An on-line detection method and system for FeO content of sinter

Abstract

The application discloses a kind of sinter FeO content online detection method, comprising the following steps, S1, the first weighing area and the second weighing area are divided on the conveying belt for transmitting sinter, first and second weighing area are provided for the sinter that passes through the weighing device for weighing, wherein the second weighing area is provided with the magnetic force device that the magnetic substance in sinter is applied with magnetic force, speed measuring device is provided on conveying belt for obtaining transmission speed;S2, when sinter is transmitted on conveying belt, the weight of sinter that passes through first weighing area and second weighing area is acquired in real time, and the transmission speed of conveying belt;S3, according to the sinter weight data and conveying belt motion data acquired in S2, the SSA-BP neural network of pre-training is used to detect the FeO content in sinter;S4, whether the FeO content in sinter is in preset range is judged, if exceed preset range, then alarm is carried out.The method realizes the real-time accurate monitoring of sinter FeO content.

Description

Technical Field

[0001] This invention relates to the field of iron and steel smelting, and specifically to an online detection method and system for FeO content in sintered ore based on an SSA-BP neural network. Background Technology

[0002] With the continuous development of the metallurgical industry, natural rich ore resources are limited, and their metallurgical performance is not as superior as that of artificial lump ore—sinter. Therefore, most modern blast furnaces use sinter, or sinter mixed with a small amount of natural rich ore for smelting. More than 80% of the raw materials in industrial furnaces come from sinter. As the main raw material for blast furnace ironmaking, the quality of sinter affects the blast furnace's smelting performance. The mineral composition of sinter varies depending on the raw materials and sintering process conditions, and can be divided into acidic sinter minerals, self-fluxing sinter minerals, and high-basicity sinter minerals. However, the main components of sinter are magnetite (Fe3O4), hematite (Fe2O3), flocculation (FeO·x), fir olivine (FeO·SiO2), calcium iron olivine (CaO·FeO·SiO2), calcium ferrite (CaO·Fe2O3), magnesium ferrite (MgO·Fe2O3), quartz (SiO2), and lime (CaO), etc.

[0003] FeO content is a relatively important indicator of sinter. The level of FeO content in sinter directly affects the particle size and cold strength of blast furnace sinter raw materials, thereby affecting the reducibility of sinter.

[0004] Currently, methods for measuring FeO content in sintered ore are mainly divided into two categories: online measurement methods and offline measurement methods. Offline measurement methods include: manual visual judgment, potassium dichromate chemical analysis, and X-ray spectroscopy. Online measurement methods include: online image recognition real-time detection, online real-time detection of exhaust gas temperature and composition, online real-time detection of magnetic permeability, and online thermal imaging recognition.

[0005] The manual visual judgment method involves judging by observing the color of the flame. However, this method is highly subjective, resulting in significant deviations in the test results, and the results cannot be quantified.

[0006] The potassium dichromate chemical analysis method determines the FeO content in sinter by titration. Although this method is accurate, it has a long lag and cannot reflect the sintering quality in a timely manner, making it difficult to participate in the real-time control of the blast furnace smelting process.

[0007] X-ray spectroscopy uses subtle differences in the angle and intensity of rays for quantitative analysis. However, due to the overlap of rays during actual measurement, the accuracy of the results is not high.

[0008] The online image recognition real-time detection method uses the image features of the sintering machine tail to determine the FeO content. Although this reduces the labor intensity of the workers to some extent, the harsh environment at the blast furnace smelting site seriously affects data acquisition.

[0009] The online real-time detection method for exhaust gas temperature and composition calculates the FeO content in sintered ore using correlation formulas. However, due to the constant changes in physical conditions at the blast furnace smelting site, the measured results differ from the actual situation.

[0010] The online real-time magnetic permeability detection method calculates the FeO content of sinter by detecting the magnetic permeability of ferromagnetic materials in sinter and using the correlation between magnetic permeability and FeO content. However, the detection device is actually an offline measurement device, which requires grinding and pressing the sample into a sheet, and the measurement results are not accurate enough due to errors.

