An Internet of Things-based monitoring system for the data of leftover materials of household boards

Abstract

The present invention discloses an Internet of Things-based home furnishing board waste data monitoring system, including: an interference feature acquisition module: collecting electromagnetic interference signal data at different positions and time periods in the workshop, extracting the interference signal features, and constructing an interference feature library based on an optimized clustering operation process; a collection array optimization module: setting up a data collection antenna array, calculating the optimal weighting coefficient through an optimization algorithm, adjusting the emission beam direction of the antenna array based on the optimal weighting coefficient to obtain an optimal collection array, and receiving initial board waste data in real time through the optimal collection array; a waste data optimization module: collecting the initial board waste data based on the optimal collection array, canceling the interference signal of the initial board waste data through the interference feature library in combination with a support vector machine to obtain the optimal board waste data, and this system can more accurately reflect the actual waste situation.

Description

Technical Field

[0001] The present invention relates to the technical field of data monitoring, and in particular to a household board surplus data monitoring system based on the Internet of Things. Background Art

[0002] With the expansion of the production scale of smart home panels, the monitoring and management of panel waste has become a key link to improve production efficiency and reduce costs. At present, the home panel production workshop is full of various metal equipment and electrical equipment, such as large electric saws, electric planers, high-frequency welders, etc. These equipment will generate strong and complex electromagnetic interference during operation. The existing monitoring system mostly uses wireless signals such as Bluetooth and ZigBee for data transmission, which will be greatly affected in this complex electromagnetic environment;

[0003] In the prior art, the wireless signal of the monitoring system is prone to signal interruption and increased bit error rate in the complex electromagnetic environment of the home furnishing board production workshop (various metal equipment and electrical equipment, such as large electric saws, electric planers, high-frequency welders, etc.). Taking the sensor nodes near large metal processing equipment as an example, the signal interruption probability is as high as 30%, which seriously affects the data integrity and will cause inaccurate subsequent residual material data monitoring results. Therefore, a home furnishing board residual material data monitoring system based on the Internet of Things is proposed herein. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention proposes the following technical solutions:

[0005] A household board waste data monitoring system based on the Internet of Things, comprising:

[0006] Interference feature acquisition module: collects electromagnetic interference signal data at different locations and time periods in the workshop, extracts interference signal features, and builds an interference feature library based on an optimized clustering operation process;

[0007] Acquisition array optimization module: set the data acquisition antenna array, calculate the optimal weighting coefficient through the optimization algorithm, adjust the transmission beam direction of the antenna array based on the optimal weighting coefficient, obtain the optimal acquisition array, and receive the initial plate residue data in real time through the optimal acquisition array;

[0008] Residue data optimization module: collects initial sheet material residue data based on the optimal collection array, and cancels the interference signal of the initial sheet material residue data through the interference feature library combined with the support vector machine to obtain the optimal sheet material residue data;

[0009] Residual material data monitoring module: Based on the real-time optimal plate residual material data, the residual material threshold is set. When the residual material data exceeds the residual material threshold, the reminder mechanism is triggered;

[0010] The optimized clustering operation is achieved by combining hierarchical clustering with K-Means clustering;

[0011] The hierarchical clustering calculates the distance of cluster centers by using a distance metric learning model based on deep learning as a clustering metric in the clustering process.

[0012] The interference signal feature acquisition process is as follows:

[0013] Based on the equipment layout, work area division and electromagnetic interference source distribution of the home panel workshop, different locations in the workshop Arrange high-precision electromagnetic interference signal collector, Indicates the position index;

[0014] In different time periods , according to the fixed sampling frequency , continuously collect electromagnetic interference signals Form a large amount of raw signal data set ;

[0015] Set the passband and stopband parameters of the filter to remove interference signals in irrelevant frequency bands and highlight the main target frequency band signals ;

[0016] By using Fourier transform to analyze the main target frequency band signal Perform feature extraction to obtain interference signal characteristics .

[0017] The interference feature library construction process is as follows:

[0018] The interference feature vector As input data for hierarchical clustering;

[0019] In the initial division stage of hierarchical clustering, an agglomerative hierarchical clustering method is used, starting from each feature vector as a separate class. When merging classes, a distance metric learning model based on deep learning is used to calculate the distance of cluster centers to obtain clustering results.

