Intelligent monitoring method and system for environment and production information in cotton processing process based on Internet of Things

Abstract

The invention provides an intelligent monitoring method and system for environment and production information in the cotton processing process based on the Internet of Things, and belongs to the technical field of intelligent monitoring of the cotton ginning process, and the method comprises the steps: obtaining the data of the cotton ginning process; performing ginning state monitoring processing on the ginning process data to obtain monitoring parameters reflecting equipment operation and production states in the ginning process; performing abnormal mode recognition processing on the embossing process data according to the monitoring parameters, and separating out embossing process abnormal parameters which are different from normal process states; performing parameter positioning processing on the abnormal parameters in the embossing process, and determining an embossing quality evaluation value corresponding to the abnormal parameters; and generating an embossing quality evaluation result according to the embossing quality evaluation value, and performing abnormal process prompting based on the evaluation result. According to the invention, cotton ginning production monitoring is realized, abnormal processes are rapidly positioned, and fault response time is shortened; the decision-making difficulty is reduced through visual display, and the production efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring technology for cotton ginning processes, specifically relating to an intelligent monitoring method and system for environmental and production information during cotton processing based on the Internet of Things. Background Technology

[0002] Cotton processing is a crucial part of the textile industry. Its core processes include ginning, cleaning, drying, lint cleaning, and packaging, involving multi-dimensional interactions between equipment operation, environmental parameters, and cotton quality.

[0003] The cotton ginning process requires effective monitoring to ensure quality. While related technologies can identify abnormal parameters, they cannot accurately assess their impact on the final product quality. For example, the specific impact of a 5°C fluctuation in drying temperature on fiber strength lacks a quantitative model, leading to delayed quality control. Related technologies use fixed weights for all parameters, ignoring the time-varying nature of parameter fluctuations and their impact on quality. For instance, during equipment startup, the impact of motor current fluctuations on quality is far greater than during stable operation, but fixed-weight models cannot reflect this difference, resulting in monitoring failing to meet process requirements. Furthermore, when new anomalies occur, manual re-analysis is required, leading to long response times. For example, a factory experienced frequent quality problems during the commissioning phase after introducing a new ginning machine due to the lack of historical data for reference. Summary of the Invention

[0004] This invention provides an intelligent monitoring method for environmental and production information during cotton processing based on the Internet of Things, enabling monitoring of cotton ginning production, rapid location of abnormal processes, and shortening fault response time; visual display reduces decision-making difficulty and improves production efficiency.

[0005] The methods include: Step S101: Obtain ginning process data, which includes multiple preset ginning process parameters; Step S102: Perform ginning status monitoring processing on the ginning process data to obtain monitoring parameters reflecting the equipment operation and production status during the ginning process; Step S103: Perform abnormal pattern recognition processing on the ginning process data according to the monitoring parameters, and separate out the abnormal parameters of the ginning process that differ from the normal process state; Step S104: Perform parameter location processing on the abnormal parameters of the ginning process to determine the ginning quality assessment value corresponding to the abnormal parameters; Step S105: Generate ginning quality assessment results based on the ginning quality assessment values, and provide abnormal process prompts based on the assessment results.

[0006] Preferably, step S102 specifically includes: Step S1021: Perform data cleaning and feature extraction on the ginning process parameters, including removing duplicate data, filling in missing values, and extracting time-domain and frequency-domain features from equipment operating parameters and temperature, humidity, and dust concentration. Step S1022: Input the preprocessed process parameters into the fusion neural network model, evaluate the process status through the weight matrix trained by historical abnormal samples, and output the process risk level including the abnormal probability. Step S1023: Construct a conditional probability model of equipment status and fault type based on Bayesian network, and calculate the posterior probability of fault for each process by combining real-time monitored parameters such as current and vibration frequency, and locate potential abnormal processes. Step S1024: Select parameters with risk coefficients higher than the threshold from the preset indicator library and generate a set of monitoring parameters including real-time monitoring frequency and early warning threshold.

[0007] Preferably, step S1023 specifically includes: Step S10231: Based on the process information of the cotton ginning equipment, define network nodes as equipment status parameters and fault types; draw directed edges of the network using the experience of process experts to represent the dependencies between parameters; Step S10232: Using historical process data, calculate the conditional probability of each node under the state of its parent node using the maximum likelihood estimation method; Step S10233: Input the currently monitored equipment status parameters as evidence into the network, and use the variable elimination algorithm to perform probabilistic reasoning; starting from the root node, update the posterior probability of each node layer by layer to obtain the joint posterior probability of each fault type; Step S10234: Set a fault probability threshold, filter out fault types with posterior probabilities exceeding the threshold; map the fault types to the corresponding processes, sort the processes by risk according to the posterior probability values, and output a list of potential abnormal processes.

[0008] Preferably, step S103 specifically includes: Step S1031: Use a convolutional layer module to extract feature data reflecting the data distribution characteristics from the cotton ginning process data. By setting the maximum number of internal nodes, the maximum number of leaf nodes, and the maximum sample radius threshold parameters of the clustering feature tree, the sample points are sequentially allocated to construct a hierarchical clustering feature tree according to the relationship between the distance between sample points and the above parameters. Step S1032: Start hierarchical node splitting from the root node of the clustering feature tree. Take the root node as the first layer of split data. When the number of samples in the split data of this layer exceeds the maximum number of internal nodes or the maximum distance between samples exceeds the maximum sample radius threshold, select the two farthest sample points as new child nodes, and redistribute the original split data according to the distance to form the next layer of split data. Repeat this process until each level of node satisfies the condition that the number of samples does not exceed the maximum number of internal nodes and the maximum distance between samples does not exceed the maximum sample radius threshold, and obtain the split data of each level. Step S1033: Based on the monitoring parameters including the preset threshold or the bottom 20% baseline, outlier screening is performed on the split data of each level; the values ​​of the indicators corresponding to the monitoring parameters in the split data are extracted, and data points that are lower than the preset threshold or the bottom 20% baseline are marked. If the data still contains multiple dimensions after screening, the level splitting process is performed again until the smallest dimension of the ginning process outlier parameter is obtained.

[0009] Preferably, step S104 specifically includes: Step S1041: Divide the abnormal parameters of the cotton ginning process into multi-dimensional grid cells according to the spatial distribution of the equipment and the time of collection. Calculate the spatiotemporal influence factor of the abnormal parameters in each grid cell. The influence factor is determined by multiplying the degree of deviation of the abnormal parameters from the threshold with the correlation weight of the equipment to the quality in the grid cell. Step S1042: Use a Gaussian mixture model to fit the probability density of the abnormal parameter combinations in each grid cell, extract the feature parameters that can characterize the quality impact, and output the quantitative impact value of the abnormal parameter combinations on the ginning quality. Step S1043: Adjust the contribution weight of each parameter to the quality assessment based on the time series fluctuation entropy value of the abnormal parameters; Step S1044: Perform K-nearest neighbor matching between the feature vector of the current abnormal parameter combination and the abnormal-quality correlation data recorded in the historical case library, and calculate the ginning quality assessment value corresponding to the current abnormal parameter by weighting the quality assessment results of the matched cases.

