A pipeline batching and conveying control method based on internet of things

Abstract

The application discloses a kind of pipeline batching delivery control methods based on Internet of Things, comprising: in glass electric melting furnace batching pipeline key position deployment hyperspectral camera, real-time identification pollutant species and location in material;Workshop is provided with environmental sensor, and the digital twin of proportionality with physical workshop is built;When detecting pollutant, digital twin starts simulation engine, predicts pollution source diffusion path and establishes feature library;System real-time tracking pollution track, accurately calculates shunt opportunity, automatically triggers shunt valve and is polluted material into waste pool;While dynamically adjusting clean room air pressure gradient and cleaning time, form airflow barrier and block diffusion;After cleaning, carry out secondary monitoring and generate traceability report.The method realizes the automatic control of pollution whole process, improves production safety and efficiency, reduces material loss.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to an IoT-based method for controlling the feeding and conveying of materials on a production line. Background Technology

[0002] In traditional glass electric melting furnace batching production, material contamination control has long relied on a post-production reverse compensation mechanism. This involves adjusting subsequent formulation parameters to offset the impact after contamination is detected through offline testing or finished product analysis. This model has significant drawbacks: contaminants continue to diffuse in the production line, leading to batch-wide quality defects. For example, trace heavy metal contamination in blind spots of hyperspectral cameras can only be identified after melting when light transmittance decreases, by which time irreversible losses have already occurred. Furthermore, the dynamic characteristics of fluctuating environmental temperature and humidity and uneven distribution of trace elements in materials make a single reverse model unsuitable for various scenarios. New models are required for scenarios with novel or complex contamination, resulting in low control efficiency and high costs.

[0003] Existing contamination control technologies for production lines suffer from data fragmentation and insufficient coordination. While hyperspectral cameras can identify pollutants, they cannot be linked with environmental sensor data to predict diffusion paths. Digital twin technology, when simulating contamination processes, fails to form a closed-loop control system with the actual diversion actuators. This fragmentation means the system can only handle known contamination types. When facing dynamically changing contamination scenarios, manual parameter adjustments are required, easily leading to misjudgments or delays. For example, traditional pressure barriers use fixed parameters and cannot dynamically adjust cleaning intensity based on contaminant concentration, resulting in excessive cleanroom energy consumption or barrier failure, further exacerbating resource waste and the risk of production interruptions. Summary of the Invention

[0004] This application provides an IoT-based assembly line material conveying control method, which enables more efficient and accurate detection of electricity meters in power management areas, thereby saving power resources.

[0005] This application provides an IoT-based method for controlling the feeding and conveying of materials on a production line, including:

[0006] S1. Install hyperspectral cameras at key locations in the glass electric melting furnace batching line to continuously scan materials and identify the types and locations of contaminants in the materials in real time; deploy multiple environmental sensor nodes in the workshop and construct a virtual environment digital twin of the physical workshop based on all the data collected by the sensors.

[0007] S2, various pollutant digital feature models are pre-established in the digital twin. When the hyperspectral camera identifies a pollutant, it immediately records the pollutant as a pollution source and triggers the simulation engine in the digital twin.

[0008] S3, the digital twin calculates the precise location and diffusion range of the pollution source on the conveyor belt based on the pollution source type and the simulated diffusion path, and establishes a pollution source feature database;

[0009] S4 tracks the central movement range of the pollution source, determines the best time to perform diversion, and automatically triggers the diversion valve at the best time to guide the contaminated material into the waste pool.

[0010] S5 dynamically adjusts the air pressure gradient and cleaning time of the cleanroom near the pollution source according to the type of pollution source, forming an airflow barrier to prevent the spread of pollutants;

[0011] S6. After cleaning, secondary pollution source monitoring is performed on the materials, and a source tracing report is generated to identify the pollution sources of the materials.

[0012] Preferably, the calculation of the precise location and diffusion range of the pollution source on the conveyor belt specifically includes: using a digital twin combined with a hyperspectral camera to identify the material location information recorded during the pollution event, as well as the running speed and direction parameters of the conveyor belt, to determine the initial location of the pollution source on the conveyor belt; and based on the simulated pollutant diffusion path, considering the movement of the conveyor belt, to calculate the precise location and diffusion range of the pollution source on the conveyor belt at different times.

[0013] Preferably, determining the optimal timing for diversion includes: obtaining the precise coordinates of the pollution source center at the current moment by interpolating historical data of the pollution source location with time points; selecting currently available diversion valves from all abandoned diversion valve location information along the conveyor belt, calculating the distance from the pollution source center to each diversion valve, and selecting the closest one as the target diversion valve; calculating the remaining time required for the pollution source to reach the diversion valve based on the distance between the current location of the pollution source and the target diversion valve, combined with the conveyor belt running speed; calculating the triggering time for diversion in advance, considering the mechanical delay of the diversion valve's action; and immediately triggering diversion if the predicted arrival time minus the delay time is less than the current time.

[0014] Preferably, in step S1, multiple environmental sensor nodes are deployed within the workshop. Specifically, this includes: installing an online mass spectrometer to determine the molecular structure of pollutants; installing a laser particle size analyzer to analyze the physical morphology of pollutants; performing multi-dimensional quantitative analysis on the identified pollutants to generate a structured pollution characteristic report, which is then uploaded in real time to the IoT production line batching and conveying control system; after receiving the structured pollution characteristic report, the IoT production line batching and conveying control system searches the pollution-process mapping knowledge base; comprehensively evaluates the strategies in the multi-strategy candidate set based on capacity requirements to determine the optimal production parameter adjustment strategy with the highest comprehensive score; before issuing the parameter adjustment strategy instruction to the physical workshop production line, the parameter adjustment strategy is first input into a digital twin to simulate and predict the production line's operating status for the next half hour; if the simulation result shows a safe pass, the parameter adjustment strategy instruction is issued to the physical workshop production line; if the simulation reveals potential risks, the production parameter adjustment strategy is re-optimized.