[0020] The results of hierarchical clustering are used as the initial cluster centers of the K-Means clustering algorithm;

[0021] The K-Means clustering algorithm starts with the initial cluster center and enters the iterative calculation process;

[0022] For each category obtained by clustering, traverse the interference signal features in this category , giving a unique identifier Create a database table as an interference signature library .

[0023] The construction process of the distance metric learning model based on deep learning is as follows:

[0024] Construct a deep neural network with the interference signal features as the input;

[0025] Use the contrastive learning loss function to train the network;

[0026] For a given random three feature vectors 、 、 , define the contrastive learning loss function:

[0027]

[0028] where represents the distance metric in the embedding space, is a preset boundary value used to ensure sufficient distance between different class vectors;

[0029] After training, obtain the distance metric learning model based on deep learning.

[0030] The process of determining the optimal acquisition array is as follows:

[0031] Arrange a uniform linear antenna array composed of antenna elements;

[0032] where the antenna element spacing is , and it satisfies , is the signal wavelength. Let the input signal received or transmitted by the th antenna element be ;

[0033] Determine an optimization objective function with the goal of maximizing the signal-to-interference-plus-noise ratio of the received signal;

[0034] Let the desired signal be , the interference signal be , and the output signal of the antenna array is expressed as: , where is the weighting coefficient of the th antenna element;

[0035] The weighting coefficient vector is , where, represents the transpose;

[0036] The formula for maximizing the signal-to-interference-plus-noise ratio is: , where denotes the mathematical expectation, is the component of the desired signal at the th antenna element, is the component of the interference signal at the th antenna element, denotes the noise;

[0037] The particle swarm optimization algorithm is combined with the objective of the optimization objective function to obtain the optimal weighting coefficient ;

[0038] After obtaining the optimal weighting coefficient the pattern function of the antenna array is changed by adjusting the weighting coefficient, so that the transmitting beam direction avoids the strong electromagnetic interference source direction, and the optimal acquisition array is obtained.

[0039] The process of combining the particle swarm optimization algorithm with the objective of the optimization objective function is as follows:

[0040] Randomly generate particles equal to the number of antenna elements The position vector of each particle is and the initial velocity vector is ;

[0041] Record the historical optimal position of each particle and initialize it to its initial position, and the global optimal position of the entire particle swarm is initialized to the initial position of the particle with the largest fitness value;

[0042] Update the position and velocity of the particle. For the position vector of each particle, substitute the position vector into the signal-to-interference-plus-noise ratio calculation formula, and the calculated result is used as the fitness value ;

[0043] Compare the current fitness value of each particle with the fitness value corresponding to its historical optimal position . When , then update to the current position;

[0044] Compare the current fitness values of all particles, find the particle position corresponding to the maximum fitness value. When the fitness value at this position is greater than the fitness value corresponding to the current global optimal position , then update to this particle position;

[0045] When the preset number of iterations is reached or the fitness value of the global optimal position is continuous The change in the current iteration is less than the minimum value and the iteration is stopped. The global optimal position at this time corresponds to the combination of weighted coefficients, which is the optimal weighted coefficient , where denotes the transpose

[0046] The process of obtaining the optimal sheet stock data is as follows

[0047] The initial sheet stock data collected by the optimal acquisition array includes the desired signal , interference signal and noise , which form the initial sheet stock data ;

[0048] For the collected initial sheet stock data , use the support vector machine to match with the interference feature library ;

[0049] According to the matching result of the interference signal, after determining the type of the interference signal, obtain the corresponding interference feature vector frequency , amplitude and phase from the interference feature library Ω, and reconstruct the interference signal according to the formula , where denotes the reconstructed interference signal corresponding to the nth interference feature vector ;

[0050] Input the reconstructed interference signal as the reference signal into the adaptive filter, and the output is close to the interference signal. Subtract the output of the filter from the collected signal to obtain the desired signal after canceling the interference ;

[0051] The signals after canceling the interference of all antenna elements form the optimal sheet stock data set , where is the number of antenna elements , and each number of antenna elements corresponds to a group of desired signals

[0052] The process of using the support vector machine to match with the interference feature library is as follows

[0053] Use the data in the constructed interference feature library Ω to form the training data set , where is the th signal of the The interference feature vectors of the training samples, is the corresponding class label;

[0054] Solve the optimization problem of the support vector machine to obtain the optimal weight vector and bias Construct a classification function ;

[0055] For the feature vectors of the initially collected data of the sheet metal remnants , substitute them into the classification function , where is the sign function, is a function that maps the input feature vectors to a high-dimensional space.