[0010] Preferably, step S1041 specifically includes: Step S10411: Construct a spatiotemporal grid positioning model and divide it into multi-dimensional grid cells; based on the physical location coordinates and time series segmentation of the ginning equipment, map the abnormal parameters of the ginning process to three-dimensional grid cells; adjust the grid cell size according to the equipment distribution density and process cycle. Step S10412: Calculate the spatiotemporal deviation of the abnormal parameters within the grid cell; for each abnormal parameter within the grid cell, calculate its positive / negative deviation and deviation amount relative to the preset threshold, where deviation amount = actual value - threshold; combine the time series trend to calculate the comprehensive deviation using a weighted method. Step S10413: Based on historical process data, the correlation between abnormal parameters and quality parameters is analyzed by linear regression, and the absolute value of the regression coefficient is extracted as the basic weight; the basic weight is corrected by combining process knowledge to obtain the correlation weight of equipment parameters on quality. Step S10414: Multiply the comprehensive deviation of the abnormal parameters within the grid cell by the correlation weight of the corresponding device to obtain the preliminary spatiotemporal influence factor; truncate and correct the influence factors that exceed the mean ± 3 times the standard deviation to finally obtain the stable spatiotemporal influence factor.

[0011] Preferably, step S1042 specifically includes: Step S10421: Standardize the abnormal parameter combinations within the mesh cells; Step S10422: Based on the preprocessed abnormal parameter combination data, the expectation-maximization algorithm is used to estimate the number of components of the GMM, the mean vector of each component, the covariance matrix and the mixing weights; the model output is the probability density distribution of the abnormal parameter combination. Step S10423: Extract three types of feature parameters from the trained GMM: ① Pattern center, representing the typical state of abnormal parameter combinations; ② Pattern diffusion, which reflects the degree of dispersion of anomalous parameter combinations; ③ Pattern probability, representing the frequency of occurrence of the abnormal pattern in historical data; Step S10424: Weighted fusion of model center, diffusion and model probability to calculate the quantitative quality impact value of the abnormal parameter combination; verify the rationality of the impact value through historical data, and calibrate the parameters that deviate from the verification results.

[0012] Preferably, step S1043 specifically includes: Step S10431: For the time series of abnormal parameters, the time domain variation coefficient, frequency domain energy ratio and nonlinear complexity index are calculated simultaneously to quantify the fluctuation characteristics of the parameters from three dimensions: fluctuation amplitude, frequency components and change patterns. Step S10432: Establish the correlation strength matrix between abnormal parameters and ginning quality indicators, and determine the direct impact weight and indirect transmission weight of each parameter on quality indicators such as fiber strength and impurity content through historical fault tree analysis. Step S10433: With the goal of minimizing the quality assessment error, the initial entropy weight method weights are globally optimized using a genetic algorithm. Under the constraints of parameter fluctuation characteristics and correlation strength matrix, the contribution weights of each parameter are iteratively adjusted. Step S10434: Based on the real-time monitored seed cotton moisture content and equipment load rate parameters, the weight coefficients are corrected through fuzzy logic reasoning to make the weight allocation adapt to the differences in the impact of parameters on quality under different production conditions.

[0013] Preferably, step S1044 specifically includes: Step S10441: Perform feature alignment on the abnormal parameter combinations within the current grid cell to generate a standardized feature vector consistent with the format of the historical case library; Step S10442: Store historical anomaly-quality correlation data in a three-level classification system: equipment type, anomaly mode, and quality impact level, and generate a unique index code for each case; at the same time, establish an inverted index table; Step S10443: Use Euclidean distance as a similarity measure and set the neighborhood size k; select the k closest similar cases to the current feature vector from the historical case library; Step S10444: Based on the quality assessment results of the matched cases, and combining the similarity between the case and the current case and the timeliness of the case, calculate the weighted evaluation value vˉ=∑(si×wt×vi) / ∑(si×wt), where vi is the evaluation value of the i-th matched case and wt is the timeliness weight; the final ginning quality assessment value is the standardized result of vˉ.

[0014] According to another embodiment of this application, an intelligent monitoring system for environmental and production information during cotton processing based on the Internet of Things is provided. The system includes: The data acquisition module is used to acquire ginning process data, which includes multiple preset ginning process parameters. The status monitoring module is used to perform ginning process data monitoring and processing to obtain monitoring parameters reflecting the equipment operation and production status during the ginning process; An anomaly identification module is used to perform anomaly pattern recognition processing on the ginning process data based on the monitoring parameters, and to separate out the abnormal parameters of the ginning process that differ from the normal process state. The status assessment module is used to perform parameter location processing on abnormal parameters in the ginning process and determine the ginning quality assessment value corresponding to the abnormal parameters. The result output module is used to generate ginning quality assessment results based on the ginning quality assessment values, and to provide abnormal process prompts based on the assessment results.

[0015] As can be seen from the above technical solutions, the present invention has the following advantages: This application provides an intelligent monitoring method for environmental and production information during cotton processing based on the Internet of Things (IoT). This method monitors and processes ginning process data to obtain monitoring parameters reflecting equipment operation and production status during the ginning process. It transforms complex collected data into meaningful monitoring parameters, focusing on key risk points in the ginning process. Based on these monitoring parameters, it performs anomaly pattern recognition processing on the ginning process data, separating abnormal parameters that differ from normal process conditions. This accurately identifies anomalies from large amounts of data, separating abnormal parameters from normal process data, narrowing the scope of data requiring subsequent attention and processing, and improving the efficiency of anomaly handling. It determines the ginning quality assessment value corresponding to the abnormal parameters. By locating the parameters, it clarifies the specific impact of the anomaly on ginning quality, presenting the severity of the anomaly in a quantitative manner. Based on the assessment results, it provides anomaly process alerts. These alerts are presented intuitively to relevant personnel, facilitating rapid response and handling.

[0016] This method integrates LSTM and Bayesian networks, training the model using historical anomaly samples to learn normal and anomaly pattern features, achieving adaptive anomaly detection. Hierarchical clustering feature trees adjust the classification granularity, accurately identifying anomaly patterns under different operating conditions. Anomaly detection accuracy is improved, and the false alarm rate is reduced. A spatiotemporal impact factor model is constructed to quantify the product of the deviation degree of anomaly parameters and their correlation weight with equipment quality. A Gaussian mixture model is used to fit the probability density of anomaly parameter combinations, extracting key quality feature parameters such as fiber strength attenuation rate. This enables quantitative assessment of the impact of anomaly parameters on quality, providing a basis for precise quality control. Weights are adjusted based on the time series fluctuation entropy value of parameters, with higher weights for parameters with more disordered fluctuations. In scenarios with rapid changes in equipment status, quality assessment accuracy is improved, effectively capturing quality risks in the production process. A K-nearest neighbor case matching mechanism is established to quickly compare the current anomaly feature vector with a historical case database, drawing on the experience of handling similar cases. This enhances the system's ability to handle unknown anomalies. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of an intelligent monitoring method for environmental and production information during cotton processing based on the Internet of Things. Figure 2 This is a schematic diagram of an intelligent monitoring system for environmental and production information during cotton processing based on the Internet of Things. Detailed Implementation

[0019] The specific process of the cotton ginning process involved in this application is as follows: seed cotton purchase → pretreatment → cotton ginning → lint cleaning → packaging → by-product processing.