[0015] Preferably, determining the optimal production parameter adjustment strategy with the highest comprehensive score specifically includes: the formula for calculating the comprehensive score S is: ,in , , These are the weighting coefficients for the impact on production capacity, the effectiveness of pollutant treatment, and the implementation cost, respectively. ; The maximum implementation cost for multiple strategy candidates. The coefficient representing the impact of the strategy on production capacity. To assess the effectiveness of the strategy in treating pollutants, The cost of implementing the strategy.

[0016] Preferably, the retrieval of the pollution-process mapping knowledge base includes: the pollution-process mapping knowledge base stores various historical pollutant events and their ultimately verified effective process adjustment schemes; the pollution-process mapping knowledge base has a dynamic update function, and as new pollutant events and effective countermeasures emerge, they are promptly added to the knowledge base; after the IoT assembly line batching and conveying control system retrieves the information, a preliminary multi-strategy candidate set is generated.

[0017] Preferably, the multi-dimensional quantitative analysis of the identified contaminants further includes: real-time detection of the types and concentrations of contaminants in the material; immediately generating a unique contamination event ID for the batch of material when a contaminant is detected, thereby identifying the contamination event; estimating the time interval from when the material enters the furnace to when the product transmittance may change based on historical data of furnace flow rate and reaction kinetics; monitoring the transmittance of the product using a spectrometer after the estimated time interval; if a decreasing trend in transmittance is detected, associating the decreasing signal with the corresponding contamination event ID to confirm whether the decrease in transmittance is caused by the contaminants in this batch; and confirming that the decrease in transmittance is caused by the contaminants in this batch. The dynamic evolution model is then triggered. The model calculates the conversion efficiency of the pollutant under the current operating conditions based on historical data. Based on the conversion efficiency and input parameters, the prediction result is determined. If no intervention is made on the prediction result, the extent to which the pollutant will eventually cause a decrease in the product transmittance is determined, and it is judged whether the decrease exceeds the tolerance range. If the judgment result is that the decrease exceeds the tolerance range, the compensation strategy generation stage is entered. The model automatically generates multiple compensation strategies, issues compensation instructions to the execution equipment, and monitors the transmittance with a spectrometer after the instructions are executed. If the transmittance decrease trend stops and begins to rise back to the target range, the compensation is judged to be successful. The entire process data of the pollution event ID is recorded in the knowledge base to optimize the dynamic evolution model.

[0018] Preferably, the model calculates the conversion efficiency of the pollutant under the current operating conditions based on historical data. Specifically, after confirming that the decrease in transmittance is caused by the batch of pollutants, the IoT production line batching and conveying control system automatically sends a trigger signal to the dynamic evolution model; based on the parameters of the current production conditions, historical data with operating conditions parameters within a set threshold range are selected from the historical data; for each set of similar operating condition historical data selected, the percentage decrease in transmittance caused by each ppm of pollutant is calculated; wherein the historical data is a set of data recorded in the past regarding the relationship between the pollutant and the change in transmittance under the same production conditions.

[0019] Preferably, the model automatically generates multiple compensation strategies, including: constructing a positive model relating pollutants and light transmittance; using the constructed positive model, combined with material characteristic data and environmental data under current production conditions, to predict pollutants that may appear in future melting cycles, generating a list of predicted pollutants, and clarifying the types and concentration ranges of pollutants; and generating an anti-interference preset formula in advance based on the list of predicted pollutants and adjusting workshop parameters to pre-eliminate pollution.

[0020] Preferably, the pollution pre-removal specifically includes: analyzing the potential impact of pollutants on the light transmittance of glass based on the predicted pollutant list, and generating an anti-interference preset formula that can pre-counteract pollutant interference by adjusting the optimal basic formula parameter Z0.

[0021] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0022] By integrating hyperspectral cameras, environmental sensors, and digital twins using IoT technology, real-time sensing, accurate simulation, and dynamic response to material contamination in production lines are achieved. This method can not only quickly identify pollutant types and predict their diffusion paths, but also effectively block contamination spread through intelligent diversion and airflow barrier control, while generating source tracing reports to assist in subsequent optimization. The entire process is highly automated and has a fast response speed, significantly improving the safety and production efficiency of production line material handling and reducing the cost of manual intervention and the risk of material loss.

[0023] By installing multiple detection devices to accurately identify pollutant characteristics and generate reports, a pollution-process mapping knowledge base is used to quickly generate a candidate set of response strategies. Then, the optimal strategy is determined by a comprehensive evaluation of multiple factors such as production capacity requirements. Finally, digital twin simulation is used for verification, which effectively improves the efficiency of pollutant treatment, ensures the stability of production line and product quality, reduces production costs and potential risks, and improves overall production efficiency.

[0024] By accurately predicting the impact of pollutants on product transmittance using a dynamic evolution model, the system automatically generates and executes optimal compensation strategies to effectively recover transmittance losses. Simultaneously, it records data throughout the entire process to optimize the model, forming a closed-loop management system that significantly improves product quality stability, reduces production costs, and enhances the controllability and intelligence of the production process.