[0056] The present invention has the following beneficial effects:

[0057] In the present invention, firstly, by combining the improved hierarchical clustering and K-Means clustering algorithms, an interference feature library is constructed. A distance metric learning model based on deep learning is introduced in the hierarchical clustering. Compared with traditional clustering methods, the distance metric learning model based on deep learning can analyze the characteristics of electromagnetic interference signals more accurately. Because traditional clustering methods such as Euclidean distance only measure the similarity of feature vectors from the spatial position and cannot consider the complex correlations between features, while the deep learning model can mine complex non-linear relationships. When dealing with interference signals with similar frequencies but different modulation methods, it can accurately distinguish and cluster, and construct a more accurate interference feature library;

[0058] Secondly, by taking the maximization of the signal-to-interference-plus-noise ratio as the goal, calculating the optimal weighting coefficient by using the particle swarm optimization algorithm, and adjusting the transmitting beam direction of the antenna array through the weighting coefficient, it is possible to effectively avoid strong electromagnetic interference sources in a complex electromagnetic environment, enhance the signal coverage of the target monitoring area, and further ensure the stability of the data acquisition process;

[0059] Finally, through the combined decision of the interference feature library and the support vector machine, the interference cancellation is performed on the data collected in real time by the optimal acquisition array. The interference feature library contains rich interference feature information, and the support vector machine can accurately identify the interference type. The combination of the two can effectively remove the interference components in the collected data, so as to obtain the optimal sheet metal remnant data and more accurately reflect the actual remnant situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 FIG. is a system block diagram of a home sheet metal remnant data monitoring system based on the Internet of Things proposed by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0061] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0062] Embodiment: As Figure 1 shown, a home furnishing board waste data monitoring system based on the Internet of Things proposed by the present invention includes:

[0063] Interference feature acquisition module: Collect electromagnetic interference signal data at different positions and time periods in the workshop, extract the interference signal features, and construct an interference feature library based on an optimized clustering operation process;

[0064] Based on the equipment layout, work area division, and electromagnetic interference source distribution in the home furnishing board workshop, at different positions in the workshop ( , , where is the number of acquisition points) arrange high-precision electromagnetic interference signal collectors, representing the position index;

[0065] At different time periods ( , , where is the number of time periods), collect electromagnetic interference signals continuously and stably at a fixed sampling frequency ; ;

[0066] The acquisition time period should cover various working states such as the regular working hours of the workshop, the start and stop moments of equipment, and the switching time periods of different production processes to ensure obtaining comprehensive and representative electromagnetic interference signal data, and finally form a large set of original signal data ;

[0067] According to the main frequency band range of electromagnetic interference in the workshop, set the passband and stopband parameters of the filter to remove interference signals in irrelevant frequency bands and highlight the signals in the main target frequency band ;

[0068] Extract the features of the signals in the main target frequency band by applying the Fourier transform to the signals in the main target frequency band to obtain the interference signal features ;

[0069] Implement an optimized clustering operation by combining improved hierarchical clustering and K-Means clustering;

[0070] Specifically, based on hierarchical clustering, the set of interference feature vectors extracted from electromagnetic interference signals is used as the input data for hierarchical clustering, and each interference feature vector represents a data point;

[0071] In the initial partitioning stage of hierarchical clustering, an agglomerative hierarchical clustering method is adopted. Starting from each feature vector as a separate class, when merging classes, a distance metric learning model based on deep learning is used to calculate the distance between cluster centers;

[0072] The specific process is as follows: Construct a deep neural network, and use the interference signal features as the input. Through network learning, a low-dimensional embedding space is obtained. In this space, the distance between similar interference signal feature vectors is closer, and that between different ones is farther;