[0020] Specifically, the seed cotton pretreatment process involves unloading and feeding, with seed cotton being drawn into the storage silo through pipelines. A spiked roller cleaning machine removes large impurities such as stones and metal. An inclined mesh screen separates branches, leaves, and bolls. The drying process can be initiated when the moisture content is >10%. The hot air temperature is controlled at ≤120℃ to prevent fiber damage. Target moisture content: 6%-8%. A toothed roller cleaner removes fine impurities. A magnetic separator adsorbs residual metal.

[0021] The cotton ginning process using a saw gin is as follows: 1. Seed cotton is pulled through the gaps between the ribs by the saw teeth. 2. Fibers separate from the cotton seeds; the cotton seed diameter is greater than the rib spacing. 3. A brush roller peels the lint off the saw teeth. Lint separation is based on airflow conveying the lint to a dust collection cage.

[0022] The cotton cleaning process uses a core-type cotton cleaning machine to remove short fibers and impurities. A spark detector prevents sparks from entering the next process. The humidification system adjusts the cotton moisture regain to 6.5%-7.5% to prevent breakage during packing.

[0023] The packing and storage process uses a hydraulic baler to control the compression density at 400-450 kg / m³. Galvanized steel wire / polyester strapping is used for binding. Each bale of lint is labeled with a barcode for traceability. By-product processing involves cottonseed utilization, such as sending it to an oil mill for seed production. Short lint recycling is based on extraction using a delinting machine.

[0024] The cotton ginning process is based on a saw-tooth gin. Seed cotton is evenly fed into the gin by a feeding mechanism. The saw-tooth rollers (600-800 rpm) rotate at high speed, and the saw teeth hook the fibers, pulling the seed cotton towards the ribs. The gaps between the ribs (2-3 mm) are smaller than the diameter of the cotton seeds (5-8 mm), so the cotton seeds are blocked, while the fibers are carried out by the saw teeth through the gaps. The brush rollers (1000-1200 rpm) are in close contact with the surface of the saw teeth, and the frictional force peels the lint off the saw teeth, causing it to fall into the cotton collection pipe.

[0025] The cotton lint separation process utilizes a negative pressure airflow of 15-20 kPa to transport the stripped cotton lint through pipes to a cotton collection dust cage. The dust cage has a dense mesh (0.5-1 mm in diameter) on its surface. The cotton lint is adsorbed onto the surface of the dust cage, while the air is discharged through the mesh, thus achieving air-cotton separation.

[0026] The cotton cleaning process utilizes a centrifugal cotton cleaning machine. The cotton enters a rotating drum (800-1000 rpm), where centrifugal force throws short fibers (<16mm in length), leaf fragments, and other impurities towards the mesh on the inner wall of the machine casing for discharge. Simultaneously, the card cloth on the drum combs the cotton, reducing fiber entanglement. A spark detector is installed in the conveying pipeline, using an infrared sensor to detect sparks (such as those generated by friction) in the cotton. If a spark is detected, the spray system is immediately activated or the machine is stopped to prevent fire hazards. The humidification system uses steam spray or ultrasonic atomization devices to spray fine water mist onto the cotton, controlling the moisture regain to 6.5%-7.5%. After humidification, the cotton fibers become more flexible and less prone to breakage during packaging, while also preventing static electricity caused by excessive dryness. The packing and storage process is based on a hydraulic baler. The cotton lint is fed into the baling box by a conveyor, and the hydraulic system (pressure 10-15MPa) pushes a piston to compress the cotton to a density of 400-450kg / m³, forming a standard bale (1400mm long × 700mm wide × 600mm high, approximately 227kg / bale). The bale is then secured 4-6 times along the longitudinal and transverse directions with galvanized steel wire (2-3mm diameter) or polyester tape to ensure a sturdy shape and facilitate transportation and storage.

[0027] In this system embodiment, sensors, PLC controllers, and edge computing devices are deployed at each stage to collect environmental, equipment, and production data in real time, enabling quality traceability, fault early warning, and efficiency optimization. The following are the key data to be monitored at each step: The data that needs to be monitored in the seed cotton pretreatment process are: Dust concentration in the unloading area (to avoid dust pollution, ≤8mg / m³); hot air temperature in the drying area (≤120℃); surface temperature of seed cotton (≤50℃, to prevent scorching); noise in the cleaning area (≤85dB, to protect workers' hearing).

[0028] Production data: Cotton unloading feed rate (tons / hour, monitored capacity); impurity separation rate of spiked roller / grid screen (large impurity removal rate ≥95%, small impurity removal rate ≥85%); moisture content before and after drying (initial moisture content, 6%-8% after drying); number of metal detections by magnetic separator (times / hour, abnormal alarm).

[0029] The data that needs to be monitored in the ginning process are: Saw-tooth gin speed (500-800 rpm, ±5% deviation alarm); linear speed ratio of brush roller to saw-tooth roller (typically 1.05-1.1, ensuring effective stripping); ginning machine current (reflects load, abnormal increase may be due to cotton seed blockage); hourly output (±10% deviation from design value warning); lint percentage (online detection, ±2% deviation requires checking ginning gap); cotton seed impurity content (≤1.5%, reflecting ginning separation effect).

[0030] The data that needs to be monitored in the cotton cleaning process are: dust concentration in the cleaning area (≤5mg / m³, to prevent explosion risk); temperature of the spark detector (monitoring the high-temperature point, >80℃ triggers an alarm); short fiber removal rate of the centrifugal cleaner (≥90%); moisture regain before and after conditioning (target 6.5%-7.5%, deviation ±0.5% requires adjustment of humidification); residual impurities in the cotton (≤1.2%, national standard requirement).

[0031] The data that needs to be monitored during the packaging and storage process includes: hydraulic baler pressure (20-30MPa, fluctuation ±5% alarm); bale weight (227kg±5kg, exceeding the standard requires calibration of the weighing system); bundling quality (steel wire tension ≥500N, breakage alarm); lint density after packaging (400-450kg / m³, online detection); completeness of traceability information (QR code clarity, data matching); storage environment: temperature and humidity (temperature ≤30℃, humidity ≤65%, mildew prevention); and inventory quantity (real-time inventory check).

[0032] The following describes in detail the intelligent monitoring method for environmental and production information during cotton processing based on the Internet of Things (IoT). Specific details, such as particular system structures and technologies, are presented for illustrative purposes and not for limitation, to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0033] It should be understood that, when used in this specification, the term "comprising" indicates the presence of the described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0034] It should be understood that "one or more" as mentioned in this application refers to one, two, or more, and "multiple" as mentioned in this application refers to two or more. In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone.

[0035] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figure 1 The diagram shows a flowchart of a method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things (IoT) in a specific embodiment. The method includes: Step S101: Obtain ginning process data, which includes multiple preset ginning process parameters.

[0038] This embodiment uses temperature sensors, current sensors, and near-infrared spectrometers deployed at key nodes of the cotton processing production line to collect ginning process data in real time. The data includes equipment operating motor current, speed, vibration frequency, temperature, humidity, dust concentration, cotton quality parameters such as fiber length, strength, impurity content, process start / end time, and material conveying flow rate. This data is transmitted to a central server for storage via 5G or industrial Ethernet.

