process, etc. In this way, it can ensure that the quick-freezing machine can perform defrosting according to the optimized strategy, thereby improving the defrosting efficiency, reducing energy consumption, and extending the service life of the quick-freezing machine.

[0079] In summary, the embodiments of the present application at least have the following technical effects:

[0080] In the embodiment of the present application, a monitoring unit is configured to be connected to a defrosting unit, monitor the status data during the defrosting process, perform feature recognition and analysis, and construct a learning data set. The control data of the control unit is obtained and supplemented and constructed into the learning data set by using the time mapping relationship. The learning data set is labeled and self-learned to generate a defrosting control analysis module, which is connected to the monitoring unit and the control unit. The information to be defrosted of the evaporator is collected by the monitoring unit, feature analysis is performed, the frosting area distribution and the frosting thickness are obtained, and the data is input into the defrosting control analysis module for processing to determine the defrosting control strategy. The defrosting control strategy is sent to the control unit, and the control unit generates a control instruction to perform operation control on the defrosting unit. In addition, steps such as dividing the area to be defrosted, arranging the divided areas in positive order, and calculating the control adjustment coefficient are also performed to further optimize the defrosting control strategy. These technical effects together solve the technical problem that in the defrosting control based on preset control parameters, it is impossible to adaptively adjust according to the real-time operating state of the quick-freezing machine and the defrosting requirements, resulting in low efficiency and high energy consumption during the defrosting process, and achieve the technical effects of real-time monitoring of the defrosting status data of the defrosting unit, automatically adapting to different defrosting requirements, and improving the efficiency and accuracy of the defrosting process.

[0081] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the specific order and continuous sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0083] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications 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 present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A defrosting control method based on a spiral quick freezer, characterized in that: The method uses a defrost control system, the system includes a control unit, a monitoring unit, and a defrost unit, and the method includes: A monitoring unit is configured to be connected to a defrost unit, to monitor defrost state data of the defrost unit during the defrost process, to perform feature recognition analysis on the defrost state data, and to construct a learning data set; Acquire control data of a control unit, and use the control data to supplement the learning data set according to a time mapping relationship between the control data and the learning data set, wherein each set of data in the supplementary learning data set includes an identification feature of the defrost state data and its corresponding control data; Annotate the learning data set according to a preset learning goal, perform self-learning on the learning data set, and generate a defrost control analysis module, wherein the defrost control analysis module is connected to the monitoring unit and the control unit; The monitoring unit collects the information of the evaporator to be defrosted, performs feature analysis on the information of the evaporator to be defrosted, obtains the distribution of frosting area and the thickness of frosting, inputs the distribution of frosting area and the thickness of frosting into the defrosting control analysis module for analysis and processing, and determines the defrosting control strategy; The defrost control strategy is sent to the control unit, and the control unit performs parameter identification according to the defrost control strategy to generate control instructions and perform operation control on the defrost unit; Among them, building a learning data set includes: The monitoring unit includes a camera, and collects a defrosting state image of the defrosting unit through the camera; Setting a parallel first convolution channel, a second convolution channel, and a post-fully connected layer, wherein the first convolution channel and the second convolution channel include different image selection steps; Recognize the defrost effect of the defrost state image through the first convolution channel, and recognize the defrost state change result through the second convolution channel; The defrost effect recognition result output by the first convolution channel is fused with the defrost change timing feature output by the second convolution channel to obtain the timing change feature data of the defrost state as the feature recognition result of the defrost state data, and to construct the learning data set.

2. The method according to claim 1, characterized in that The acquiring control data of the control unit, and supplementing and constructing the learning data set by using the control data according to the time mapping relationship between the control data and the learning data set, comprises: Obtaining a control angle of a defrost unit, a control parameter of a solenoid valve and a corresponding output nozzle pressure, and a control sending time as the control data; According to the alignment of the data collection time in the learning data set and the control sending time, a mapping relationship between the control data and the learning data set is established, the control data is added to the learning data set, and the learning data set is supplemented and constructed.