[0073] A contrastive learning loss function is used to train the network. For each interference signal feature vector , some vectors similar to it (belonging to the same type of interference) and dissimilar (belonging to different types of interference) are randomly selected as contrast samples;

[0074] During the training process, by adjusting the network parameters, the distance between similar vectors in the embedding space is made as small as possible, and the distance between dissimilar vectors is made as large as possible. For three randomly given feature vectors 、 、 , the contrastive learning loss function is defined as:

[0075]

[0076] where represents the distance metric in the embedding space, and is a preset boundary value used to ensure a sufficient distance interval between vectors of different classes;

[0077] For a given feature vector , some similar vectors belonging to the same type of interference as it , and dissimilar vectors belonging to different types of interference are randomly selected as contrast samples. The reason for doing this is that through these samples with clear similarity or dissimilarity relationships, the model can better learn the differences in interference signals at the feature level, and then perform reasonable distance metrics on them in the embedding space;

[0078] After training, the distance metric learning model based on deep learning is used as the distance metric method for hierarchical clustering to replace the traditional distance metric method (such as Euclidean distance);

[0079] Specifically, traditional distance measurement methods (such as Euclidean distance) only calculate the distance from the spatial position of the data point, which makes it difficult to consider the complex internal connections between features. The distance measurement learning model based on deep learning can mine the complex nonlinear relationship between features by building a deep neural network to learn the feature vectors of interference signals, thereby more accurately measuring the similarity of feature vectors. For example, when processing interference signals with similar frequencies but different modulation methods, the model can learn the impact of modulation method differences on signal similarity, and distinguish different interference signals more accurately than Euclidean distance.

[0080] The results of hierarchical clustering are used as the initial cluster centers of the K-Means clustering algorithm;

[0081] The K-Means clustering algorithm starts with the initial cluster center and enters the iterative calculation process. In each iteration, the Euclidean distance clustering calculation method is replaced by the Mahalanobis distance. By calculating each eigenvector (interference signal feature ) to the Mahalanobis distance of the cluster center , the feature vector (interference signal feature ) is assigned to the category of the nearest cluster center, and then the location of the cluster center is updated according to the new category assignment. This iterative process is repeated until the location of the cluster center no longer changes significantly or the preset number of iterations is reached;

[0082] Specifically, the distance between the feature vector and the cluster center is calculated by replacing the Euclidean distance with the Mahalanobis distance in the clustering process. The traditional Euclidean distance does not consider the correlation between features. When processing a data set such as an electromagnetic interference signal where there may be complex correlations between features, it is easy to cause clustering bias. The Mahalanobis distance can consider the correlation between features and can more accurately measure the similarity between samples.

[0083] For each category obtained by clustering, traverse the interference signal features in this category , based on the interference signal characteristics Get detailed parameters and assign a unique identifier Create a database table as an interference signature library ;

[0084] Among them, the interference feature library The detailed parameters (in database table form) include the following fields: Category ID , Interference signal characteristics And the corresponding collection location and the corresponding time period , establish a composite index based on category identification and collection location;

[0085] Specifically, by combining the improved hierarchical clustering and K-Means clustering as an overall optimized clustering operation process, the advantages of both algorithms are brought into play;

[0086] Hierarchical clustering grasps the overall structure of the data, can avoid falling into local optimal solutions, and makes a preliminary reasonable division of the eigenvectors of electromagnetic interference signals to form a rough clustering framework. The K-Means clustering algorithm then starts from the results of hierarchical clustering and iteratively calculates to locally optimize the clustering results. This fusion method is not a simple superposition of algorithms, but allows the two algorithms to complement each other according to the characteristics of the electromagnetic interference signal features, improving the accuracy and stability of clustering.