[0039] Step S102: Perform ginning status monitoring processing on the ginning process data to obtain monitoring parameters that reflect the equipment operation and production status during the ginning process.

[0040] This embodiment performs multi-layer processing on the collected process data. Root mean square (RMS), peak value, power spectral density, and process features such as lint percentage and moisture regain are extracted through feature engineering. Then, the preprocessed data is input into a neural network model fusing LSTM and random forest, combined with a weight matrix trained using historical anomaly samples, to output the process anomaly probability. Finally, a conditional probability model of equipment status and faults is constructed based on a Bayesian network to calculate the posterior probability of faults in each process, and parameters with risk coefficients exceeding thresholds are selected as monitoring parameters.

[0041] In this way, by integrating machine learning and probabilistic reasoning methods, LSTM captures the temporal features of data, random forests enhance classification capabilities, and Bayesian networks quantify fault probabilities, the transformation from data to risk assessment is achieved.

[0042] In some specific embodiments, step S102 specifically includes: Step S1021 involves performing data cleaning and feature extraction on the ginning process parameters, including removing duplicate data, filling in missing values, and extracting time-domain and frequency-domain features from equipment operating parameters, temperature, humidity, and dust concentration.

[0043] Step S1022: Input the preprocessed process parameters into the fusion neural network model, evaluate the process status through the weight matrix trained by historical abnormal samples, and output the process risk level including the probability of abnormality.

[0044] Step S1023: Construct a conditional probability model of equipment status and fault type based on Bayesian network, and calculate the posterior probability of fault for each process by combining real-time monitored parameters such as current and vibration frequency, and locate potential abnormal processes.

[0045] Step S1023 specifically includes: Step S10231: Based on the process information of the cotton ginning equipment, define network nodes as equipment status parameters and fault types; draw directed edges of the network using the experience of process experts to represent the dependencies between parameters.

[0046] Step S10232: Using historical process data, calculate the conditional probability of each node under the state of its parent node using the maximum likelihood estimation method.

[0047] Step S10233: Input the currently monitored equipment status parameters as evidence into the network and use the variable elimination algorithm for probabilistic reasoning; starting from the root node, update the posterior probability of each node layer by layer to obtain the joint posterior probability of each fault type.

[0048] Step S10234: Set a fault probability threshold, filter out fault types with posterior probabilities exceeding the threshold; map the fault types to the corresponding processes, sort the processes by risk according to the posterior probability values, and output a list of potential abnormal processes.

[0049] This embodiment utilizes historical process data and quantifies the conditional probability of each node under its parent node's state using maximum likelihood estimation, forming a probability knowledge base. For example, when the current exceeds the rated value by 20%, the motor overload probability is 0.7. Real-time monitored equipment status parameters are input into the network as evidence. A variable elimination algorithm is used to update the posterior probability of nodes layer by layer, starting from the root node. The probability of each fault type is calculated through joint probability calculation. Based on a preset fault probability threshold, high-risk fault types are selected and mapped to corresponding processes. A list of potential abnormal processes is output, sorted by posterior probability. Real-time updates to network evidence and calculation of posterior probabilities reflect the impact of changes in equipment operating status on fault risk. For example, when the ambient temperature rises, adjusting the posterior probability of motor overheating faults provides a 3-5 hour earlier warning of potential risks compared to static threshold alarms. Based on the mapping relationship between fault types and processes, abstract fault risks are concretized into specific production links. The risk ranking function allows maintenance personnel to prioritize high-probability abnormal processes, improving fault handling efficiency.

[0050] Step S1024: Select parameters with risk coefficients higher than the threshold from the preset indicator library and generate a set of monitoring parameters including real-time monitoring frequency and early warning threshold.

[0051] This embodiment employs a neural network model. It takes the drying temperature curve of the past 24 hours as input, captures parameter change trends through a memory unit, and outputs the probability distribution of anomalies for each process. An attention mechanism is introduced during model training to focus on parameters that significantly impact quality, such as the correlation between ginning machine speed and lint percentage. A conditional probability table is constructed for equipment state variables, including current I, vibration frequency f, and bearing wear and motor overload. The posterior probability is updated using real-time monitoring data. For example, when current I exceeds the rated value by 20% and vibration frequency f shows an abnormal peak of 100Hz, the probability of bearing wear is inferred to be higher than 90%, thus locating the abnormal process.

[0052] This embodiment establishes a three-layer filtering method for a preset index library: The first layer: outlier identification based on data mining, which uses box plot analysis and a random forest neural network model to determine the outlier thresholds for each parameter.

[0053] The second layer: The risk level output by the neural network is weighted and fused with the Bayesian inference result to generate a parameter risk coefficient (e.g., the abnormal dust concentration risk coefficient of a certain process = neural network probability × 0.6 + Bayesian posterior probability × 0.4).

[0054] The third layer: Combine production quality soft measurement models (such as the PLSR model of near-infrared spectroscopy and fiber strength) to select parameters with a weight of >15% influencing quality as core monitoring indicators.

[0055] This embodiment integrates a neural network model of LSTM and random forest. LSTM is used to capture the temporal dependencies of process parameters, and the multi-feature classification capability of random forest is combined to output the probability distribution of process anomalies. The Bayesian network is based on the historical probability association between equipment status and fault type, combined with real-time parameters to calculate the posterior probability, locate potential abnormal processes, and display the potential abnormal processes.

[0056] Step S103: Perform abnormal pattern recognition processing on the ginning process data according to the monitoring parameters, and separate out the abnormal parameters of the ginning process that differ from the normal process state.

[0057] This embodiment performs structured analysis on cotton ginning process data. First, data distribution features are extracted using convolutional layers to construct a clustering feature tree. Tree parameters such as the number of nodes and sample radius thresholds are defined, and nodes are allocated according to sample distance rules. Then, the tree is hierarchically split. When the number of node samples or the sample spacing exceeds the limit, a new node is generated using the farthest sample point as the seed, until the convergence condition is met. Finally, based on the preset threshold of the monitoring parameters or the bottom 20% of the data, outliers in each level of data are filtered. If multidimensional data exists, splitting continues until the smallest dimension of the outlier parameter is located.

[0058] In this way, based on data clustering and hierarchical decomposition strategies, complex data is mapped into a tree structure, and abnormal patterns are filtered out by parameter thresholds, achieving accurate positioning from the whole to the part.

[0059] In some specific embodiments, step S103 specifically includes the following steps: Step S1031: Perform feature extraction and cluster feature tree construction on the cotton ginning process data to obtain a hierarchical cluster feature tree. The feature extraction uses a convolutional layer module to extract features from the original process data to obtain feature data reflecting the data distribution characteristics. The cluster feature tree construction includes defining tree parameters: the maximum number of internal nodes, the maximum number of leaf nodes, and the maximum sample radius threshold for leaf nodes. Read the first sample point from the dataset as the root node, and read subsequent sample points in sequence, calculating the distance between the current sample point and the nearest sample point in the existing nodes. If the distance is less than or equal to the maximum sample radius threshold, add the current sample point to the node. If the distance exceeds the maximum sample radius threshold and the number of samples in the node does not reach the maximum number of internal nodes, combine the current sample point with the nearest sample point to form a new node. If the number of samples in the node reaches the maximum number of internal nodes, select the two sample points with the farthest distance in the node as seeds for the new leaf node, and redistribute the other sample points in the original node to the new leaf node according to the distance. Repeat the above process until all sample points are assigned, forming a hierarchical cluster feature tree.

