Monitoring method and system for flue gas flow of aluminum-silicon alloy eutectic tank
By combining an integrated intelligent measuring instrument and adaptive filtering technology with an LSTM neural network, the real-time performance and accuracy issues of flue gas flow monitoring in aluminum-silicon alloy eutectoid tanks were resolved, achieving sensor stability and energy consumption optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ORDOS MENGTAI ALUMINUM CO LTD
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-09
Smart Images
Figure CN122170969A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of non-ferrous metal smelting technology, and in particular to a method and system for monitoring the flow rate of flue gas in an aluminum alloy eutectoid cell. Background Technology
[0002] Aluminum-silicon alloys, as important lightweight alloy materials, are widely used in aerospace, automotive manufacturing, and electronic packaging. In the production of aluminum-silicon alloys, the eutectoid cell is one of the core smelting equipment, obtaining alloy products through a eutectoid reaction of aluminum and silicon raw materials under high-temperature conditions. During the operation of the eutectoid cell, the exhaust system plays a crucial role in discharging high-temperature reaction gases, maintaining pressure balance within the cell, and ensuring a stable smelting environment. The flue gas flow rate, as a key parameter reflecting the operating status of the eutectoid cell, directly relates to process conditions such as the uniformity of raw material supply, the normal operation of electrodes, and the absence of blockages in the exhaust pipes. Therefore, accurate monitoring of the flue gas flow rate in the aluminum-silicon alloy eutectoid cell is of great significance for ensuring production safety, optimizing process parameters, and reducing energy consumption.
[0003] Currently, the non-ferrous metal smelting industry primarily monitors flue gas parameters using traditional manual inspection methods or a combination of separate sensors. Manual inspection relies on maintenance personnel periodically using handheld devices to perform single-point measurements at flue gas duct boreholes, recording parameters such as temperature and pressure. This method is not only labor-intensive and lacks real-time accuracy, but the measurement results are also significantly affected by the operator's experience and the location of the measurement points, making it difficult to reflect the true state of the flue gas across the flue gas duct cross-section. Some companies have experimented with a combination of separate sensors, where temperature sensors, pressure sensors, and Pitot tube flow meters are installed at different locations within the flue gas duct. A data acquisition module then aggregates the signals from each sensor and uploads them to the monitoring system.
[0004] However, such solutions have the following inherent drawbacks: First, multiple sensors are installed independently, occupying multiple openings in the flue, making installation complex and maintenance difficult; second, there is a lack of coordination between the sensors, resulting in low data fusion and making it difficult to achieve synchronous acquisition and correlation analysis of multiple parameters; third, the sensors are directly exposed to the high temperature, high dust, and strong corrosion environment of aluminum-silicon alloy flue gas, making the pressure tapping holes prone to clogging and the probes prone to corrosion, resulting in short measurement failure cycles and high maintenance costs. Summary of the Invention
[0005] This invention provides a method and system for monitoring flue gas flow in an aluminum-silicon alloy eutectoid tank, which solves the problem of poor real-time performance in monitoring flue gas flow in existing technologies.
[0006] On one hand, the present invention provides a method for monitoring the flow rate of flue gas in an aluminum-silicon alloy eutectoid cell, comprising: An integrated intelligent measuring instrument was installed in the aluminum-silicon alloy eutectoid cell, and the humidity compensation coefficient of the flue gas composition was calibrated. The flue gas parameters of the flue section are collected by the integrated intelligent measuring instrument. The flue gas parameters are subjected to adaptive Kalman median-mean fusion sliding window filtering to obtain preprocessed data; Based on the flue gas composition humidity compensation coefficient, the preprocessed data is input into the dual compensation density model to calculate the instantaneous flow rate; The preprocessed data and the instantaneous flow rate are fused with the eutectoid tank production process parameters, input into an LSTM neural network hierarchical early warning model, and combined with a fault knowledge graph for operating condition diagnosis, and the early warning level is output. Based on the instantaneous flow rate of each eutectoid cell, the operation strategy of the exhaust fan is optimized by a genetic algorithm and a scheduling command is output.
[0007] Optionally, the integrated intelligent measuring instrument integrates a multi-array micro probe, a self-cleaning module, a constant temperature heating module, and a sensor self-sensing module, and the integrated intelligent measuring instrument is installed in a single hole in the mainstream gas state area of the aluminum-silicon alloy eutectoid flue. The multi-array microprobe consists of three sets of microsensor arrays arranged in a triangular pattern. The multi-array microprobe simultaneously collects temperature, static pressure, differential pressure, flue gas solids concentration, and humidity data at multiple points on the flue gas cross-section. The flue gas solids concentration is the mixed concentration of alumina and silica dust. The self-cleaning module cleans the probe by blowing the pressure tapping hole with a high-pressure air knife and then vibrating the probe surface with ultrasonic waves according to a preset cycle or differential pressure signal fluctuation threshold. The constant temperature heating module controls the probe temperature to be 10-15°C higher than the flue gas dew point temperature. The sensor self-sensing module collects the working status data of the multi-array microprobe, constructs a sensor health assessment model based on the working status data, and generates a software compensation coefficient based on the working status data when data drift occurs in the data collected by the multi-array microprobe, thereby completing remote online calibration.