[0087] Acquisition array optimization module: Set up a data acquisition antenna array, calculate the optimal weighting coefficient through an optimization algorithm, adjust the transmitting beam direction of the antenna array based on the optimal weighting coefficient to obtain an optimal acquisition array, and receive the initial sheet stock data in real time through the optimal acquisition array;

[0088] Arrange a uniform linear antenna array composed of antenna elements;

[0089] Among them, the antenna element spacing is , and it satisfies ( is the signal wavelength), the th antenna element receives or transmits signals, and its input signal is set as ;

[0090] In order to obtain the optimal weighting coefficient , it is necessary to first determine an optimization objective function. With maximizing the signal-to-interference-plus-noise ratio ( ) of the received signal as the goal, the signal-to-interference-plus-noise ratio reflects the relative magnitude of the desired signal power and the interference and noise power. The higher the value, the better the signal quality;

[0091] Specifically, let the desired signal be , the interference signal be , and the output signal of the antenna array is expressed as: where is the weighting coefficient of the th antenna element;

[0092] The weighting coefficient vector is , where represents the transpose, is the number of antenna elements, and one antenna element corresponds to one weighting coefficient;

[0093] The formula for maximizing the signal-to-interference-plus-noise ratio is: , where represents the mathematical expectation, is the component of the desired signal at the th antenna element, is the component of the interference signal at the th antenna element, represents the noise;

[0094] The objective of the optimization objective function is to maximize the signal-to-interference-plus-noise ratio , so the signal-to-interference-plus-noise ratio maximization calculation formula is the optimization objective function;

[0095] Considering the complexity and real-time variability of the electromagnetic environment in the furniture board workshop, the particle swarm optimization algorithm is combined with the objective of the optimization objective function (maximizing the signal-to-interference-plus-noise ratio ) to obtain the optimal weighting coefficient ;

[0096] Randomly generate particles ( is equal to the number of antenna elements), and the position vector of each particle is , and the initial velocity vector is ;

[0097] Record the historical best position of each particle , initialize it to its initial position, and the global best position of the entire particle swarm is initialized to the initial position of the particle with the largest fitness value;

[0098] Update the position and velocity of the particle, and the formula is expressed as:

[0099] Velocity update:

[0100] Where: represents the velocity at the next position , ω is the inertia weight, which is set to a larger value (such as 0.9) at the beginning of the iteration for global search and gradually decreases (such as to 0.4) as the iteration progresses for local search;

[0101] and are learning factors, usually taking the value of 1.5;

[0102] and are random numbers between 0 and 1;

[0103] Position update:

[0104] = +

[0105] For the position vector of each particle , substitute it into the signal-to-interference-plus-noise ratio calculation formula, and take the calculation result as the fitness value ;

[0106] Compare the current fitness value of each particle with the fitness value corresponding to its historical optimal position . When , update to the current position;

[0107] Compare the current fitness values of all particles, find the particle position corresponding to the maximum fitness value. When the fitness value at this position is greater than the fitness value corresponding to the current global optimal position , update to this particle position;

[0108] When the preset number of iterations is reached or the fitness value of the global optimal position changes less than the minimum value in consecutive iterations, stop the iteration. At this time, the combination of weighted coefficients corresponding to the global optimal position is the optimal weighted coefficient , where represents the transpose, is the number of weighted coefficients ( is equal to the number of antenna elements);

[0109] After obtaining the optimal weighted coefficient , calculate the radiation characteristics of the antenna array through the formula:

[0110]

[0111] where is the wave number, is the radiation direction angle of the antenna array;

[0112] By adjusting the weighted coefficients, change the pattern function of the antenna array to avoid the direction of strong electromagnetic interference sources for the transmitting beam direction, while enhancing the signal coverage of the target monitoring area to obtain the optimal acquisition array;

[0113] For example, devices such as phase shifters and attenuators can be used to achieve weighted adjustment of the signals of each antenna element;

[0114] Specifically, the superposition of the signals of each antenna element in the antenna array will form a specific radiation pattern, and the weighted coefficient It will affect the amplitude and phase of the antenna unit signal, and further change the radiation pattern function of the antenna array. Therefore, by adjusting the weighting coefficient, the direction of the transmitting beam can be changed according to the specified requirements to avoid interference sources and enhance the signal coverage in the target area;

[0115] After the antenna array is put into use, the signal data of the initial sheet stock data is received in real time and transmitted to the sheet stock data optimization module.