[0025] By constructing a forward model, a direct mapping from the light transmittance target to the formula parameters is achieved, avoiding the lag of traditional reverse compensation. By using forward-looking pollution prediction to generate anti-interference preset formulas, production parameters can be adjusted in advance, effectively preventing the impact of pollutants on the light transmittance of glass, significantly improving product quality stability and production efficiency, and reducing material loss and energy costs. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart of an IoT-based assembly line material feeding and conveying control method according to an embodiment of the present invention. Detailed Implementation

[0027] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0029] Example 1: Figure 1 This is a schematic flowchart of an IoT-based assembly line material feeding and conveying control method according to an embodiment of the present invention.

[0030] like Figure 1 As shown, an IoT-based assembly line material feeding and conveying control method includes the following steps:

[0031] S1. Install hyperspectral cameras at key locations on the glass electric melting furnace batching line to continuously scan materials and identify the types and locations of contaminants in the materials in real time; deploy multiple environmental sensor nodes in the workshop and construct a virtual environment digital twin based on all the data collected by the sensors, which is proportional to the physical workshop.

[0032] Key locations include, but are not limited to, material silo outlets, conveyor belt turning points, and mixer inlets; multiple environmental sensor nodes collect data on airflow speed and direction, air pressure gradient, temperature, and humidity within the workshop.

[0033] S2: Digital characteristic models of various pollutants are pre-established within the digital twin. When the hyperspectral camera identifies a pollutant, it immediately records the pollutant as a pollution source and triggers the simulation engine within the digital twin.

[0034] The feature models include, but are not limited to, dust (chromium compound powder), aerosols (acid mist, volatile organic solvents), and microorganisms; model parameters include particle size distribution, density, diffusion coefficient, and natural settling rate (for dust); volatility, suspension time, and reaction with humidity (for aerosols); and survival rate, reproduction rate, and spread capability (for microorganisms). The simulation engine automatically calls the corresponding digital feature model based on the type of pollutant, and combines it with real-time airflow data, equipment layout, pressure gradient, and other parameters to simulate and calculate the diffusion path, impact range, and concentration distribution probability of the pollutant within the next 30 seconds.

[0035] S3, the digital twin calculates the precise location and diffusion range of the pollution source on the conveyor belt based on the pollution source type and the simulated diffusion path, and establishes a pollution source feature database.

[0036] Specifically, the digital twin combines the material location information recorded by a hyperspectral camera when identifying a contamination event, with the conveyor belt's speed and direction parameters, to determine the initial position of the contamination source on the conveyor belt. Assume that when the hyperspectral camera identifies the contamination event, the recorded position of the material in the workshop coordinate system is ( , The conveyor belt runs in the x-direction at a speed of... (Unit: m / s, obtained from the conveyor belt control system), the time interval from the identification time to the start of diffusion path calculation is... (Unit: seconds, obtainable from system time records), then the initial position of the pollution source on the conveyor belt. .

[0037] Based on the simulated pollutant diffusion path, considering the movement of the conveyor belt, the precise position of the pollution source on the conveyor belt at different times is calculated. Let the displacement of a point on the simulated pollutant diffusion path relative to the initial pollution source position in the x-direction be... (Unit: m, obtained from diffusion path data output by the simulation engine), combined with the operation of the conveyor belt, the precise x-direction position of the pollution source on the conveyor belt at time t. y-direction position (Assume that the position in the y-direction remains unchanged during the operation of the conveyor belt).

[0038] Calculate the diffusion range of the pollutant characteristic model. Within the horizontal diffusion range, for particulate pollutants, calculate based on their diffusion coefficient. Given a simulation time t, the dust diffusion range perpendicular to the conveyor belt's running direction (assumed to be the y-direction) can be approximately described by a normal distribution. Let the diffusion range be... ,but Here, k is an empirical coefficient that can be determined based on actual experimental data, generally ranging from 2 to 3. For example, by conducting multiple dust diffusion experiments in a simulated environment and measuring the dust diffusion distance in the y-direction at different times, the value of k can be obtained through fitting. In the longitudinal diffusion range, the natural settling rate of the dust is considered. Within time t, the diffusion range of dust in the longitudinal direction (assuming it's the z-direction). Where h is the initial height of the pollution source from the conveyor belt surface (unit: m, which can be determined based on the workshop equipment layout and the location of the pollution source). When When the value is ≤0, it means that the dust has settled onto the conveyor belt.

[0039] The diffusion range of aerosols and microorganisms is analyzed by combining the simulated diffusion paths and airflow data. In the direction perpendicular to the conveyor belt's running direction, the aerosol's suspension time can be used as a reference. The diffusion range is estimated using the microbial propagation capacity function P(d). For example, for aerosols, assuming that their diffusion range perpendicular to the conveyor belt during suspension time is related to the extent of airflow disturbance in that direction, an approximate diffusion distance can be obtained by analyzing airflow data. For microorganisms, according to the transmission capability function P(d), the distance corresponding to when the transmission probability drops to a certain threshold (e.g., 10%) is... As its diffusion range.

[0040] The characteristic parameters such as pollution source type, initial location, time sequence of precise location change, and time sequence of diffusion range change are stored in the pollution source characteristic database.

[0041] The propagation capability function P(d) quantifies the probability of microbial contaminants spreading over a spatial distance d. Its core logic is to use a mathematical model to describe how the propagation capability of contaminants decreases with increasing distance as they diffuse from the pollution source. The specific function formula is defined according to the application scenario.