[0008] Optionally, the flue gas parameters are subjected to adaptive Kalman median-mean fusion sliding window filtering to obtain preprocessed data, including: Calculate the fluctuation coefficient of the flue gas parameter, where the fluctuation coefficient is the ratio of the standard deviation to the mean of the flue gas parameter; The sliding window capacity is dynamically adjusted based on the fluctuation coefficient. The flue gas parameters are input into a Kalman filter to filter out continuous electromagnetic interference and flue gas steady-state pulsation noise in the industrial field, and the first filtering data is obtained. The filtered data is input into the sliding window capacity. The data within the sliding window is sorted, and the maximum and minimum values of 1 / 3 after sorting are removed. The arithmetic mean of the remaining 1 / 3 of the data is calculated to obtain the preprocessed data.
[0009] Optionally, dynamically adjusting the sliding window capacity based on the fluctuation coefficient includes: If the fluctuation coefficient is less than 5%, the sliding window capacity will be adjusted to the first preset number of data points. If the fluctuation coefficient is between 5% and 15%, the sliding window capacity will be adjusted to the second preset number of data points. If the fluctuation coefficient is greater than 15%, the sliding window capacity will be adjusted to the third preset data point.
[0010] Optionally, based on the flue gas composition humidity compensation coefficient, the preprocessed data is input into a dual-compensation density model to calculate the instantaneous flow rate, including: The temperature, pressure, flue gas solids concentration, humidity data, and flue gas composition humidity compensation coefficient in the preprocessed data are input into the double-compensation flue gas density model to calculate the corrected actual flue gas density. The dual-compensation flue gas density model is as follows: ; in, For actual flue gas density, This refers to the density of flue gas under standard conditions. , To measure the flue gas pressure and temperature, , For pressure and temperature under standard conditions, This refers to the concentration of solids in the flue gas. For the humidity of the flue gas, The component compensation coefficient, This is the humidity compensation coefficient; The flue gas velocity data is extracted from the preprocessed data, and combined with the effective cross-sectional area of the eutectoid flue and the actual flue gas density, the instantaneous flow rate of the flue gas is calculated using the fluid dynamics flow rate formula.
[0011] Optionally, the preprocessed data and the instantaneous flow rate are fused with the eutectoid tank production process parameters, input into an LSTM neural network hierarchical early warning model, and combined with a fault knowledge graph for operating condition diagnosis, outputting an early warning level, including: The preprocessed data, the instantaneous flow rate, and the eutectoid cell production process parameters are fused to obtain fused data; the production process parameters include the raw material aluminum-silicon ratio, electrode operating current, and in-cell melting temperature. Multi-dimensional features, including time-domain and frequency-domain features, are extracted from the fused data; the time-domain features are the mean, variance, mutation rate, and duration of continuous anomalies in the data, and the frequency-domain features are the main frequency and spectral energy after Fourier transform. The multi-dimensional features are input into the LSTM neural network hierarchical early warning model to obtain preliminary working condition diagnosis results; The preliminary diagnostic results are matched with the knowledge graph of aluminum-silicon alloy eutectoid tank faults. The preliminary diagnostic results are then corrected based on the matching results, and the operating condition diagnostic results and corresponding warning levels are output.
[0012] Optionally, based on the instantaneous flow rate of each eutectoid cell, a genetic algorithm is used to optimize the operation strategy of the exhaust fan and output scheduling instructions, including: Collect instantaneous flow rate and static pressure data of flue gas in the parallel production scenario of the eutectoid cell, and acquire fan equipment parameters including rated power, speed-volume characteristic curve, and energy consumption characteristic curve of the exhaust fan. With the objective function of minimizing the total energy consumption of the exhaust fans and balancing the exhaust load of each fan, and with the rated speed and maximum air volume of the fans as constraints, an optimization model for the operation of exhaust fans based on genetic algorithm is constructed. The instantaneous flow rate of the eutectoid tank, the static pressure data of the flue, and the parameters of the fan equipment are input into the exhaust fan operation optimization model. The model is solved by encoding, crossover, mutation and iteration through a genetic algorithm to obtain the operating power adjustment value, speed adjustment value and standby fan start-up and shutdown strategy of each exhaust fan. The operating power adjustment value, the speed adjustment value, and the standby fan start-stop strategy are converted into industrial control scheduling instructions and sent to the frequency conversion control system and start-stop control system of the exhaust fan.
[0013] Optional, also includes: The flue gas parameters, instantaneous flow rate, warning level, and dispatch instructions are dynamically displayed in multiple dimensions on the front-end interface.
[0014] Optional, also includes: The flue gas parameters and instantaneous flow rate are stored as time-series data in the InfluxDB time-series database; The process parameters, early warning records, and scheduling instructions for the eutectoid tank are stored as structured data in a MySQL database.
[0015] On the other hand, the present invention also provides an aluminum-silicon alloy eutectoid cell flue gas flow monitoring system, comprising: An integrated intelligent measuring instrument is installed in the main flow area of the aluminum-silicon alloy eutectoid flue to collect flue gas parameters of the flue section and calibrate the flue gas composition humidity compensation coefficient. A host computer is communicatively connected to the integrated intelligent measuring instrument, and the host computer includes: The filtering module is used to perform adaptive Kalman median-mean fusion sliding window filtering on the received flue gas parameters to obtain preprocessed data; The flow calculation module is used to input the preprocessed data into the dual-compensation density model to calculate the instantaneous flow rate based on the pre-calibrated flue gas composition humidity compensation coefficient; The operating condition early warning module is used to integrate the preprocessed data and the instantaneous flow rate with the eutectoid tank production process parameters, input them into the LSTM neural network hierarchical early warning model, combine them with the fault knowledge graph to perform operating condition diagnosis, and output the early warning level. The collaborative scheduling module is used to optimize the operation strategy of the exhaust fan and output scheduling instructions based on the instantaneous flow rate of each eutectoid tank using a genetic algorithm.