[0060] Step S1032: Perform hierarchical node splitting on the clustering feature tree to obtain split data for each level. The node splitting process starts from the root node and proceeds layer by layer. First, the root node is used as the first level of split data. If the number of samples in the first level of split data exceeds the maximum number of internal nodes or the maximum distance between samples exceeds the maximum sample radius threshold, then the first level of split data is split a second time. The two sample points with the farthest distance in the split data are selected as new child nodes, and the original split data is redistributed to the child nodes according to the distance to form the second level of split data. This splitting process is repeated until all levels of nodes meet the conditions that the number of samples does not exceed the maximum number of internal nodes and the maximum distance between samples does not exceed the maximum sample radius threshold, thus obtaining the split data for each level.

[0061] Step S1033: Perform outlier filtering on the split data at each level according to the monitoring parameters to obtain abnormal parameters of the cotton ginning process; the monitoring parameters include a preset threshold or a baseline benchmark of the bottom 20%; for the split data at each level, extract the value of the corresponding indicator of the monitoring parameter, and mark the data points whose indicator value is lower than the preset threshold or lower than the bottom 20% baseline benchmark; if the split data still contains multiple dimensions after filtering, perform hierarchical splitting on the remaining data again until the filtered abnormal parameters are the smallest dimension data, and finally obtain the abnormal parameters of the cotton ginning process.

[0062] As can be seen, the convolutional layer module extracts the core features of the data, reduces redundant information, and improves the efficiency of subsequent analysis. Meanwhile, the clustering feature tree transforms massive amounts of process data into a structured tree structure, facilitating rapid retrieval and analysis of similarities and differences between data. The hierarchical node splitting mechanism adjusts the granularity of data partitioning through parameter thresholds, preventing excessive data aggregation from masking anomalies and avoiding noise interference from overly fine partitioning, ensuring accurate localization of anomaly patterns. For example, when a small fluctuation occurs in the equipment parameters of a certain process, this mechanism can pinpoint the specific anomalous parameter through layer-by-layer splitting, improving the accuracy of identification. Through repeated hierarchical splitting and filtering, it ensures that the finally located anomalous parameter has the smallest dimension, avoiding attributing anomalies to general processes or data sets, and shortening the problem-solving time.

[0063] Step S104: Perform parameter location processing on the abnormal parameters of the ginning process to determine the ginning quality assessment value corresponding to the abnormal parameters.

[0064] This embodiment addresses the identified abnormal parameters by dividing the equipment into spatial and temporal grid cells based on a spatiotemporal grid positioning model, calculating the abnormal parameters, and combining the deviation threshold with the weights associated with equipment quality. A Gaussian mixture model is used to fit the probability density of the abnormal parameter combinations to extract quality impact features. The parameter weights are adjusted using a weighted entropy method, with higher fluctuation entropy values ​​resulting in greater weights. Finally, the abnormal parameter features are matched with a historical case library using K-nearest neighbor matching, and the current quality assessment value is calculated by weighting the quality assessment results of the matched cases.

[0065] In this way, by integrating spatiotemporal analysis, probabilistic modeling, and case-based reasoning, the impact of anomalous parameters on quality can be quantified, and historical experience can be combined to improve the accuracy of the assessment.

[0066] Step S105: Generate ginning quality assessment results based on the ginning quality assessment values, and provide abnormal process prompts based on the assessment results.

[0067] This embodiment generates a cotton ginning quality assessment report based on the quality assessment values. The report includes a list of abnormal parameters, the degree of impact, related processes, and historical comparison data. The assessment results are displayed through a visual interface, abnormal processes are highlighted, and early warning information is pushed out. This allows for quick location of abnormal processes, shortening fault response time; the visual display reduces decision-making difficulty and improves management efficiency.

[0068] In one embodiment of the present invention, based on step S104, the following will provide a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S104 specifically includes: Step S1041: Divide the abnormal parameters of the cotton ginning process into multi-dimensional grid cells according to the spatial distribution of the equipment and the time of collection. Calculate the spatiotemporal influence factor of the abnormal parameters in each grid cell. The influence factor is determined by multiplying the degree of deviation of the abnormal parameters from the threshold with the correlation weight of the equipment to the quality in the grid cell. Step S1042: Use a Gaussian mixture model to fit the probability density of the abnormal parameter combinations in each grid cell, extract the feature parameters that can characterize the quality impact, and output the quantitative impact value of the abnormal parameter combinations on the ginning quality. Step S1043: Adjust the contribution weight of each parameter to the quality assessment based on the time series fluctuation entropy value of the abnormal parameters; Step S1044: Perform K-nearest neighbor matching between the feature vector of the current abnormal parameter combination and the abnormal-quality correlation data recorded in the historical case library, and calculate the ginning quality assessment value corresponding to the current abnormal parameter by weighting the quality assessment results of the matched cases.

[0069] In this embodiment, the impact of anomalies on quality is quantified by analyzing the correlation between abnormal parameters and quality, and finally, a quantifiable and comparable quality assessment value is generated.

[0070] It should be noted that since equipment malfunctions can affect quality through complex pathways, such as excessively high temperatures causing fiber thermal damage, it is necessary to mine the correlation between the two through historical data. For example, the co-occurrence frequency of "abnormal ginning machine temperature" and "reduced fiber strength" over the past year can be statistically analyzed, and the Pearson correlation coefficient can be calculated. If r=0.75, it can be verified whether "abnormal temperature" significantly predicts "reduced strength," thereby determining which abnormal parameters are directly related to quality and recording the correlation strength, such as strong correlation, moderate correlation, and weak correlation.

[0071] The weights of the impact of abnormal parameters on quality should be determined. Correlation strength only reflects the correlation; the degree of influence needs further quantification. For strongly correlated abnormal parameters, such as "abnormal temperature in the cotton ginning machine," combined with process knowledge (e.g., for every 10°C increase in temperature, fiber strength decreases by 5%), the absolute values ​​of the parameter coefficients are extracted as weights, such as a temperature weight of 0.6 and a current weight of 0.4.

[0072] Next, the overall quality impact value is calculated. The deviation of each abnormal parameter, such as an actual temperature value of 120℃, a threshold of 100℃, and a deviation of 20%, is normalized to the [0,1] interval, such as 20% / maximum possible deviation = 0.5. After multiplying by the corresponding weight, the values ​​are accumulated to obtain the overall quality impact value. For example, the total impact value after accumulating multiple abnormal parameters is 0.7.

[0073] Finally, standardized quality assessment values ​​are determined. The comprehensive impact value is compared with a preset benchmark (e.g., industry standards specify "impact value > 0.6 is unacceptable") to classify the assessment level, such as 0.8-1.0 as "serious nonconformity", 0.5-0.8 as "moderate nonconformity", and 0-0.5 as "minor nonconformity". Standardization converts continuous impact values ​​into discrete assessment levels, facilitating subsequent abnormal process alerts.