[0042] S4 tracks the central movement range of the pollution source, determines the optimal time to perform diversion, and automatically triggers the diversion valve at the optimal time to guide the contaminated material into the waste pool.

[0043] The optimal time to perform a diversion is the precise time required for the contaminant source to reach the nearest discarded diversion valve.

[0044] Specifically, by using historical data on the location of pollution sources and interpolating with time points, the precise coordinates of the pollution source center at the current moment can be obtained. For example, if the current time falls between two known locations, the center point location is estimated based on the time ratio.

[0045] From the location information of all abandoned diversion valves along the conveyor belt, the currently available diversion valves are selected, and the distance from the center of the pollution source to each diversion valve is calculated. The one with the closest distance is selected as the target diversion valve.

[0046] Assuming the pollution source moves at a constant speed with the conveyor belt, the remaining time required for the pollution source to reach the diversion valve is calculated based on the distance between the current position of the pollution source and the target diversion valve, combined with the conveyor belt speed. If there are interfering factors such as airflow or mechanical vibration, the time can be fine-tuned using experimentally calibrated correction factors.

[0047] Considering the mechanical delays in the diversion valve's operation (such as valve opening time and signal transmission time), calculate the timing for triggering the diversion in advance. If the predicted arrival time minus the delay time is less than the current time, trigger the diversion immediately.

[0048] When the trigger time is reached, an opening signal is sent to the target diversion valve. During the diversion process, the location of the contamination source is continuously monitored. If the source deviates from the diversion valve by more than a preset threshold, the diversion stops and an alarm is issued. A diversion time limit is also set to prevent excessive material accumulation.

[0049] Record key information about diversion events (such as trigger time, diversion valve number, and final location of the pollution source) for subsequent analysis and optimization. Adjust pollutant diffusion model parameters based on diversion effects (such as the proportion of waste material) to improve the accuracy of subsequent predictions.

[0050] S5 dynamically adjusts the air pressure gradient and cleaning time of the cleanroom near the pollution source according to the type of pollution source, forming an airflow barrier to prevent the spread of pollutants.

[0051] S6. After cleaning, secondary pollution source monitoring is performed on the materials, and a source tracing report is generated to identify the pollution sources of the materials.

[0052] The source tracing report records the time, location, type of pollutants, measures taken, and results of the incident.

[0053] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0054] By integrating hyperspectral cameras, environmental sensors, and digital twins using IoT technology, real-time sensing, accurate simulation, and dynamic response to material contamination in production lines are achieved. This method can not only quickly identify pollutant types and predict their diffusion paths, but also effectively block contamination spread through intelligent diversion and airflow barrier control, while generating source tracing reports to assist in subsequent optimization. The entire process is highly automated and has a fast response speed, significantly improving the safety and production efficiency of production line material handling and reducing the cost of manual intervention and the risk of material loss.

[0055] Example 2: While Example 1 utilized technologies such as hyperspectral cameras and digital twins to identify and treat material contaminants, it still has shortcomings. Firstly, the pre-established digital characteristic model of contaminants is relatively fixed and struggles to adapt to the differences in material composition between different batches and the impact of dynamic changes in the workshop environment on contaminant characteristics. This leads to deviations in the contaminant diffusion simulation results, affecting the accuracy of subsequent diversion and airflow barrier control. Secondly, this solution does not fully consider the actual production capacity requirements of the factory. Under different production capacity scenarios, enterprises have different strategies and requirements for contaminant treatment. When capacity is tight, production must be prioritized; when capacity is ample, thorough treatment can be implemented. However, Example 1 lacks a mechanism for dynamically adjusting treatment strategies based on production capacity, making it difficult to balance contaminant treatment with production efficiency. To address these issues and improve the accuracy, flexibility, and adaptability of contaminant treatment to better meet the production needs of different factories, [further details are needed].

[0056] In some embodiments, step S1, which involves arranging multiple environmental sensor nodes within the workshop, further includes:

[0057] S11. Install an online mass spectrometer in the workshop to determine the molecular structure of pollutants; install a laser particle size analyzer to analyze the physical morphology of pollutants.

[0058] S12 performs multi-dimensional quantitative analysis on the identified pollutants, generates a structured pollution characteristic report, and uploads it to the IoT assembly line batching and conveying control system in real time.

[0059] The structured pollution characteristics report includes pollutant types, pollutant concentrations, pollutant change trends, and spatial distribution of pollutants.

[0060] S13, after receiving a structured contamination feature report, the IoT assembly line batching and conveying control system searches the contamination-process mapping knowledge base.

[0061] The pollution-process mapping knowledge base stores historical pollution events and their proven effective process adjustment solutions. For example, if the pollutant is "iron oxide dust" with a concentration >50 ppm and an upward trend, effective countermeasures might include: slightly increasing the smelting temperature by 20°C, increasing the dosage of clarifying agent A by 5%, and reducing the conveyor belt speed by 10%. If the pollutant is "organic silicone oil," effective countermeasures might include: initiating ultrasonic cleaning at a specific frequency and temporarily adjusting the material mixing sequence. Furthermore, the pollution-process mapping knowledge base should be dynamically updated, promptly incorporating new pollution events and effective countermeasures to continuously improve and enrich its content. After retrieval by the IoT-based production line batching and conveying control system, a preliminary multi-strategy candidate set is generated.

[0062] S14. Based on capacity demand, a comprehensive evaluation of the strategies in the multi-strategy candidate set is conducted to determine the optimal production parameter adjustment strategy with the highest comprehensive score.