[0016] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method as described above.
[0017] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method as described above.
[0018] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method as described above.
[0019] The present invention provides a method and system for monitoring flue gas flow in an aluminum-silicon alloy eutectoid tank. The method addresses the problem of sensor clogging and corrosion by installing an integrated intelligent measuring instrument and calibrating the compensation coefficient; it eliminates on-site noise interference by adaptively fusing and filtering flue gas parameters; it overcomes the shortcomings of traditional temperature and pressure correction methods that do not consider the influence of flue gas components by introducing a dual-compensation density model to calculate instantaneous flow, thus improving the accuracy of flue gas flow calculation; it achieves quantitative fault identification from parameter monitoring by fusing flue gas parameters with process parameters and inputting the result into an LSTM model combined with a knowledge graph for operational condition diagnosis; and it optimizes the fan operation strategy using a genetic algorithm, realizing multi-tank collaborative scheduling and energy consumption optimization, thus solving the problem of poor real-time performance in monitoring flue gas flow in an aluminum-silicon alloy eutectoid tank. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0021] Figure 1 This is a schematic flowchart of the flue gas flow monitoring method for an aluminum-silicon alloy eutectoid tank provided in an embodiment of the present invention; Figure 2This is a schematic diagram of the structure of the aluminum-silicon alloy eutectoid cell flue gas flow monitoring system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0023] Figure 1 This is a schematic flowchart of the flue gas flow monitoring method for a silicon-aluminum alloy eutectoid tank provided in an embodiment of the present invention.
[0024] like Figure 1 As shown in the embodiment of the present invention, the method for monitoring flue gas flow in a silicon-aluminum alloy eutectoid cell mainly includes the following steps: 101. Install an integrated intelligent measuring instrument in the aluminum-silicon alloy eutectoid cell and calibrate the humidity compensation coefficient of the flue gas composition.
[0025] Among them, the integrated intelligent measuring instrument integrates a multi-array micro probe, a self-cleaning module, a constant temperature heating module and a sensor self-sensing module, and the integrated intelligent measuring instrument is installed in a single hole in the mainstream gas state area of the aluminum-silicon alloy eutectoid flue. The multi-array microprobe consists of three sets of microsensor arrays arranged in a triangular pattern. The multi-array microprobe simultaneously collects temperature, static pressure, differential pressure, flue gas solids concentration, and humidity data at multiple points on the flue gas cross-section. The flue gas solids concentration is the mixed concentration of alumina and silica dust. The self-cleaning module cleans the probe by using a combination of high-pressure air knife to blow the pressure tapping hole and ultrasonic vibration to beat the probe surface, based on a preset cycle or differential pressure signal fluctuation threshold. The constant temperature heating module controls the probe temperature to be 10-15℃ higher than the flue gas dew point temperature.
[0026] The sensor self-sensing module collects the working status data of the multi-array micro probes, builds a sensor health assessment model based on the working status data, and generates software compensation coefficients based on the working status data when data drift occurs in the data collected by the multi-array micro probes, thus completing remote online calibration.
[0027] In addition, when calibrating the flue gas composition and humidity compensation coefficient, on-site calibration tests were conducted in conjunction with the process characteristics of aluminum-silicon alloy smelting and the actual emission conditions of the eutectoid tank flue gas. By taking multiple samples of flue gas in the flue, the corresponding relationship between flue gas solids concentration, humidity and flue gas density under different process conditions was obtained. After data fitting and error correction, the composition compensation coefficient and humidity compensation coefficient that are suitable for the characteristics of the aluminum-silicon alloy eutectoid tank flue gas were determined, providing correction parameters for the subsequent calculation of flue gas density by the dual-compensation density model.
[0028] 102. Collect flue gas parameters of the flue section using an integrated intelligent measuring instrument.
[0029] Specifically, flue gas parameters are collected using an integrated intelligent measuring instrument. This instrument simultaneously acquires data from multiple points on the flue gas cross-section using a multi-array micro-probe, thus comprehensively reflecting the state distribution characteristics of the flue gas. Simultaneously, the coordinated operation of the self-cleaning module and the constant temperature heating module effectively avoids data deviations caused by probe contamination or temperature fluctuations, further improving the reliability of the measurement results.
[0030] During data acquisition, the sensor self-sensing module continuously monitors the operating status of the multi-array microprobes and generates a health assessment report in real time. If drift in the probe data is detected, the software compensation coefficient is automatically calculated based on a pre-built health assessment model, and the data is corrected through remote online calibration. This mechanism not only reduces the workload of manual maintenance but also significantly improves the long-term stability and measurement accuracy of the system.
[0031] 103. Adaptive Kalman median-mean fusion sliding window filtering is applied to the flue gas parameters to obtain preprocessed data.
[0032] Specifically, adaptive Kalman median-mean fusion sliding window filtering is applied to the flue gas parameters to obtain preprocessed data, including: Calculate the fluctuation coefficient of the flue gas parameters, which is the ratio of the standard deviation to the mean of the flue gas parameters.
[0033] The fluctuation coefficient calculation process takes into account the variation characteristics of flue gas parameters within a certain time range. By analyzing the relationship between the standard deviation and mean of the flue gas parameters, it can effectively reflect the dispersion and stability of the flue gas parameters. When the fluctuation coefficient is high, it indicates that the flue gas parameters change drastically, and there may be abnormal disturbances or measurement noise; while when the fluctuation coefficient is low, it indicates that the flue gas parameters are relatively stable and the system operation is relatively stable.