[0011] Online thermal imaging identification uses image processing methods to extract features, but it only extracts image information and cannot directly reflect the sintering state, so the measurement results are not accurate enough. Summary of the Invention

[0012] The purpose of this invention is to provide an online detection method and system for FeO content in sinter, which enables real-time and accurate monitoring of FeO content in sinter.

[0013] To solve the above-mentioned technical problems, the present invention provides a technical solution: an online detection method for FeO content in sintered ore, comprising the following steps:

[0014] S1. Divide the conveyor belt used for transporting sinter into a first weighing area and a second weighing area. Weighing devices for weighing the passing sinter are set on the first and second weighing areas. The second weighing area is equipped with a magnetic device that applies magnetic force to the magnetic material in the sinter. A speed measuring device for obtaining the transmission speed is set on the conveyor belt.

[0015] S2. When the sinter is being transported on the conveyor belt, the weight of the sinter passing through the first weighing area and the second weighing area, as well as the conveyor belt speed, are obtained in real time.

[0016] S3. Based on the sinter weight data and conveyor belt motion data obtained in S2, the FeO content in the sinter is detected using a pre-trained SSA-BP neural network.

[0017] S4. Determine whether the FeO content in the sinter is within the preset range. If it exceeds the preset range, an alarm will be triggered.

[0018] The online detection method for FeO content in sintered ore according to claim 1 is characterized in that: the training steps of the SSA-BP neural network are as follows:

[0019] S301. Initialize the SSA-BP neural network by setting parameters including weight range, error function, computational precision, and maximum number of learning iterations.

[0020] S302. Initialize sparrow population parameters and define the maximum number of iterations;

[0021] S303. Input the training data into the SSA-BP neural network, collect and distribute the population fitness, and determine the optimal position of the current population; wherein, the training data are the weight values ​​of the first weighing area, the weight values ​​of the second weighing area, and the transmission speed of the conveyor belt for each batch of sintered ore with different quality and known FeO content.

[0022] S304. Obtain the optimal weights and thresholds using the sparrow search algorithm and assign them to the SSA-BP neural network. Calculate the inputs and outputs of each neuron in the hidden and output layers.

[0023] S305. The actual output value of the output layer is compared with the expected value, and the error E is calculated.

[0024] S306. When the error E is within the preset range, it is determined that the error E meets the requirements, and the current SSA-BP neural network is used as the final prediction model; otherwise, it is determined whether the current number of learning iterations has reached the maximum number of learning iterations. If it has reached the maximum number of learning iterations, the current SSA-BP neural network is used as the final prediction model. If it has not reached the maximum number of learning iterations, it returns to S304 for iterative iteration.

[0025] According to the above scheme, S304 specifically refers to...

[0026] The output of the i-th node in the input layer is x i , i = (1,2,3);

[0027] Where x1, x2, and x3 are the weight values ​​of the first weighing area, the weight values ​​of the second weighing area, and the conveying speed of the conveyor belt for each batch of sintered ore with different masses and known FeO content, respectively.

[0028] The output of the j-th node in the hidden layer is

[0029] Among them, w ij It is the connection weight between the i-th node in the input layer and the j-th node in the hidden layer, b j f1() is the threshold at the j-th node of the hidden layer, and f1() is the activation function of the hidden layer node. The number of nodes in the hidden layer is mainly selected based on the following formula: j = log2(m + n), where j represents the number of nodes in the hidden layer, and m and n represent the number of nodes in the input layer and the output layer, respectively.

[0030] According to the above scheme, in S305, the method for calculating the error E is as follows: The error e of the kth node in the output layer k =O k -y k O k It is the expected result of the k-th node in the output layer, y k It is the output result of the kth node of the output layer; the expected value is the known FeO content of the sinter.

[0031] According to the above scheme, the training data is randomly divided into training set, validation set and test set at a ratio of 70%, 15% and 15% respectively.

[0032] According to the above scheme, the preset range corresponding to error E is ±10%.