[0063] Specifically, strategies from a multi-strategy candidate set are comprehensively evaluated based on capacity demand. Capacity demand can be measured by the quantity of products that need to be produced per unit of time. Let the target capacity be... Current actual production capacity is Taking the following factors into consideration:

[0064] The impact of strategies on capacity is assessed by evaluating the degree to which each strategy increases or decreases capacity after implementation. For example, some strategies may reduce capacity by adding process steps, while others may increase capacity by optimizing processes. The impact coefficient of a strategy on capacity can be estimated using historical data or simulation experiments. Its value range is generally between -1 and 1, with negative numbers indicating reduced production capacity and positive numbers indicating increased production capacity.

[0065] The effectiveness of strategies in pollutant treatment is evaluated based on the characteristics of the pollutants and the treatment objectives. Each strategy can be scored by establishing a pollutant treatment effectiveness scoring standard. The scoring range can be set according to the actual situation, such as 0-10 points. The higher the score, the better the processing effect.

[0066] The cost of implementing the strategy should be estimated by considering factors such as equipment modifications, raw material consumption, and labor costs. .

[0067] The formula for calculating the overall score S is: ,in , , These are the weighting coefficients for the impact on production capacity, the effectiveness of pollutant treatment, and the implementation cost, respectively. The weighting coefficient can be set according to the focus and objectives of the production process. For example, if more emphasis is placed on pollutant treatment, the weighting coefficient can be appropriately increased. The value; The maximum implementation cost for multiple strategy candidates. The coefficient representing the impact of the strategy on production capacity. To assess the effectiveness of the strategy in treating pollutants, To determine the cost of implementing the strategy, data normalization is required before calculating the overall score. By calculating the overall score for each strategy, the optimal production parameter adjustment strategy with the highest overall score is identified.

[0068] S15, before issuing parameter adjustment strategy instructions to the physical workshop production line, first input the parameter adjustment strategy into the digital twin to simulate and predict the working status of the production line in the next half hour.

[0069] S16. If the simulation results show that the simulation is successful, the parameter adjustment strategy instruction will be sent to the physical workshop production line; if the simulation finds potential risks, the production parameter adjustment strategy will be re-optimized.

[0070] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0071] By installing multiple detection devices to accurately identify pollutant characteristics and generate reports, a pollution-process mapping knowledge base is used to quickly generate a candidate set of response strategies. Then, the optimal strategy is determined by a comprehensive evaluation of multiple factors such as production capacity requirements. Finally, digital twin simulation is used for verification, which effectively improves the efficiency of pollutant treatment, ensures the stability of production line and product quality, reduces production costs and potential risks, and improves overall production efficiency.

[0072] Example 3: In Example 2, a pollutant identification system incorporating multi-dimensional detection methods such as online mass spectrometry and laser particle size analyzer was constructed. Combined with a pollution-process mapping knowledge base and digital twin simulation technology, dynamic optimization and adjustment of production parameters were achieved. However, in practical applications, the construction of the pollution-process mapping knowledge base heavily relies on historical data accumulation. For newly emerging pollutant types or complex pollutant coupling events, the knowledge base may have coverage blind spots. Simultaneously, while the digital twin simulation can predict short-term risks, it lacks a systematic assessment capability for the cumulative effects on the long-term process chain, leading to a phenomenon where some parameter adjustment strategies exhibit "short-term safety but long-term instability" during implementation. Furthermore, the weighting coefficients in the comprehensive evaluation stage rely on human experience. When facing dynamic changes in different production conditions, the static weighting system struggles to match the optimal decision-making requirements in real time. To overcome these limitations, a dynamic decision-making framework with self-learning capabilities needs to be constructed, and the evaluation mechanism for the long-term stability of the process chain needs to be strengthened. Example 3 will propose improvement solutions to address these issues.

[0073] In some embodiments, step S12, which involves multi-dimensional quantitative analysis of the identified pollutants, further includes:

[0074] S121, real-time detection of the types and concentrations of contaminants in materials. When a contaminant is detected, a unique contamination event ID is immediately generated for the batch of materials to identify the contamination event.

[0075] The pollution event ID is associated with the pollutant type, concentration, and timestamp of the material entering the furnace (T=0 seconds), and this information is stored in the IoT production line batching and conveying control system.

[0076] S122, based on historical data of furnace flow rate and reaction kinetics, estimates the time interval from when the material enters the furnace to when the product transmittance may change.

[0077] S123, after the estimated time interval, use a spectrometer to monitor the transmittance of the product; if a downward trend in transmittance is detected, associate the downward signal with the corresponding pollution event ID to confirm whether the decrease in transmittance is caused by this batch of pollutants.

[0078] Specifically, after completing step S122, which involves estimating the time interval from when the material enters the furnace to when the product transmittance may change based on historical data of furnace flow rate and reaction kinetics (denoted as...),... After that, when the system time reaches the time when the material enters the furnace ( ) plus At that time, the spectrometer is activated to monitor the transmittance of the product. The spectrometer continuously acquires the transmittance value of the product according to the set sampling frequency (e.g., data is collected once every 5 seconds), and records the transmittance value collected each time. , where i=1,2,3,⋯, represents the sampling sequence number.

[0079] To determine whether transmittance shows a decreasing trend, a sliding window method is used. The size of the sliding window is set to n (generally n = 5-10, but can be adjusted according to the stability and sensitivity requirements of the actual monitoring), meaning that transmittance data from n consecutive sampling points are analyzed each time. The average rate of change v of transmittance within the window is calculated using the formula: in, It is the transmittance value of the last sampling point within the window. It is the transmittance value of the first sampling point within the window, and Δt is the sampling interval time of the spectrometer (e.g., the 5 seconds mentioned earlier).