[0034] The sliding window capacity is dynamically adjusted based on the fluctuation coefficient.
[0035] Specifically, the sliding window capacity is dynamically adjusted based on the fluctuation coefficient, including: If the fluctuation coefficient is less than 5%, the sliding window capacity will be adjusted to the first preset data point. If the fluctuation coefficient is between 5% and 15%, the sliding window capacity will be adjusted to the second preset number of data points. If the fluctuation coefficient is greater than 15%, the sliding window capacity will be adjusted to the third preset data point.
[0036] Specifically, when the fluctuation coefficient is less than 5%, the flue gas condition is considered stable, and the sliding window capacity is adjusted to the first preset 20 data points to balance response speed and filtering effect; when the fluctuation coefficient is between 5% and 15%, the condition is considered to have moderate fluctuations, and the sliding window capacity is adjusted to the second preset 30 data points to enhance the suppression of moderate amplitude noise; when the fluctuation coefficient is greater than 15%, the condition is considered to have drastic fluctuations, and the sliding window capacity is adjusted to the third preset 40 data points to improve robustness to outliers under drastic fluctuation conditions, thereby achieving dynamic adaptation between the sliding window capacity and the flue gas condition to ensure that the filtering effect is always in the optimal state.
[0037] The flue gas parameters are input into a Kalman filter to filter out continuous electromagnetic interference and flue gas steady-state pulsation noise in the industrial environment, thus obtaining primary filtered data.
[0038] Input the filtered data into the sliding window capacity, sort the data within the sliding window, remove the maximum and minimum values of 1 / 3 after sorting, calculate the arithmetic mean of the remaining 1 / 3 of the data, and obtain the preprocessed data.
[0039] The process involves inputting flue gas parameters into a Kalman filter. Using a preset state transition matrix and observation matrix, systematic noise such as continuous electromagnetic interference and flue gas pulsations in the industrial environment is filtered out, resulting in primary filtered data. This primary filtered data is then input into an adjusted sliding window. The data within the window is sorted in ascending order, and the maximum values in the first third and the minimum values in the last third are removed. Only the arithmetic mean of the data points in the middle third is calculated. This arithmetic mean is the final preprocessed data, effectively filtering out abnormal values such as sudden dust impacts and equipment vibrations while preserving the true trend of flue gas parameter changes.
[0040] 104. Based on the flue gas composition and humidity compensation coefficient, the preprocessed data is input into the dual-compensation density model to calculate the instantaneous flow rate.
[0041] Specifically, based on the flue gas composition and humidity compensation coefficient, the preprocessed data is input into the dual-compensation density model to calculate the instantaneous flow rate, including: The temperature, pressure, flue gas solids concentration, humidity data, and flue gas composition humidity compensation coefficient from the preprocessed data are input into the double-compensated flue gas density model to calculate the corrected actual flue gas density.
[0042] The composition compensation coefficient and temperature compensation coefficient are parameters determined based on field calibration tests and are used to correct deviations in the flue gas density calculation. The dual-compensation flue gas density model combines the effects of flue gas composition and humidity on density, and calculates the flue gas density value under actual operating conditions using a multivariate nonlinear regression algorithm.
[0043] The dual-compensation flue gas density model is as follows: ; in, For actual flue gas density, This refers to the density of flue gas under standard conditions. , To measure the flue gas pressure and temperature, , For pressure and temperature under standard conditions, This refers to the concentration of solids in the flue gas. For the humidity of the flue gas, The component compensation coefficient, This is the humidity compensation coefficient.
[0044] Flue gas velocity data is extracted from the preprocessed data. Combined with the effective cross-sectional area of the eutectoid flue and the actual flue gas density, the instantaneous flow rate of the flue gas is calculated using the fluid dynamics flow rate formula.
[0045] Specifically, the filtered flue gas velocity data is extracted from the preprocessed data. This flue gas velocity data is then combined with the pre-measured effective cross-sectional area of the eutectoid duct and the actual flue gas density. The results are then substituted into the fluid dynamics flow formula to calculate the instantaneous flow rate of the aluminum-silicon alloy eutectoid duct flue gas. The fluid dynamics flow formula is: Instantaneous volumetric flow rate = flue gas velocity × effective cross-sectional area of the duct, and Instantaneous mass flow rate = flue gas velocity × effective cross-sectional area of the duct × actual flue gas density. The instantaneous volumetric flow rate and instantaneous mass flow rate of the flue gas can be calculated simultaneously.
[0046] It is understandable that instantaneous flow rate includes instantaneous volumetric flow rate and instantaneous mass flow rate.
[0047] 105. Integrate the preprocessed data and instantaneous flow rate with the eutectoid cell production process parameters, input them into the LSTM neural network hierarchical early warning model, combine them with the fault knowledge graph to perform operating condition diagnosis, and output the early warning level.
[0048] Specifically, preprocessed data and instantaneous flow rates are integrated with eutectoid tank production process parameters, input into an LSTM neural network hierarchical early warning model, and combined with a fault knowledge graph for operational condition diagnosis, outputting early warning levels, including: The preprocessed data, instantaneous flow rate, and eutectoid tank production process parameters are fused to obtain fused data.
[0049] The production process parameters include the aluminum-silicon ratio of raw materials, electrode working current, and melting temperature in the bath.