[0116] Sheet stock data optimization module: Based on the optimal acquisition array, the initial sheet stock data is collected, and the interference signal of the initial sheet stock data is cancelled by combining the interference feature library with the support vector machine to obtain the optimal sheet stock data;

[0117] The initial sheet stock data collected by the optimal acquisition array includes the desired signal (optimal sheet stock data) , interference signal and noise , which constitute the initial sheet stock data ;

[0118] For the collected initial sheet stock data , use the support vector machine to match it with the interference feature library ;

[0119] Using the data in the constructed interference feature library Ω, a training data set is formed , where is the interference feature vector of the th training sample of the th antenna unit signal, is the corresponding class label (representing different interference types);

[0120] By solving the optimization problem of the support vector machine , the constraint conditions are: ;

[0121] Among them, is the weight vector, is the bias, is the slack variable, is the penalty parameter, is the function that maps the input feature vector to a high-dimensional space, and the optimal and are obtained, so as to determine the classification hyperplane, is the th interference feature;

[0122] Specifically, in industrial environments such as actual home furnishing board production workshops, there are complex electromagnetic interferences that can affect the collected data of board leftovers. By collecting the initial data containing interferences and using a support vector machine and an interference feature library to match and classify the interference signals, and then taking corresponding processing measures, the core principle is to classify different types of data by finding the optimal hyperplane;

[0123] By solving the above optimization problem, the optimal weight vector is obtained and the bias , obtaining the weight vector and the bias After that, substitute them into the expression of the hyperplane , and then combine it with the sign function , and it constitutes the classification function;

[0124] For the feature vector of the initially collected board leftover data, substitute it into the classification function , where is the sign function;

[0125] Judge its interference type through the classification function, find the element in the interference feature library Ω that matches it, and complete the classification and matching of the interference signal;

[0126] Through the result output by the sign function , the collected feature vectors can be classified into different interference types. For example, it is stipulated that +1 represents high-frequency electromagnetic interference and -1 represents low-frequency electromagnetic interference. After obtaining the judgment result of the interference type, find the element in the interference feature library Ω that matches it, and then obtain the corresponding detailed parameters of the interference feature vector ;

[0127] According to the interference signal matching result, after determining the interference signal type, obtain the corresponding detailed parameters of the interference feature vector from the interference feature library Ω;

[0128] The detailed parameters further include frequency , amplitude and phase , and reconstruct the interference signal according to the formula , where represents the reconstructed interference signal corresponding to the nth interference feature vector;

[0129] Specifically, the feature vector of the initially collected board leftover data is collectedAfter that, substitute it into the classification function of the support vector machine, calculate the relationship between the feature vector and the optimal hyperplane through the classification function, obtain the class label, and thus judge the type of interference to which it belongs, which can effectively distinguish different types of interference signals. Moreover, after determining the type of interference signal, obtain the corresponding interference feature vector from the interference feature library Ω for the detailed parameters (frequency , amplitude and phase ). Usually, the interference signal is approximately described by a sine wave. According to the formula reconstruct the interference signal, which provides an accurate reference signal for subsequent interference cancellation;

[0130] Input the reconstructed interference signal as a reference signal into the adaptive filter, continuously adjust the filter weight coefficient, so that the filter output is close to the interference signal. Then, subtract the output of the filter from the collected signal to obtain the desired signal after interference cancellation ;

[0131] The signals after interference cancellation of all antenna elements constitute the optimal sheet stock data set , where is the number of antenna elements , and the number of each antenna element corresponds to a group of desired signals;

[0132] Specifically, input the reconstructed interference signal as a reference signal into the adaptive filter. The adaptive filter continuously adjusts the weight coefficient according to the input signal and the reference signal, aiming to make the filter output as close as possible to the interference signal. Since the collected signal contains the desired signal and the interference signal, when the filter output is close to the interference signal, subtract the filter output from the collected signal to cancel the interference signal, thereby obtaining the desired signal after interference cancellation. Then, after performing the above interference cancellation operation on the signals collected by each antenna element, integrate the signals after interference cancellation of all antenna elements to form the optimal sheet stock data set .