[0080] Set a threshold for judging the decreasing trend of light transmittance. The reference range is 0.01–0.05 (unit: % / second; the specific value can be adjusted according to the sensitivity requirements for light transmittance changes in actual production). If the calculated average rate of change... (A negative sign indicates a decrease in light transmittance), so it is determined that the light transmittance is showing a downward trend.

[0081] Once a decreasing trend in transmittance is detected, the contamination event ID corresponding to the current batch of materials is retrieved from the IoT-based batching and conveying control system (generated and stored in step S121). This transmittance decrease signal is then associated with the contamination event ID. This association can be achieved by creating a record in the system's database containing the contamination event ID, the start time of the transmittance decrease (i.e., the time corresponding to the first sampling point within the window), and the average rate of change of the decreasing trend.

[0082] By using this correlation, combined with the information on contaminant types and concentrations recorded in step S121, it can be preliminarily confirmed whether the decrease in transmittance is caused by this batch of contaminants. If, after this batch of material enters the furnace, a decrease in transmittance occurs within a reasonable timeframe, and there are no other obvious interfering factors (such as furnace equipment malfunction, cross-contamination from other batches of material, etc., which can be eliminated by setting corresponding monitoring and judgment rules in the system), then it can be considered that the decrease in transmittance is related to this batch of contaminants.

[0083] S124. Once it is confirmed that the decrease in light transmittance is caused by this batch of pollutants, a dynamic evolution model is triggered. The model calculates the conversion efficiency of the pollutant under the current operating conditions based on historical data.

[0084] Specifically, after completing step S123, which confirms that the decrease in light transmittance is caused by the batch of pollutants, the IoT production line batching and conveying control system automatically sends a trigger signal to the dynamic evolution model to start the model for subsequent calculations.

[0085] Historical data is a collection of records showing the relationship between a contaminant and changes in transmittance under similar production conditions in the past. Specifically, it includes the following types of information: Contaminant concentration information: the concentration of the contaminant in the material during different batches of production in the past, in ppm. This data reflects different levels of contamination. Transmittance change information: corresponding to the batches with different contaminant concentrations, the changes in transmittance of the product during the production process, including the start time and magnitude of the transmittance decrease. For example, recording the transmittance decrease from an initial value to a final value within a specific time period, and then calculating the percentage decrease. Current operating condition related parameters: furnace temperature, material composition, and other parameters that may be related to the current production conditions in the past. These parameters are used to filter historical data similar to the current operating conditions to improve the accuracy of model calculations.

[0086] Based on the parameters of the current production conditions, such as the current furnace temperature (set to...). The similarity thresholds for operating parameters are selected from historical data, including the content of other major components in the material. For example, the similarity threshold for furnace temperature can be set to ±5℃ (the specific value can be adjusted according to the actual production requirements for operating stability). Only historical data whose operating parameters are within the set threshold range are selected for subsequent calculations.

[0087] For each set of historical data under similar operating conditions selected, calculate the percentage decrease in transmittance caused by each ppm of pollutant. Let the pollutant concentration in a certain set of historical data be... (ppm), corresponding to a percentage decrease in transmittance of Then, for this set of data, what is the percentage decrease in transmittance per ppm of pollutants? The calculation formula is: The calculations were performed on all the selected historical data of similar operating conditions. The average value is taken to obtain the conversion efficiency E of the pollutant under the current operating conditions, and the formula is: Where m is the number of sets of historical data with similar working conditions selected. It is the percentage decrease in light transmittance per ppm of pollutants calculated from the kth set of historical data.

[0088] S125: Determine the prediction result based on the conversion efficiency and input parameters. If no intervention is made on the prediction result, determine the extent to which the pollutants in this batch will eventually cause a decrease in the product's light transmittance, and determine whether the decrease exceeds the tolerance range. If the determination result is that the decrease exceeds the tolerance range, proceed to the compensation strategy generation stage.

[0089] The input parameters include the type of pollutant, its concentration, the current furnace temperature, and the current rate of decrease in transmittance.

[0090] S126, the model automatically generates multiple compensation strategies and sends compensation commands to the execution device. After the commands are executed, the transmittance is monitored by a spectrometer. If the transmittance decreases and begins to rise back to the target range, the compensation is considered successful.

[0091] Specifically, the model automatically generates multiple compensation strategies, such as chemical compensation and process compensation. Chemical compensation, taking the addition of CeO2 as an example, calculates the theoretical compensation dose required to neutralize contaminants based on chemical reaction principles. Considering reaction efficiency, the actual compensation dose to be injected is determined. Process compensation, such as fine-tuning the heating curve, alters the redox state of the glass by changing the temperature in the furnace back zone, promoting the existence of iron ions in a more transparent valence state, and predicting the transmittance loss that this operation can recover. The model evaluates multiple compensation strategies, comprehensively considering factors such as compensation effect, operational feasibility, and cost, and selects the optimal combination of compensation strategies. For example, a combined scheme with chemical compensation as the primary method and process compensation as a secondary method may be chosen.

[0092] The model sends compensation commands to the compensator injection system, IoT-based batching and conveying control system, and electrode power regulator, specifying the injection rate, injection time, and electrode power adjustment parameters. The executing equipment follows the commands, injecting the compensator into the furnace at a specific rate and adjusting the electrode power to raise the temperature in a specific area of ​​the furnace. After compensation begins, the transmittance is continuously monitored using a spectrometer. If the decreasing transmittance stops and begins to rise, eventually stabilizing within the target range, the compensation is considered successful.