[0050] Specifically, due to differences in sampling frequencies among different data sources—for example, the instantaneous flow rate sampling frequency is 1Hz, the electrode operating current is 0.5Hz, and the aluminum-silicon ratio is recorded every 10 minutes—it is necessary to first unify all data to a time step of 1 minute / step using linear interpolation.
[0051] In this process, high-frequency data is downsampled and averaged, while low-frequency data is filled with values at adjacent time points using linear interpolation. After time alignment, each sequence is normalized using min-max normalization, mapping the values to the [0,1] interval. Finally, all normalized variables are concatenated along the time dimension, and a multi-dimensional vector including preprocessed data, instantaneous flow rate, aluminum-silicon ratio, electrode current, and smelting temperature is output at each time step, completing the data fusion.
[0052] Multi-dimensional features, including time-domain and frequency-domain features, are extracted from the fused data. The time-domain features include the mean, variance, mutation rate, and duration of continuous anomalies in the data, while the frequency-domain features include the main frequency and spectral energy after Fourier transform.
[0053] Specifically, a sliding window method is used to extract time-domain and frequency-domain features from the fused data. The window size is set to 10 minutes, and the step size is 1 minute. During time-domain feature extraction, four types of indicators are calculated for all data sequences within each window. These four types of indicators include: Mean: The arithmetic mean of the data points within the window, reflecting the central tendency.
[0054] Variance: The average of the squared differences between data points and the mean, representing the degree of fluctuation.
[0055] Mutation rate: Calculated by first-order difference, it represents the percentage of times the mutation rate exceeds 0.1 within a statistical window; Duration of consecutive abnormalities: Based on the 3σ principle, the maximum number of time points within the statistical window that continuously exceed the normal range.
[0056] For frequency domain feature extraction, fast Fourier transforms (FFTs) need to be performed on the instantaneous flow rate and smelting temperature sequences respectively. First, the window data is detrended, then an FFT of length equal to the window size is used to calculate the spectrum. When extracting the dominant frequency, the frequency component with the largest amplitude in the spectrum is obtained; when extracting the spectral energy, the sum of squares of the amplitudes of the frequency components is calculated. Finally, all time-domain and frequency-domain features are concatenated into a multi-dimensional feature vector to complete feature extraction.
[0057] By inputting multi-dimensional features into the LSTM neural network hierarchical early warning model, preliminary diagnostic results of the working conditions are obtained.
[0058] In this process, inputting multi-dimensional features into the LSTM neural network hierarchical early warning model requires constructing the LSTM neural network hierarchical early warning model. The LSTM neural network hierarchical early warning model includes an input layer, an LSTM layer, a Dropout layer, a fully connected layer, and an output layer.
[0059] The dimension of the input layer is the same as that of the multi-dimensional feature vector. For example, if the dimension of the multi-dimensional feature vector is 5 variables × 4 time-domain features + 2 variables × 2 frequency-domain features = 24 dimensions, then the dimension of the input layer is 24 dimensions.
[0060] The LSTM layer is set to 2 layers, with 128 hidden units in the first layer and 64 in the second layer. The activation function is tanh, which is used to capture long-term dependencies in time series.
[0061] The Dropout layer adds one layer after each of the two LSTM layers, with a Dropout rate of 0.2, randomly discarding neurons to prevent overfitting.
[0062] The fully connected layer uses the ReLU activation function to map the LSTM output to a low-dimensional space.
[0063] The output layer is activated with Softmax and outputs four dimensions, including the probability values of normal, level 1, level 2, and level 3 warnings.
[0064] During model training, historical labeled data was used, and the training, validation, and test sets were divided in a 7:2:1 ratio. The loss function was multi-class cross-entropy, the optimizer was Adam, the initial learning rate was 0.001, the training epochs were 100, and an early stopping strategy was adopted.
[0065] After the LSTM neural network hierarchical early warning model is trained, the multidimensional feature vector is input into the LSTM neural network hierarchical early warning model, and the level with the highest probability is taken as the preliminary diagnosis result of the working condition.
[0066] The preliminary diagnosis results are matched with the knowledge graph of aluminum-silicon alloy eutectoid tank faults. The preliminary diagnosis results are corrected based on the matching results, and the operating condition diagnosis results and corresponding warning levels are output.
[0067] Specifically, the knowledge graph of aluminum-silicon alloy eutectoid cell faults adopts an ontology instance structure. The ontology layer defines four types of nodes: fault type, fault cause, fault characteristic, and treatment measures, as well as the relationship edges between fault type and fault characteristic, and between fault type and fault cause. The instance layer is populated based on expert knowledge and historical data. For example, the fault types include instances such as excessive electrode wear and abnormally high melting temperature, and the fault characteristics include instances such as electrode current variance > 0.5 and continuous abnormal temperature duration > 5 minutes. The fault types and characteristics are associated through corresponding edges.
[0068] During the matching and correction process, features exceeding the normal range in the preliminary diagnosis results are first extracted as a feature set. The cosine similarity between this feature set and the corresponding feature sets for each fault type in the knowledge graph is then calculated. If the similarity exceeds 0.8, the fault type is considered a candidate. If multiple candidates exist, they are ranked using LSTM probability weighting, and the one with the highest overall score is selected as the final operational condition diagnosis result. Finally, based on the severity attribute of the fault type in the knowledge graph, a corresponding warning level is output, completing the diagnosis and warning output.
[0069] 106. Based on the instantaneous flow rate of each eutectoid cell, optimize the operation strategy of the exhaust fan using a genetic algorithm and output scheduling instructions.