[0074] It can be seen that step S104 achieves accurate quantification from abnormal parameters to quality assessment values ​​through association mapping, weight allocation, comprehensive calculation and standardization.

[0075] In some specific embodiments, step S1041 specifically includes: Step S10411: Construct a spatiotemporal grid positioning model and divide it into multi-dimensional grid cells; based on the physical location coordinates and time series segmentation of the ginning equipment, map the abnormal parameters of the ginning process to three-dimensional grid cells; adjust the grid cell size according to the equipment distribution density and process cycle.

[0076] Step S10412: Calculate the spatiotemporal deviation of the abnormal parameters within the grid cell; for the abnormal parameters within each grid cell, calculate their positive / negative deviation and deviation amount relative to the preset threshold, where deviation amount = actual value - threshold; combine the time series trend to calculate the comprehensive deviation using a weighted method.

[0077] Step S10413: Based on historical process data, the correlation between abnormal parameters and quality parameters is analyzed by linear regression, and the absolute value of the regression coefficient is extracted as the basic weight; the basic weight is corrected by combining process knowledge to obtain the correlation weight of equipment parameters on quality.

[0078] Step S10414: Multiply the comprehensive deviation of the abnormal parameters within the grid cell by the correlation weight of the corresponding device to obtain the preliminary spatiotemporal influence factor; truncate and correct the influence factors that exceed the mean ± 3 times the standard deviation to finally obtain the stable spatiotemporal influence factor.

[0079] In this embodiment, the spatiotemporal impact of anomalous parameters on quality is quantified through spatiotemporal gridding analysis and multi-dimensional parameter fusion, thus constructing a spatiotemporal grid positioning model. Since the cotton ginning machines and dryers are spatially distributed within the workshop, and process parameters change over time, anomalous parameters need to be mapped to grid cells along a dual dimension of "spatial location + time node." For example, the cotton ginning machines are arranged along the X-axis (0-10m), and the dryers are arranged along the Y-axis (0-8m), with the spatial grid divided into 1m × 1m sections. The time dimension is divided by hour, such as 8:00-9:00 as a time unit, ensuring that each grid cell covers the parameters of a specific device during a specific time period. The spatiotemporal deviation of the anomalous parameters is calculated. The deviation of the anomalous parameters is not only reflected in their numerical magnitude but also related to their temporal persistence.

[0080] For example, if the temperature of a cotton gin remains above the threshold between 8:00 and 9:00, or suddenly jumps above the threshold at 9:00, the impact on quality will differ. Quantifying the direction and degree of deviation using a weighted method can more accurately reflect the severity of the anomaly.

[0081] Determine the association weight of equipment with quality. Different equipment parameters have varying degrees of impact on quality; for example, temperature has a greater impact on fiber strength than vibration, which needs to be verified using historical data. For instance, the correlation between "temperature deviation" and "reduction in fiber strength" over the past year was statistically analyzed, with a regression coefficient of 0.8. This coefficient was adjusted to 0.9 based on expert experience and used as the association weight for that equipment.

[0082] Finally, the spatiotemporal impact factors are calculated and calibrated. The initial impact factors (deviation × weight) may be affected by random noise (such as single-shot sensor errors). A sliding window is used to smooth the data (taking the average of three time units) to reduce fluctuations. Abnormally high impact factors are truncated to avoid individual extreme values ​​affecting the overall assessment, ensuring the stability and representativeness of the impact factors. This achieves the quantification of the spatiotemporal quality impact from outlier parameters.

[0083] In some specific embodiments, step S1042 specifically includes: Step S10421: Standardize the abnormal parameter combinations within the mesh cells; Step S10422: Based on the preprocessed abnormal parameter combination data, the expectation-maximization algorithm is used to estimate the number of components of the GMM, the mean vector of each component, the covariance matrix and the mixing weights; the model output is the probability density distribution of the abnormal parameter combination. Step S10423: Extract three types of feature parameters from the trained GMM: ① Pattern center, representing the typical state of abnormal parameter combinations; ② Pattern diffusion, which reflects the degree of dispersion of anomalous parameter combinations; ③ Pattern probability, representing the frequency of occurrence of the abnormal pattern in historical data; Step S10424: Weighted fusion of model center, diffusion and model probability to calculate the quantitative quality impact value of the abnormal parameter combination; verify the rationality of the impact value through historical data, and calibrate the parameters that deviate from the verification results.

[0084] Step S1042 quantifies the impact of abnormal parameter combinations on ginning quality through multi-parameter fusion modeling. In this step, constructing and training a Gaussian Model (GMM) aims to capture the complex distribution characteristics of abnormal parameter combinations. Traditional single Gaussian models can only describe a single-center distribution, while actual abnormal patterns may exhibit multi-center characteristics, such as the typical abnormal patterns of "high temperature, low strength" and "low temperature, high impurity content." By iteratively optimizing the number of components, mean, covariance, and weights of the GMM using the EM algorithm, the true distribution of abnormal parameter combinations can be accurately fitted. The BIC criterion is used to determine the optimal number of components, avoiding overfitting due to too many components or underfitting due to too few components.

[0085] Extracting characteristic parameters of quality impact is a crucial step in mining key information from the model. The mode center directly reflects the typical combination of anomalous parameters, such as "temperature 110℃ + vibration 5mm / s + intensity 30cN / tex" which is a common condition in a certain process and represents the "baseline state" of quality impact. The diffusion degree indicates the stability of the anomalous mode, and the mode probability reflects the frequency of occurrence of the anomalous mode.

[0086] Finally, calculating the quantified quality impact value is the core of transforming the model output into a comparable quality assessment indicator. By weighted fusion of three types of feature parameters—pattern center reflecting the degree of impact, diffusion reflecting the controllability of the impact, and pattern probability reflecting the persistence of the impact—a comprehensive impact value is obtained. For example, an anomalous pattern whose pattern center corresponds to a 20% reduction in intensity has a high degree of impact, low controllability, and high persistence; its comprehensive impact value will be significantly higher than a pattern whose pattern center corresponds to a 10% reduction in intensity but has low diffusion and low probability. Verification with historical actual quality defect rates ensures the clear physical meaning of the impact value, thus providing a reliable basis for subsequent quality assessments. In this way, a quantitative mapping from anomalous parameter combinations to quality impact is achieved.

[0087] In some specific embodiments, step S1043 specifically includes: Step S10431: For the time series of abnormal parameters, the time domain variation coefficient, frequency domain energy ratio and nonlinear complexity index are calculated simultaneously to quantify the fluctuation characteristics of the parameters from three dimensions: fluctuation amplitude, frequency components and change patterns. Step S10432: Establish the correlation strength matrix between abnormal parameters and ginning quality indicators, and determine the direct impact weight and indirect transmission weight of each parameter on quality indicators such as fiber strength and impurity content through historical fault tree analysis. Step S10433: With the goal of minimizing the quality assessment error, the initial entropy weight method weights are globally optimized using a genetic algorithm. Under the constraints of parameter fluctuation characteristics and correlation strength matrix, the contribution weights of each parameter are iteratively adjusted. Step S10434: Based on the real-time monitored seed cotton moisture content and equipment load rate parameters, the weight coefficients are corrected through fuzzy logic reasoning to make the weight allocation adapt to the differences in the impact of parameters on quality under different production conditions.