[0133] Surplus material data monitoring module: Based on the real-time optimal sheet stock data, set the surplus material threshold. When the surplus material data exceeds the surplus material threshold, trigger the reminder mechanism;

[0134] After obtaining the optimal sheet stock data set , traverse the surplus material quantities of all surplus material threshold types in the surplus material data set , and set the corresponding alarm threshold. When a certain surplus material data exceeds the threshold, the system issues a reminder;

[0135] Let a certain type of board ( ) have a minimum quantity threshold of , and a maximum quantity threshold of ;

[0136] Specifically, for a certain type of board the maximum threshold and the minimum threshold are obtained by referring to historical production data and statistically calculating the average quantity of this type of board required for each batch of production under normal production conditions;

[0137] Compare the extracted real-time data with the set thresholds to determine whether a reminder needs to be triggered;

[0138] If , then trigger a low quantity reminder;

[0139] If , then trigger a high quantity reminder.

[0140] In the application, several formulas involved are calculated by taking their numerical values after dimensionless, and the establishment of the formulas is obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so no more details will be elaborated here.

[0141] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution.

[0142] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A home furnishing board waste data monitoring system based on the Internet of Things, characterized in that Including: Interference feature acquisition module: Collect electromagnetic interference signal data at different positions and time periods in the workshop, extract the interference signal feature vectors, and construct an interference feature library based on an optimized clustering operation process; The process of constructing the interference feature library is as follows: Use the interference signal feature vector as the input data for hierarchical clustering; In the preliminary division stage of hierarchical clustering, the agglomerative hierarchical clustering method is used. Starting from each feature vector as a separate class, when merging classes, a distance metric learning model based on deep learning is used to calculate the distance between cluster centers to obtain the clustering result; Take the result obtained by hierarchical clustering as the initial cluster center of the K-Means clustering algorithm; The K-Means clustering algorithm starts from the initial cluster center and enters the iterative calculation process. In each iteration, Replace the Euclidean distance clustering algorithm with the Mahalanobis distance, and calculate the Mahalanobis distance from the feature vector of each interference signal to the clustering center ; The K-Means clustering algorithm starts from the initial cluster center and enters the iterative calculation process; For each category obtained by clustering, traverse the interference signal feature vectors in this category , and assign a unique identifier Create a database table as the interference feature library ; Acquisition array optimization module: Set up a data acquisition antenna array, calculate the optimal weighting coefficient through an optimization algorithm, adjust the transmitting beam direction of the antenna array based on the optimal weighting coefficient to obtain an optimal acquisition array, and receive the initial sheet stock data in real time through the optimal acquisition array; Remnant data optimization module: Based on the optimal acquisition array, collect the initial sheet stock data, and use the interference feature library combined with the support vector machine to cancel the interference signal in the initial sheet stock data to obtain the optimal sheet stock data; Remnant data monitoring module: Based on the real-time optimal sheet stock data, set a remnant threshold. When the remnant data exceeds the remnant threshold, trigger a reminder mechanism; The optimized clustering operation is realized by combining hierarchical clustering and K-Means clustering; In the hierarchical clustering, a distance metric learning model based on deep learning is used as the clustering metric method to calculate the distance between cluster centers during the clustering process; The process of constructing the distance metric learning model based on deep learning is as follows: Construct a deep neural network with the interference signal feature vector as the input; A low-dimensional embedding space is obtained through network learning, and the network is trained using a contrastive learning loss function. For each interference signal feature vector , vectors belonging to the same type of interference and different types of interference are randomly selected as contrast samples; For a given set of three random feature vectors , , , a contrastive learning loss function is defined as follows: ; wherein represents a distance metric in the embedding space, is a preset boundary value for ensuring a sufficient distance interval between vectors of different classes; After training, a distance metric learning model based on deep learning is obtained; The process of determining the optimal acquisition array is as follows: Arrange a uniform linear antenna array composed of antenna elements; Among them, the antenna element spacing is , and it satisfies , is the signal wavelength. Let the input signal received or transmitted by the -th antenna element be ; Determine an optimization objective function with the goal of maximizing the signal-to-interference-plus-noise ratio (SINR) of the received signal as the objective; Let the desired signal be , and the interference signal be . The output signal of the antenna array is expressed as:​ , where is the weighting coefficient of the th antenna element; The weighted coefficient vector is , where denotes transpose. Signal-to-interference-plus-noise ratio The maximization calculation formula is as follows: , where represents the mathematical expectation, is the component of the desired signal at the th antenna element, is the component of the interference signal at the th antenna element, represents noise; The particle swarm optimization algorithm is combined with the objective of the optimization objective function to obtain the optimal weighting coefficient ; Obtain the optimal weighting coefficient After that, by adjusting the weighting coefficient to change the pattern function of the antenna array, the direction of the transmitting beam is made to avoid the direction of strong electromagnetic interference sources, and an optimal acquisition array is obtained.