[0070] Specifically, based on the instantaneous flow rate of each eutectoid cell, a genetic algorithm is used to optimize the operation strategy of the exhaust fan and output scheduling instructions, including: Data on instantaneous flue gas flow rate and flue static pressure in parallel production scenarios of the eutectoid cell were collected. At the same time, fan equipment parameters, including rated power, speed-volume characteristic curves, and energy consumption characteristic curves of the exhaust fan, were obtained.
[0071] In the parallel production scenario of the co-extrusion tank, the instantaneous flow rate and static pressure data of flue gas are collected by an integrated intelligent measuring instrument. The speed-volume characteristic curve and energy consumption characteristic curve are obtained by the performance test report of the fan or by on-site measurement using frequency conversion speed regulation + air volume meter + power analyzer.
[0072] With the objective function of minimizing the total energy consumption of the exhaust fans and balancing the exhaust load of each fan, and with the rated speed and maximum air volume of the fans as constraints, an optimization model for the operation of exhaust fans based on a genetic algorithm is constructed.
[0073] An optimization model based on a genetic algorithm is constructed with the objectives of minimizing the total energy consumption of the exhaust fans and balancing the exhaust load of each fan. The objective function is transformed from a multi-objective model into a single objective using a weighted summation method. The objective function is: ; in, The number of operating wind turbines, For the first The energy consumption of typhoon generators For the first The actual air volume of the typhoon fan. For the first The rated air volume of the typhoon fan The variance of the load factor, , This is a weighting coefficient, set according to on-site requirements.
[0074] The constraints include fan speed constraints, air volume supply and demand balance constraints, and standby fan start-stop logic constraints: standby fans must be included in the operating fan group after starting, and are not included in energy consumption calculations after stopping.
[0075] The instantaneous flow rate of the eutectoid cell, the static pressure data of the flue, and the parameters of the fan equipment are input into the exhaust fan operation optimization model. The model is solved by encoding, crossover, mutation and iteration through a genetic algorithm to obtain the operating power adjustment value, speed adjustment value and standby fan start-up and shutdown strategy of each exhaust fan.
[0076] Specifically, the coding design uses a hybrid real-number and binary encoding. Each individual is a vector, with the first N bits being real numbers representing the rotational speed of the operating fan and the last K bits being binary numbers representing the start / stop status of the K standby fans. The population size is set to 50. The rotational speed of the operating fan is randomly generated within the minimum and rated values, and the start / stop status of the standby fans is randomly assigned a value of 0 or 1.
[0077] The reciprocal of the objective function value is used as the fitness, with higher fitness indicating a better individual. Genetic operations employ a roulette wheel selection method, choosing parents based on fitness ratios. In the crossover operation, arithmetic crossover is used for the real-valued portion, and single-point crossover is used for the binary portion. In the mutation operation, Gaussian mutation is used for the real-valued portion, and bit-flip mutation is used for the binary portion.
[0078] For individuals exceeding the speed range, the speed is forcibly corrected to the boundary value; for individuals that do not meet the airflow balance requirement, the speed of all working fans is increased proportionally until the constraint is met.
[0079] When the fitness change over 20 consecutive generations is less than If the maximum number of iterations reaches 100, the optimal individual will be output, which will be decoded into the speed adjustment value of each wind turbine and the start-stop strategy of the standby wind turbine.
[0080] The operating power adjustment value, speed adjustment value, and standby fan start-stop strategy are converted into industrial control scheduling instructions and sent to the frequency conversion control system and start-stop control system of the exhaust fan.
[0081] Specifically, since the fan speed is proportional to the inverter frequency, the speed adjustment value is converted into the target frequency. For standby fans that need to be started, a start command is first sent through the PLC's digital output module, and after a 30-second delay, a target frequency command is sent to the inverter. For fans that need to be stopped, the frequency is first reduced to the frequency corresponding to the lowest stable speed, and then a stop command is sent.
[0082] All control commands are sent to the field frequency converters and PLCs via Modbus TCP or PROFINET industrial bus. A feedback mechanism is also implemented: the system waits for the frequency converter to return an acknowledgment signal indicating that the frequency has reached the target value ±0.5Hz. If no acknowledgment is received within 5 seconds, the command is resent, up to a maximum of 3 resentments. If resentment fails, a local audible and visual alarm is triggered, and the abnormal information is uploaded to the server. Ultimately, through the coordinated operation of the frequency converter control system and the start / stop control system, optimized operation of the exhaust fan is achieved.
[0083] In some embodiments, it also includes: The flue gas parameters, instantaneous flow rate, warning level, and dispatch instructions are displayed dynamically in multiple dimensions on the front-end interface.
[0084] The system features a multi-dimensional dynamic display on the front-end interface, allowing operators to monitor the real-time operating status of the aluminum-silicon alloy eutectoid tank flue gas flow monitoring system. The display includes trend curves of flue gas parameters, real-time instantaneous flow values, color-coded warning levels, and feedback on the execution of dispatch commands. The trend curves, with time as the horizontal axis, clearly show the changing patterns of key parameters such as flue gas temperature, pressure, and humidity. Real-time values are highlighted and dynamically refreshed to ensure instantaneous flow information is readily available. Warning levels are indicated by red, orange, and yellow to differentiate different degrees of operational anomalies, accompanied by brief text descriptions. The execution feedback section records in detail the issuance time of dispatch commands, target equipment, and completion status, providing comprehensive decision support for operators.
[0085] In some embodiments, it also includes: The flue gas parameters and instantaneous flow rate are stored as time series data in the InfluxDB time series database; The process parameters, early warning records, and scheduling instructions for the eutectoid tank are stored as structured data in a MySQL database.