[0088] In some specific embodiments, step S1044 specifically includes: Step S10441: Perform feature alignment on the abnormal parameter combinations within the current grid cell to generate a standardized feature vector consistent with the format of the historical case library; Step S10442: Store historical anomaly-quality correlation data in a three-level classification system: equipment type, anomaly mode, and quality impact level, and generate a unique index code for each case; at the same time, establish an inverted index table; Step S10443: Use Euclidean distance as a similarity measure and set the neighborhood size k; select the k closest similar cases to the current feature vector from the historical case library; Step S10444: Based on the quality assessment results of the matched cases, and combining the similarity between the case and the current case and the timeliness of the case, calculate the weighted evaluation value vˉ=∑(si×wt×vi) / ∑(si×wt), where vi is the evaluation value of the i-th matched case and wt is the timeliness weight; the final ginning quality assessment value is the standardized result of vˉ.

[0089] In some embodiments, step S1044 maps the current abnormal parameter combination to a referenceable quality assessment result through precise matching and weighted fusion of historical cases. Constructing a historical case library index structure improves matching efficiency. The matching scope can be narrowed from the entire library to a subset of similar equipment and similar anomalies. For example, if the current anomaly is "high temperature and low strength in cotton ginning machines," the index table directly locates historical cases under the "cotton ginning machine - high temperature and low strength" category, avoiding traversing methods like those used for dryers.

[0090] Next, multi-dimensional KNN matching is used to find the most relevant historical reference cases. Euclidean distance ensures the similarity of numerical features such as temperature and vibration frequency, and the neighborhood size k=5 is determined through cross-validation. Filtering similar cases, such as those with the same "cotton ginning machine malfunction" and the same "quality impact level," ensures that the matched cases are consistent with the "problem essence" of the current malfunction, avoiding mismatching "dryer malfunction" cases to the "cotton ginning machine malfunction" scenario.

[0091] Finally, the weighted calculation of the quality assessment value is to integrate information from multiple cases and improve the reliability of the results. The similarity 's' reflects the closeness of the current case to the present case, while the timeliness weight 'wt' considers the impact of process improvements. For example, if the current case is 0.3 distance from historical case A (s=0.77, weight 1.2) and 0.5 distance from case B (s=0.67, weight 1.0), and if case A's evaluation value is 0.7 and case B's is 0.5, then the weighted evaluation value vˉ=(0.77×1.2×0.7+0.67×1.0×0.5) / (0.77×1.2+0.67×1.0)≈0.65, ultimately mapping to "moderate non-compliance".

[0092] In summary, step S1044, through feature preprocessing, case library index optimization, multi-dimensional KNN matching, and weighted fusion, achieves a precise mapping from current anomalies to historical experience, making more comprehensive use of effective information in historical data and improving the credibility and practicality of quality assessment.

[0093] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0094] The following are embodiments of the intelligent monitoring system for environment and production information in cotton processing based on the Internet of Things provided in this disclosure. This system and the intelligent monitoring methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the intelligent monitoring system for environment and production information in cotton processing based on the Internet of Things, please refer to the embodiments of the above intelligent monitoring methods.

[0095] The system includes: Data acquisition module 201 is used to acquire ginning process data, which includes multiple preset ginning process parameters; The status monitoring module 202 is used to perform ginning status monitoring processing on the ginning process data to obtain monitoring parameters that reflect the equipment operation and production status during the ginning process; Anomaly identification module 203 is used to perform anomaly pattern recognition processing on the ginning process data according to the monitoring parameters, and separate out the abnormal parameters of the ginning process that differ from the normal process state. The status assessment module 204 is used to perform parameter location processing on the abnormal parameters of the ginning process and determine the ginning quality assessment value corresponding to the abnormal parameters. The result output module 205 is used to generate a ginning quality assessment result based on the ginning quality assessment value, and to provide an abnormal process prompt based on the assessment result.

[0096] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0097] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a full understanding of embodiments of the invention. However, those skilled in the art will recognize that the technical solutions of the invention can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the invention.

[0098] The IoT-based intelligent monitoring system for environmental and production information during cotton processing comprises the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein. These units and steps can be implemented using electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can implement the described functions using different methods for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0099] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things, characterized in that, The methods include: Step S101: Obtain ginning process data, which includes multiple preset ginning process parameters; Step S102: Perform ginning status monitoring processing on the ginning process data to obtain monitoring parameters reflecting the equipment operation and production status during the ginning process; Step S103: Perform abnormal pattern recognition processing on the ginning process data according to the monitoring parameters, and separate out the abnormal parameters of the ginning process that differ from the normal process state; Step S104: Perform parameter location processing on the abnormal parameters of the ginning process to determine the ginning quality assessment value corresponding to the abnormal parameters; Step S105: Generate ginning quality assessment results based on the ginning quality assessment values, and provide abnormal process prompts based on the assessment results.

2. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things as described in claim 1, characterized in that, Step S102 specifically includes: Step S1021: Perform data cleaning and feature extraction on the ginning process parameters, including removing duplicate data, filling in missing values, and extracting time-domain and frequency-domain features from equipment operating parameters and temperature, humidity, and dust concentration. Step S1022: Input the preprocessed process parameters into the fusion neural network model, evaluate the process status through the weight matrix trained by historical abnormal samples, and output the process risk level including the abnormal probability. Step S1023: Construct a conditional probability model of equipment status and fault type based on Bayesian network, and calculate the posterior probability of fault for each process by combining real-time monitored current and vibration frequency, and locate potential abnormal processes. Step S1024: Select parameters with risk coefficients higher than the threshold from the preset indicator library and generate a set of monitoring parameters including real-time monitoring frequency and early warning threshold.

3. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things as described in claim 2, characterized in that, Step S1023 specifically includes: Step S10231: Based on the process information of the cotton ginning equipment, define network nodes as equipment status parameters and fault types; draw directed edges of the network using the experience of process experts to represent the dependencies between parameters; Step S10232: Using historical process data, calculate the conditional probability of each node under the state of its parent node using the maximum likelihood estimation method; Step S10233: Input the currently monitored equipment status parameters as evidence into the network, and use the variable elimination algorithm to perform probabilistic reasoning; starting from the root node, update the posterior probability of each node layer by layer to obtain the joint posterior probability of each fault type; Step S10234: Set a fault probability threshold, filter out fault types with posterior probabilities exceeding the threshold; map the fault types to the corresponding processes, sort the processes by risk according to the posterior probability values, and output a list of potential abnormal processes.

4. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things as described in claim 1, characterized in that, Step S103 specifically includes: Step S1031: Use a convolutional layer module to extract feature data reflecting the data distribution characteristics from the cotton ginning process data. By setting the maximum number of internal nodes, the maximum number of leaf nodes, and the maximum sample radius threshold parameters of the clustering feature tree, the sample points are sequentially allocated to construct a hierarchical clustering feature tree according to the relationship between the distance between sample points and the above parameters. Step S1032: Start hierarchical node splitting from the root node of the clustering feature tree. Take the root node as the first layer of split data. When the number of samples in the split data of this layer exceeds the maximum number of internal nodes or the maximum distance between samples exceeds the maximum sample radius threshold, select the two farthest sample points as new child nodes, and redistribute the original split data according to the distance to form the next layer of split data. Repeat this process until each level of node satisfies the condition that the number of samples does not exceed the maximum number of internal nodes and the maximum distance between samples does not exceed the maximum sample radius threshold, and obtain the split data of each level. Step S1033: Based on the monitoring parameters including the preset threshold or the bottom 20% baseline, outlier screening is performed on the split data of each level; the values ​​of the indicators corresponding to the monitoring parameters in the split data are extracted, and data points that are lower than the preset threshold or the bottom 20% baseline are marked. If the data still contains multiple dimensions after screening, the level splitting process is performed again until the smallest dimension of the ginning process outlier parameter is obtained.

5. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things as described in claim 1, characterized in that, Step S104 specifically includes: Step S1041: Divide the abnormal parameters of the cotton ginning process into multi-dimensional grid cells according to the spatial distribution of the equipment and the time of collection. Calculate the spatiotemporal influence factor of the abnormal parameters in each grid cell. The influence factor is determined by multiplying the degree of deviation of the abnormal parameters from the threshold with the correlation weight of the equipment to the quality in the grid cell. Step S1042: Use a Gaussian mixture model to fit the probability density of the abnormal parameter combinations in each grid cell, extract the feature parameters that can characterize the quality impact, and output the quantitative impact value of the abnormal parameter combinations on the ginning quality. Step S1043: Adjust the contribution weight of each parameter to the quality assessment based on the time series fluctuation entropy value of the abnormal parameters; Step S1044: Perform K-nearest neighbor matching between the feature vector of the current abnormal parameter combination and the abnormal-quality correlation data recorded in the historical case library, and calculate the ginning quality assessment value corresponding to the current abnormal parameter by weighting the quality assessment results of the matched cases.

6. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things, as described in claim 5, is characterized in that... Step S1041 specifically includes: Step S10411: Construct a spatiotemporal grid positioning model and divide it into multi-dimensional grid cells; based on the physical location coordinates and time series segmentation of the ginning equipment, map the abnormal parameters of the ginning process to three-dimensional grid cells; adjust the grid cell size according to the equipment distribution density and process cycle. Step S10412: Calculate the spatiotemporal deviation of the abnormal parameters within the grid cell; for each abnormal parameter within the grid cell, calculate its positive / negative deviation and deviation amount relative to the preset threshold, where deviation amount = actual value - threshold; combine the time series trend to calculate the comprehensive deviation using a weighted method. Step S10413: Based on historical process data, the correlation between abnormal parameters and quality parameters is analyzed by linear regression, and the absolute value of the regression coefficient is extracted as the basic weight; the basic weight is corrected by combining process knowledge to obtain the correlation weight of equipment parameters on quality. Step S10414: Multiply the comprehensive deviation of the abnormal parameters within the grid cell by the correlation weight of the corresponding device to obtain the preliminary spatiotemporal influence factor; truncate and correct the influence factors that exceed the mean ± 3 times the standard deviation to finally obtain the stable spatiotemporal influence factor.

7. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things as described in claim 5, characterized in that, Step S1042 specifically includes: Step S10421: Standardize the abnormal parameter combinations within the mesh cells; Step S10422: Based on the preprocessed abnormal parameter combination data, the expectation-maximization algorithm is used to estimate the number of components of the GMM, the mean vector of each component, the covariance matrix and the mixing weights; the model output is the probability density distribution of the abnormal parameter combination. Step S10423: Extract three types of feature parameters from the trained GMM: ① Pattern center, representing the typical state of abnormal parameter combinations; ② Pattern diffusion, which reflects the degree of dispersion of anomalous parameter combinations; ③ Pattern probability, representing the frequency of occurrence of the abnormal pattern in historical data; Step S10424: Weighted fusion of model center, diffusion and model probability to calculate the quantitative quality impact value of the abnormal parameter combination; verify the rationality of the impact value through historical data, and calibrate the parameters that deviate from the verification results.

8. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things, as described in claim 5, is characterized in that... Step S1043 specifically includes: Step S10431: For the time series of abnormal parameters, the time domain variation coefficient, frequency domain energy ratio and nonlinear complexity index are calculated simultaneously to quantify the fluctuation characteristics of the parameters from three dimensions: fluctuation amplitude, frequency components and change patterns. Step S10432: Establish the correlation strength matrix between abnormal parameters and ginning quality indicators, and determine the direct impact weight and indirect transmission weight of each parameter on quality indicators such as fiber strength and impurity content through historical fault tree analysis. Step S10433: With the goal of minimizing the quality assessment error, the initial entropy weight method weights are globally optimized using a genetic algorithm. Under the constraints of parameter fluctuation characteristics and correlation strength matrix, the contribution weights of each parameter are iteratively adjusted. Step S10434: Based on the real-time monitored seed cotton moisture content and equipment load rate parameters, the weight coefficients are corrected through fuzzy logic reasoning to make the weight allocation adapt to the differences in the impact of parameters on quality under different production conditions.

9. The method for intelligent monitoring of environmental and production information during cotton processing based on the Internet of Things, as described in claim 5, is characterized in that... Step S1044 specifically includes: Step S10441: Perform feature alignment on the abnormal parameter combinations within the current grid cell to generate a standardized feature vector consistent with the format of the historical case library; Step S10442: Store historical anomaly-quality correlation data in a three-level classification system: equipment type, anomaly mode, and quality impact level, and generate a unique index code for each case; at the same time, establish an inverted index table; Step S10443: Use Euclidean distance as a similarity measure and set the neighborhood size k; select the k closest similar cases to the current feature vector from the historical case library; Step S10444: Based on the quality assessment results of the matched cases, and combining the similarity between the case and the current case and the timeliness of the case, calculate the weighted evaluation value vˉ=∑(si×wt×vi) / ∑(si×wt), where vi is the evaluation value of the i-th matched case and wt is the timeliness weight; the final ginning quality assessment value is the standardized result of vˉ.

10. An intelligent monitoring system for environmental and production information during cotton processing based on the Internet of Things, characterized in that, The system is used to implement the intelligent monitoring method for environmental and production information during cotton processing based on the Internet of Things as described in any one of claims 1 to 9; The system includes: The data acquisition module is used to acquire ginning process data, which includes multiple preset ginning process parameters. The status monitoring module is used to perform ginning process data monitoring and processing to obtain monitoring parameters reflecting the equipment operation and production status during the ginning process; An anomaly identification module is used to perform anomaly pattern recognition processing on the ginning process data based on the monitoring parameters, and to separate out the abnormal parameters of the ginning process that differ from the normal process state. The status assessment module is used to perform parameter location processing on abnormal parameters in the ginning process and determine the ginning quality assessment value corresponding to the abnormal parameters. The result output module is used to generate ginning quality assessment results based on the ginning quality assessment values, and to provide abnormal process prompts based on the assessment results.

Images