[0093] S127, record the entire process data of pollution event IDs into the knowledge base to optimize the dynamic evolution model.

[0094] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0095] By accurately predicting the impact of pollutants on product transmittance using a dynamic evolution model, the system automatically generates and executes optimal compensation strategies to effectively recover transmittance losses. Simultaneously, it records data throughout the entire process to optimize the model, forming a closed-loop management system that significantly improves product quality stability, reduces production costs, and enhances the controllability and intelligence of the production process.

[0096] Example 4: In Example 3, when glass batching control is carried out based on a traditional reverse reasoning model, the model uses data on pollutants that have appeared in actual production and the corresponding deviations in glass transmittance as input. It adjusts the formulation parameters through reverse analysis to compensate for the impact of pollution. However, due to significant differences in material characteristic data (such as fluctuations in trace element content) and environmental data (such as changes in temperature and humidity) across different production batches, the types and concentration ranges of pollutants actually introduced have considerable uncertainty and dynamic variation characteristics. Faced with this complex and ever-changing pollution scenario, the unified reverse compensation model used in Example 3 is difficult to fully cover various potential pollutant combinations. When dealing with new pollutants or abnormal fluctuations in pollutant concentrations, it is prone to insufficient or over-compensation, thus affecting the accurate achievement of the glass transmittance target.

[0097] In some embodiments, in step S126, the model automatically generates multiple compensation strategies, and further includes:

[0098] S1261, Construct a positive model relating pollutants and light transmittance.

[0099] The model takes the ideal glass transmittance target value T_target, material quality big data (uniformly expressed as material characteristic data such as trace element spectral analysis data of the current batch of materials), and environmental data as inputs. By analyzing the intrinsic relationship between these data and glass transmittance, the output is the optimal basic formula parameter Z0 required to achieve the transmittance target. Z0 includes key parameters such as the proportion of raw materials and the amount of clarifying agent. This forward model achieves a direct mapping from performance requirements to formula parameters through target-driven approach, which is different from the logic of traditional reverse reasoning where pollution occurs first and then compensation is provided.

[0100] S1262 utilizes the established forward model, combined with material characteristic data and environmental data under current production conditions, to predict pollutants that may appear during future melting cycles, generate a list of predicted pollutants, and clarify the types and concentration ranges of pollutants.

[0101] S1263, based on the predicted pollutant inventory, generate an anti-interference preset formula in advance and adjust workshop parameters to pre-exclude pollutants.

[0102] Specifically, based on the predicted pollutant inventory, the potential impact of pollutants on the light transmittance of glass is analyzed. By adjusting the optimal basic formula parameter Z0, an anti-interference preset formula that can preemptively offset the interference of pollutants is generated. At the same time, relevant parameters in the workshop (such as melting temperature and time) are adjusted according to the preset formula to ensure that the production system is in an anti-interference state before the actual introduction of pollutants, thus achieving pre-rejection of pollution.

[0103] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0104] By constructing a forward model, a direct mapping from the light transmittance target to the formula parameters is achieved, avoiding the lag of traditional reverse compensation. By using forward-looking pollution prediction to generate anti-interference preset formulas, production parameters can be adjusted in advance, effectively preventing the impact of pollutants on the light transmittance of glass, significantly improving product quality stability and production efficiency, and reducing material loss and energy costs.

[0105] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for controlling the feeding and conveying of materials on an assembly line based on the Internet of Things, characterized in that, include: S1. Install hyperspectral cameras at key locations in the glass electric melting furnace batching line to continuously scan materials and identify the types and locations of contaminants in the materials in real time; deploy multiple environmental sensor nodes in the workshop and construct a virtual environment digital twin of the physical workshop based on all the data collected by the sensors. S2, various pollutant digital feature models are pre-established in the digital twin. When the hyperspectral camera identifies a pollutant, it immediately records the pollutant as a pollution source and triggers the simulation engine in the digital twin. S3, the digital twin calculates the precise location and diffusion range of the pollution source on the conveyor belt based on the pollution source type and the simulated diffusion path, and establishes a pollution source feature database; The calculation of the precise location and diffusion range of the pollution source on the conveyor belt specifically includes: using a digital twin combined with a hyperspectral camera to identify the material location information recorded when the pollution event is identified, as well as the running speed and direction parameters of the conveyor belt, to determine the initial location of the pollution source on the conveyor belt; and based on the simulated pollutant diffusion path, considering the movement of the conveyor belt, to calculate the precise location and diffusion range of the pollution source on the conveyor belt at different times. S4. Track the center movement range of the pollution source, determine the optimal time to perform diversion, and automatically trigger the diversion valve at the optimal time to guide the contaminated material into the waste pool. Determining the optimal time to perform diversion includes: using historical data of the pollution source location and interpolating with the time point to obtain the precise coordinates of the pollution source center at the current moment; filtering out the currently available diversion valves from all waste diversion valve location information along the conveyor belt, calculating the distance from the pollution source center to each diversion valve, and selecting the closest one as the target diversion valve; calculating the remaining time required for the pollution source to reach the diversion valve based on the distance between the current location of the pollution source and the target diversion valve, combined with the conveyor belt running speed; considering the mechanical delay of the diversion valve action, calculating the triggering time for diversion in advance; if the predicted arrival time minus the delay time is less than the current time, diversion is triggered immediately. S5 dynamically adjusts the air pressure gradient and cleaning time of the cleanroom near the pollution source according to the type of pollution source, forming an airflow barrier to prevent the spread of pollutants; S6. After cleaning, secondary pollution source monitoring is performed on the materials, and a source tracing report is generated to identify the pollution sources of the materials.