[0086] Specifically, the InfluxDB time-series database meets the high-frequency acquisition and real-time analysis needs of flue gas parameters and instantaneous flow data through its efficient write and query performance. Its flexible data retention strategy automatically manages historical data according to the monitoring cycle, ensuring optimal utilization of storage resources. Meanwhile, InfluxDB supports rich aggregate functions and time window operations, facilitating the generation of trend curves and statistical reports, providing a reliable basis for comprehensive evaluation of system operation status. The portion of structured data stored in the MySQL database is designed with standardization in mind, ensuring clear and traceable correlations between production process parameters, early warning records, and scheduling instructions, thereby improving the overall efficiency and accuracy of data management.
[0087] Based on the same inventive concept, this invention also protects an aluminum-silicon alloy eutectoid cell flue gas flow monitoring system. The aluminum-silicon alloy eutectoid cell flue gas flow monitoring system provided by this invention will be described below. The aluminum-silicon alloy eutectoid cell flue gas flow monitoring system described below and the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method described above can be referred to in correspondence with each other.
[0088] In some embodiments, such as Figure 2 As shown, some embodiments of the present invention also provide an aluminum-silicon alloy eutectoid cell flue gas flow monitoring system, comprising: The integrated intelligent measuring instrument 210 is installed in the main flow area of the aluminum-silicon alloy eutectoid flue to collect flue gas parameters of the flue section and calibrate the flue gas composition humidity compensation coefficient. The host computer 220 communicates with the integrated intelligent measuring instrument. The host computer includes: The filtering module 2201 is used to perform adaptive Kalman median mean fusion sliding window filtering on the received flue gas parameters to obtain preprocessed data; The flow calculation module 2202 is used to calculate the instantaneous flow rate by inputting the pre-processed data into the dual-compensation density model according to the pre-calibrated flue gas composition humidity compensation coefficient. The operating condition early warning module 2203 is used to integrate preprocessed data and instantaneous flow rate with the production process parameters of the eutectoid tank, input the data into the LSTM neural network hierarchical early warning model, combine the fault knowledge graph to perform operating condition diagnosis, and output the early warning level. The collaborative scheduling module 2204 is used to optimize the operation strategy of the exhaust fan and output scheduling instructions based on the instantaneous flow rate of each eutectoid tank using a genetic algorithm.
[0089] Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention.
[0090] like Figure 3 As shown, the electronic device may include a processor 310, a communication interface 320, a memory 330, and a communication bus 340. The processor 310, communication interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions from the memory 330 to execute a flue gas flow monitoring method for an aluminum-silicon alloy eutectoid tank.
[0091] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0092] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method provided by the above methods.
[0093] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the aluminum-silicon alloy eutectoid cell flue gas flow monitoring method provided by the above methods.
[0094] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for monitoring flue gas flow rate in an aluminum-silicon alloy eutectoid cell, characterized in that, include: An integrated intelligent measuring instrument was installed in the aluminum-silicon alloy eutectoid cell, and the humidity compensation coefficient of the flue gas composition was calibrated. The flue gas parameters of the flue section are collected by the integrated intelligent measuring instrument. The flue gas parameters are subjected to adaptive Kalman median-mean fusion sliding window filtering to obtain preprocessed data; Based on the flue gas composition humidity compensation coefficient, the preprocessed data is input into the dual compensation density model to calculate the instantaneous flow rate; The preprocessed data and the instantaneous flow rate are fused with the eutectoid tank production process parameters, input into an LSTM neural network hierarchical early warning model, and combined with a fault knowledge graph for operating condition diagnosis, and the early warning level is output. Based on the instantaneous flow rate of each eutectoid cell, the operation strategy of the exhaust fan is optimized by a genetic algorithm and a scheduling command is output.
2. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, The integrated intelligent measuring instrument integrates a multi-array micro probe, a self-cleaning module, a constant temperature heating module, and a sensor self-sensing module. The integrated intelligent measuring instrument is installed in a single hole in the mainstream gas state area of the aluminum-silicon alloy eutectoid flue. The multi-array microprobe consists of three sets of microsensor arrays arranged in a triangular pattern. The multi-array microprobe simultaneously collects temperature, static pressure, differential pressure, flue gas solids concentration, and humidity data at multiple points on the flue gas cross-section. The flue gas solids concentration is the mixed concentration of alumina and silica dust. The self-cleaning module cleans the probe by blowing the pressure tapping hole with a high-pressure air knife and then vibrating the probe surface with ultrasonic waves according to a preset cycle or differential pressure signal fluctuation threshold. The constant temperature heating module controls the probe temperature to be 10-15°C higher than the flue gas dew point temperature. The sensor self-sensing module collects the working status data of the multi-array microprobe, constructs a sensor health assessment model based on the working status data, and generates a software compensation coefficient based on the working status data when data drift occurs in the data collected by the multi-array microprobe, thereby completing remote online calibration.
3. The method for monitoring flue gas flow in a silicon-aluminum alloy eutectoid cell according to claim 1, characterized in that, The flue gas parameters are subjected to adaptive Kalman median-mean fusion sliding window filtering to obtain preprocessed data, including: Calculate the fluctuation coefficient of the flue gas parameter, where the fluctuation coefficient is the ratio of the standard deviation to the mean of the flue gas parameter; The sliding window capacity is dynamically adjusted based on the fluctuation coefficient. The flue gas parameters are input into a Kalman filter to filter out continuous electromagnetic interference and flue gas steady-state pulsation noise in the industrial field, and the first filtering data is obtained. The filtered data is input into the sliding window capacity. The data within the sliding window is sorted, and the maximum and minimum values of 1 / 3 after sorting are removed. The arithmetic mean of the remaining 1 / 3 of the data is calculated to obtain the preprocessed data.
4. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 3, characterized in that, Dynamically adjusting the sliding window capacity based on the fluctuation coefficient includes: If the fluctuation coefficient is less than 5%, the sliding window capacity will be adjusted to the first preset number of data points. If the fluctuation coefficient is between 5% and 15%, the sliding window capacity will be adjusted to the second preset number of data points. If the fluctuation coefficient is greater than 15%, the sliding window capacity will be adjusted to the third preset number of data points.
5. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Based on the flue gas composition humidity compensation coefficient, the preprocessed data is input into a dual-compensation density model to calculate the instantaneous flow rate, including: The temperature, pressure, flue gas solids concentration, humidity data, and flue gas composition humidity compensation coefficient in the preprocessed data are input into the double-compensation flue gas density model to calculate the corrected actual flue gas density. The dual-compensation flue gas density model is as follows: ； in, For actual flue gas density, The density of flue gas under standard conditions. , To measure the flue gas pressure and temperature, , For pressure and temperature under standard conditions, This refers to the concentration of solids in the flue gas. For the humidity of the flue gas, The component compensation coefficient, This is the humidity compensation coefficient; The flue gas velocity data is extracted from the preprocessed data, and combined with the effective cross-sectional area of the eutectoid flue and the actual flue gas density, the instantaneous flow rate of the flue gas is calculated using the fluid dynamics flow rate formula.
6. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, The preprocessed data and the instantaneous flow rate are fused with the eutectoid tank production process parameters, input into an LSTM neural network hierarchical early warning model, and combined with a fault knowledge graph for operating condition diagnosis, outputting early warning levels, including: The preprocessed data, the instantaneous flow rate, and the eutectoid cell production process parameters are fused to obtain fused data; the production process parameters include the raw material aluminum-silicon ratio, electrode operating current, and in-cell melting temperature. Multi-dimensional features, including time-domain and frequency-domain features, are extracted from the fused data; the time-domain features are the mean, variance, mutation rate, and duration of continuous anomalies in the data, and the frequency-domain features are the main frequency and spectral energy after Fourier transform. The multi-dimensional features are input into the LSTM neural network hierarchical early warning model to obtain preliminary working condition diagnosis results; The preliminary diagnostic results are matched with the knowledge graph of aluminum-silicon alloy eutectoid tank faults. The preliminary diagnostic results are then corrected based on the matching results, and the operating condition diagnostic results and corresponding warning levels are output.
7. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Based on the instantaneous flow rate of each eutectoid cell, a genetic algorithm is used to optimize the operation strategy of the exhaust fan and output scheduling instructions, including: Collect instantaneous flow rate and static pressure data of flue gas in the parallel production scenario of the eutectoid cell, and acquire fan equipment parameters including rated power, speed-volume characteristic curve, and energy consumption characteristic curve of the exhaust fan. With the objective function of minimizing the total energy consumption of the exhaust fans and balancing the exhaust load of each fan, and with the rated speed and maximum air volume of the fans as constraints, an optimization model for the operation of exhaust fans based on genetic algorithm is constructed. The instantaneous flow rate of the eutectoid tank, the static pressure data of the flue, and the parameters of the fan equipment are input into the exhaust fan operation optimization model. The model is solved by encoding, crossover, mutation and iteration through a genetic algorithm to obtain the operating power adjustment value, speed adjustment value and standby fan start-up and shutdown strategy of each exhaust fan. The operating power adjustment value, the speed adjustment value, and the standby fan start-stop strategy are converted into industrial control scheduling instructions and sent to the frequency conversion control system and start-stop control system of the exhaust fan.
8. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Also includes: The flue gas parameters, instantaneous flow rate, warning level, and dispatch instructions are dynamically displayed in multiple dimensions on the front-end interface.
9. The method for monitoring flue gas flow in an aluminum-silicon alloy eutectoid cell according to claim 1, characterized in that, Also includes: The flue gas parameters and instantaneous flow rate are stored as time-series data in the InfluxDB time-series database; The process parameters, early warning records, and scheduling instructions for the eutectoid tank are stored as structured data in a MySQL database.
10. A flue gas flow monitoring system for an aluminum-silicon alloy eutectoid cell, characterized in that, include: An integrated intelligent measuring instrument is installed in the main flow area of the aluminum-silicon alloy eutectoid flue to collect flue gas parameters of the flue section and calibrate the flue gas composition humidity compensation coefficient. A host computer is communicatively connected to the integrated intelligent measuring instrument, and the host computer includes: The filtering module is used to perform adaptive Kalman median-mean fusion sliding window filtering on the received flue gas parameters to obtain preprocessed data; The flow calculation module is used to input the preprocessed data into the dual-compensation density model to calculate the instantaneous flow rate based on the pre-calibrated flue gas composition humidity compensation coefficient. The operating condition early warning module is used to integrate the preprocessed data and the instantaneous flow rate with the eutectoid tank production process parameters, input them into the LSTM neural network hierarchical early warning model, combine them with the fault knowledge graph to perform operating condition diagnosis, and output the early warning level. The collaborative scheduling module is used to optimize the operation strategy of the exhaust fan and output scheduling instructions based on the instantaneous flow rate of each eutectoid tank using a genetic algorithm.