2. The home furnishing board waste data monitoring system based on the Internet of Things according to claim 1, wherein, The process of obtaining the interference signal feature vector is as follows: Based on the equipment layout, work area division, and distribution of electromagnetic interference sources in the home furnishing board workshop, at different positions in the workshop arrange high-precision electromagnetic interference signal collectors, indicating the position index; At different time periods ,at a fixed sampling frequency ,continuously collect electromagnetic interference signals to form a large collection of original signal data ; Set the passband and stopband parameters of the filter to remove the interference signals in the irrelevant frequency bands and highlight the signals in the main target frequency band ; By applying Fourier transform to the signals in the main target frequency band feature extraction is performed to obtain the feature vector of the interference signal .

3. The data monitoring system for leftover materials of household boards based on the Internet of Things according to claim 1, characterized in that, The process of combining the particle swarm optimization algorithm with the objective of the optimization objective function is as follows: Randomly generate particles equal to the number of antenna elements The position vector of each particle is , and the initial velocity vector is ; Record the historical optimal position of each particle , initialize it to its initial position, and the global optimal position of the entire particle swarm Initialize it to the initial position of the particle with the maximum fitness value; Update the positions and velocities of the particles, and for the position vector of each particle , substitute the position vector into the calculation formula of the signal-to-interference-plus-noise ratio , and use the calculated result as the fitness value ; Compare the current fitness value of each particle with its historical optimal position corresponding fitness value. When , update to the current position; Compare the current fitness values of all particles, find the particle position corresponding to the maximum fitness value, and when the fitness value at this position is greater than the fitness value of the current global optimal position corresponding to, then update to this particle position; When the preset number of iterations is reached or the fitness value of the global optimal position remains unchanged within consecutive iterations by less than a minimum value stop the iteration. The combination of weighted coefficients corresponding to the global optimal position at this time is the optimal weighted coefficient, where denotes the transpose. ​ 4. The home furnishing board waste data monitoring system based on the Internet of Things according to claim 1, characterized in that, The process of obtaining the optimal sheet stock data is as follows: The initial sheet stock surplus data collected by the optimal acquisition array includes desired signals , interference signals and noise , which constitute the initial sheet stock surplus data ; For the initially collected data of the leftover sheet materials , use the support vector machine to match with the interference feature library ; After determining the interference signal type based on the interference signal matching result, obtain the corresponding interference signal feature vector from the interference feature library Ω frequency amplitude and phase According to the formula reconstruct the interference signal where represents the reconstructed interference signal corresponding to the nth interference signal feature vector; The reconstructed interference signal is input into the adaptive filter as a reference signal, and the output is close to the interference signal. The collected signal is subtracted from the output of the filter to obtain the desired signal after canceling the interference ; The signals of all antenna units after interference cancellation constitute an optimal sheet stock data set , where is the number of antenna units , and the number of each antenna unit corresponds to a group of desired signals.

5. The data monitoring system for the remaining materials of household boards based on the Internet of Things according to claim 1, characterized in that, Using a Support Vector Machine and an Interference Feature Library The process of performing the matching is as follows: Form a training data set using the data in the constructed interference feature library Ω , where is the interference signal feature vector of the th training sample of the th antenna unit signal, and is the corresponding class label; Solve the optimization problem of the support vector machine to obtain the optimal weight vector and bias Construct a classification function ; The feature vector of the initially collected data of the leftover sheet materials , is substituted into the classification function , where is the sign function, is a function that maps the input feature vector to a high-dimensional space.

Images