2. The IoT-based assembly line material dispensing and conveying control method as described in claim 1, characterized in that, The deployment of multiple environmental sensor nodes within the workshop specifically includes: installing an online mass spectrometer to determine the molecular structure of pollutants; installing a laser particle size analyzer to analyze the physical morphology of pollutants; performing multi-dimensional quantitative analysis on the identified pollutants to generate a structured pollution characteristic report, which is then uploaded in real time to the IoT production line batching and conveying control system; upon receiving the structured pollution characteristic report, the IoT production line batching and conveying control system searches the pollution-process mapping knowledge base; comprehensively evaluates the strategies in the multi-strategy candidate set based on capacity requirements to determine the optimal production parameter adjustment strategy with the highest comprehensive score; before issuing the parameter adjustment strategy instruction to the physical workshop production line, the parameter adjustment strategy is first input into a digital twin to simulate and predict the production line's operating status for the next half hour; if the simulation results show a safe pass, the parameter adjustment strategy instruction is issued to the physical workshop production line; if the simulation reveals potential risks, the production parameter adjustment strategy is re-optimized.

3. The IoT-based assembly line material dispensing and conveying control method as described in claim 2, characterized in that, The determination of the optimal production parameter adjustment strategy with the highest comprehensive score specifically includes: comprehensive score S The calculation formula is: ,in These are the weighting coefficients for the impact on production capacity, the effectiveness of pollutant treatment, and the implementation cost, respectively. ; The maximum implementation cost for multiple strategy candidates. The coefficient representing the impact of the strategy on production capacity. To assess the effectiveness of the strategy in treating pollutants, The cost of implementing the strategy.

4. The IoT-based assembly line material dispensing and conveying control method as described in claim 2, characterized in that, The retrieval of the pollution-process mapping knowledge base includes: the pollution-process mapping knowledge base stores various historical pollutant events and their ultimately verified effective process adjustment schemes; the pollution-process mapping knowledge base has a dynamic update function, and as new pollutant events and effective countermeasures emerge, they are promptly added to the knowledge base; after the IoT assembly line batching and conveying control system retrieves the information, a preliminary multi-strategy candidate set is generated.

5. The IoT-based assembly line material dispensing and conveying control method as described in claim 2, characterized in that, The multi-dimensional quantitative analysis of the identified contaminants also includes: real-time detection of the types and concentrations of contaminants in the material; immediately generating a unique contamination event ID for the batch of material upon detection of a contaminant to identify the contamination event; estimating the time interval from when the material enters the furnace to when the product transmittance may change based on historical data of furnace flow rate and reaction kinetics; monitoring the transmittance of the product using a spectrometer after the estimated time interval; if a decreasing trend in transmittance is detected, associating the signal determination result of the decreasing transmittance trend with the corresponding contamination event ID to confirm whether the decrease in transmittance is caused by the contaminants in this batch; and confirming that the decrease in transmittance is related to the contaminants in this batch. Once triggered, a dynamic evolution model is established. The model calculates the conversion efficiency of the pollutant under current operating conditions based on historical data. Based on the conversion efficiency and input parameters, a prediction result is determined. If no intervention is made regarding the prediction result, the extent to which the pollutant will ultimately cause a decrease in product transmittance is assessed, and it is determined whether this decrease exceeds the tolerance range. If the determination result indicates that the decrease exceeds the tolerance range, the compensation strategy generation stage begins. The model automatically generates multiple compensation strategies, issues compensation commands to the execution equipment, and monitors the transmittance using a spectrometer after execution. If the transmittance decrease stops and begins to recover to the target range, the compensation is considered successful. The entire process data of the pollution event ID is recorded in the knowledge base to optimize the dynamic evolution model.

6. The IoT-based assembly line material dispensing and conveying control method as described in claim 5, characterized in that, The model calculates the conversion efficiency of the pollutant under the current operating conditions based on historical data. Specifically, after confirming that the decrease in transmittance is caused by this batch of pollutants, the IoT production line batching and conveying control system automatically sends a trigger signal to the dynamic evolution model. Based on the parameters of the current production conditions, historical data with operating conditions parameters within the set threshold range are selected from the historical data. For each set of similar historical data, the percentage decrease in transmittance caused by each ppm of pollutant is calculated. Historical data is a collection of records showing the relationship between the pollutant and changes in light transmittance under the same production conditions in the past.

7. The IoT-based assembly line material dispensing and conveying control method as described in claim 5, characterized in that, The model automatically generates multiple compensation strategies, including: constructing a positive model relating pollutants to light transmittance; using the constructed positive model, combined with material characteristic data and environmental data under current production conditions, to predict pollutants that may appear in future melting cycles, generating a list of predicted pollutants, and clarifying the types and concentration ranges of pollutants; and generating an anti-interference preset formula in advance based on the list of predicted pollutants and adjusting workshop parameters to pre-exclude pollutants.

8. The IoT-based assembly line material dispensing and conveying control method as described in claim 7, characterized in that, The pollution pre-rejection specifically includes: analyzing the potential impact of pollutants on the light transmittance of glass based on a predicted pollutant inventory, and adjusting the optimal basic formula parameters. This generates an anti-interference preset formula that can pre-counteract pollutant interference.

Images