3. The method according to claim 2, characterized in that The step of labeling the learning data set according to the preset learning target, performing self-learning on the learning data set, and generating a defrost control analysis module includes: The preset learning objectives include control angle, nozzle pressure, and control strategy optimization. When the preset learning objective is the control angle, a sample set of control angle adjustment is obtained based on the learning data set, a mapping relationship between the control angle and the defrost state data is established, the control angle is identified to construct an angle learning sample set, and a control angle analysis model is obtained by training and learning the angle learning sample set; When the preset learning target is nozzle pressure, a mapping relationship between nozzle pressure and defrost state data is established, the nozzle pressure is marked to construct a pressure control learning sample set, and a control pressure analysis model is obtained by training and learning the pressure control learning sample set; When the preset learning goal is control strategy optimization, the control angle analysis model and the control pressure analysis model are combined, and a post-optimization processing model is provided to comprehensively optimize the angle control analysis results and the pressure control analysis results, including setting optimization target values ​​and optimization iteration rules to obtain a comprehensive optimization analysis model; Configure the target task, obtain the adaptation model from the control angle analysis model, the control pressure analysis model, and the comprehensive optimization analysis model, and generate the defrost control analysis module.

4. The method according to claim 1, characterized in that The monitoring unit collects the information of the evaporator to be defrosted, performs feature analysis on the information of the evaporator to be defrosted, and obtains the distribution of frosting area and frosting thickness, including: Configuring the camera of the monitoring unit to cover the evaporator to collect images of frost on the surface of the evaporator; The coverage, thickness and area distribution of the frost on the evaporator surface are identified, wherein the frost on the evaporator surface image includes multiple shooting angles, and the coverage and area distribution of the frost are identified by the front view angle, and the thickness of the frost is identified by the side view angle; The coverage range, frosting thickness, and frosting area distribution identification are position matched and integrated to establish a corresponding relationship between the coverage range, frosting distribution position, and frosting thickness.

5. The method according to claim 3, characterized in that The frosting area distribution and frosting thickness are input into the defrosting control analysis module for analysis and processing to determine the defrosting control strategy, including: Obtaining a defrost control parameter of a local spiral quick freezer and determining a target task, wherein the target task is used to describe an adjustment target parameter of the defrost control parameter; When the adjustment target parameter is the nozzle angle, the control angle analysis model is configured to generate a defrost control analysis module, and the control angle analysis is performed according to the frosting area distribution and frosting thickness, and the control angle is obtained as the defrost control strategy for output; When the adjustment target parameter is the nozzle pressure, the control pressure analysis model is configured to generate a defrost control analysis module, and the control pressure analysis is performed according to the frosting area distribution and frosting thickness, and the pressure control parameter is obtained as the defrost control strategy for output; When the adjustment target parameters include nozzle angle and nozzle pressure, the comprehensive optimization analysis model is configured to generate a defrost control analysis module, and the control angle and control pressure are optimized according to the frosting area distribution and frosting thickness. According to the preset target value, an evaluation and optimization is performed to find the comprehensive control strategy of control angle and control pressure with the highest target value evaluation, which is output as the defrost control strategy.

6. The method according to claim 5, characterized in that Before inputting the frosting area distribution and frosting thickness into the defrost control analysis module for analysis and processing, the method further includes: According to the frosting area distribution and frosting thickness, the defrosting area is segmented, and the segmented areas include different frosting thicknesses; According to the frosting thickness of the segmented areas, the segmented areas are arranged in positive order, the minimum frosting thickness is used as the basic thickness, set as the basic identifier, the thickness difference between adjacent arranged areas is used as the level compensation value, and the defrosting level identifier of each segmented area is generated, and the level identifier of the segmented area is equal to the basic identifier plus the level compensation value; The numerical value of the defrost level identifier is multiplied by the area size of the segmented area, and then the value is calculated by the proportion of the total amount to convert it into a control adjustment coefficient of the segmented area. The control adjustment coefficient is used to describe the weight proportion value of the control parameter during the defrost control analysis.

7. The method according to claim 6, characterized in that The operational expression of the control adjustment coefficient is: ,in, is the control adjustment coefficient of the i-th segmentation area, The level compensation value of the i-th segmented area represents the difference between the frost thickness and the base thickness of the area. is the area of ​​the ith segmented region, j is the total number of segmented regions, B is the basic identifier, and B=1.

8. The method according to claim 4, characterized in that The identification of frosting thickness by side viewing angle includes: Performing frame recognition on the image captured at the side viewing angle and performing binarization processing; Pixel distribution recognition is performed based on the binarization of the image, and the number is superimposed according to the distribution recognition to determine the frosted area; A reference cross section is set, and the reference cross section is used as a starting point, and an estimate is made based on the frosting area to determine the frosting thickness distribution.

Images