[0033] An online detection system for FeO content in sintered ore, comprising,

[0034] Conveyor belt; the conveyor belt is equipped with a first weighing area and a second weighing area;

[0035] Weighing device; including weighing sensors disposed in the first and second weighing areas of the conveyor belt, used to acquire the weight of sinter when it is conveyed to the first and second weighing areas.

[0036] A magnetic device is installed in the second weighing area to apply magnetic force to magnetic materials in the sinter that have passed through the second weighing area.

[0037] Speed ​​measuring device; used to obtain the conveyor speed of the conveyor belt;

[0038] The monitoring platform is electrically connected to the weighing and speed measuring devices. It includes a pre-trained SSA-BP neural network, which is used to determine the FeO content in the sinter based on the real-time acquired sinter weight data and the conveyor belt speed, and to issue an alarm when the FeO content in the sinter exceeds a preset range.

[0039] According to the above scheme, the magnetic device is a neodymium iron boron magnet.

[0040] According to the above scheme, the speed measuring device includes a speed measuring wheel installed on the return section of the conveyor belt.

[0041] The beneficial effects of this invention are as follows: This invention improves the convergence speed of the BP neural network by using the Sparrow Search Algorithm (SSA). Based on the self-learning and adaptive capabilities of the BP neural network, the pre-trained SSA-BP neural network is used to predict the FeO content in the sintered ore under real-time detection. It can deeply explore the complex relationship between the output and the input data, and can more accurately approximate complex nonlinear functions, thereby achieving the goal of online real-time and accurate detection of the FeO content in sintered ore, improving the accuracy of the detection results and enhancing their portability. Attached Figure Description

[0042] Figure 1 This is a flowchart of an online detection method for FeO content in sinter based on an SSA-BP neural network according to an embodiment of the present invention;

[0043] Figure 2 This is a flowchart of data transmission to a monitoring platform according to an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram showing the approximate composition of sinter produced by steel plants.

[0045] Figure 4 The results of magnetic simulation of sinter with different Fe3O4 contents under a constant magnetic field are shown in the figure.

[0046] Figure 5 This is a magnetic simulation diagram of sinter with a fixed Fe3O4 content, when the distance between the NdFeB permanent magnet and the sinter is adjusted.

[0047] Figure 6 This is a flowchart of the SSA-BP neural network training process according to an embodiment of the present invention;

[0048] Figure 7 This is a simplified diagram of a neural network structure according to an embodiment of the present invention;

[0049] Figure 8 This is a schematic diagram of the overall system device according to an embodiment of the present invention;

[0050] Figure 9 This is a comparison chart of experimental results from one embodiment of the present invention;

[0051] Figure 10 This is an absolute error diagram according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure 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 disclosure. All other embodiments obtained by those skilled in the art based on the described embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0053] This embodiment provides an online detection method for FeO content in sintered ore based on SSA-BP neural network;

[0054] like Figure 3 As shown, the main components of sinter are magnetite (Fe3O4), hematite (Fe2O3), plutonium (FeO·x), fir olivine (FeO·SiO2), calcium iron olivine (CaO·FeO·SiO2), calcium ferrite (CaO·Fe2O3), magnesium ferrite (MgO·Fe2O3), quartz (SiO2), and lime (CaO). FeO exists in sinter in the forms of magnetite, fir olivine, calcium iron olivine, and plutonium, with magnetite (Fe3O4) being the primary form. The main magnetic material in sinter is magnetite (Fe3O4), which exhibits ferrimagnetism; the remaining materials are weakly magnetic. Therefore, by measuring the magnetite content in sinter, the FeO content can be determined using a neural network.

[0055] The formula for calculating the attractive force between ferrimagnetic materials and neodymium iron boron permanent magnets is as follows:

[0056] B represents the magnetic induction intensity at the interface between the magnetic field generated by the permanent magnet and the sintered ore, and S represents the cross-sectional area of ​​the NdFeB permanent magnet. Because the magnetization capability of the magnetic field generated by the NdFeB permanent magnet has a certain range... Figure 4 The image shows the magnetic simulation results of sintered ores with different Fe3O4 contents under a constant magnetic field. Figure 5 A magnetic simulation diagram of sinter with fixed Fe3O4 content when adjusting the distance between the NdFeB permanent magnet and the sinter; through... Figure 4 and Figure 5 It can be seen that as the Fe3O4 content in the sinter increases, the magnetic force on the sinter gradually increases, while as the distance between the NdFeB permanent magnet and the sinter increases, the magnetic force on the sinter gradually decreases. Therefore, adjusting the position appropriately can ensure that the sinter is fully magnetized; in this embodiment, a spacing of 20 cm is used.

[0057] like Figure 1As shown, an online detection method for FeO content in sintered ore based on SSA-BP neural network is presented. This method is based on the existence forms and characteristics of FeO in sintered ore. It employs a magnetic method to measure relevant data using a weighing device and a velocity measuring device. A pre-trained SSA-BP neural network is used to construct an optimal prediction model for FeO content in sintered ore, thereby detecting the FeO content. The method includes:

[0058] S101: The weight data of sinter in the first weighing area, the weight data of sinter magnetized by neodymium iron boron magnets in the second weighing area, and the speed data of the conveyor belt rotating at a constant speed are obtained in real time by the weighing device and speed measuring device installed on the belt conveyor.

[0059] S102: The weight data of the sinter in the first weighing area, the weight data of the sinter magnetized by the neodymium iron boron strong magnet in the second weighing area, and the speed data of the conveyor belt rotating at a constant speed are transmitted to the monitoring platform.

[0060] S103: The pre-trained SSA-BP neural network in the monitoring platform processes various data and detects the FeO content of sintered ore through neural network algorithms;

[0061] S104: The pre-trained SSA-BP neural network outputs the FeO content of sinter and indicates whether the FeO content of sinter meets the requirements (the FeO content of sinter is 7%-12%, which meets the requirements); if the FeO content of sinter does not meet the requirements (the FeO content value of sinter is less than 7% or greater than 12%), the monitoring platform will issue an alarm.

[0062] As one or more embodiments, in S101, the weight and speed data of sintered ore at a specific location on the conveyor belt in the industrial site are acquired in real time. Specific steps include:

[0063] By using weighing and speed measuring devices installed under the conveyor belt, the weight and speed data of sintered ore at specific locations on the conveyor belt in an industrial setting can be collected.

[0064] Furthermore, this system takes the industrial site as the research area, and the overall system diagram is as follows: Figure 8 As shown. The weighing device includes: a plurality of weighing sensors, a microcontroller, wherein the plurality of weighing sensors are respectively connected to a junction box, the junction box is connected to a weighing force measuring instrument, the weighing force measuring instrument is connected to the microcontroller, and the weight data measured by the weighing sensors is converted from analog signal to digital signal by the weighing force measuring instrument, and the conversion result is collected by the microcontroller;

[0065] The speed measuring device includes: a measuring wheel, an analog-to-digital converter module, and a microcontroller. The measuring wheel is installed on the return section of the conveyor belt and connected to a speed sensor. The speed sensor is connected to the microcontroller. The speed data measured by the speed sensor is converted from an analog signal to a digital signal through the analog-to-digital converter module, and the conversion result is collected by the microcontroller.

[0066] As one or more embodiments, in S102, the acquired weight data and speed data are transmitted to a monitoring platform, which includes a serial communication module, a data processing module, a database management module, a data display module, and a system parameter setting module; the specific steps include:

[0067] The weighing device and the speed measuring device respectively acquire weight data and speed data, which are converted from analog signals to digital signals, through a microcontroller;

[0068] The microcontroller communicates with the monitoring platform via a serial communication module, and the monitoring platform processes the transmitted weight and speed data via a data processing module.

[0069] Specifically, in S102, as Figure 2 As shown, the acquired weight and speed data are transmitted to the monitoring platform. The purpose of serial communication is not only to transmit data information, but more importantly, to ensure accurate transmission. The specific steps include:

[0070] S1021: The system powers on and performs microcontroller initialization, LCD display initialization, and monitoring platform serial port initialization.

[0071] S1022: Collects weight and speed data converted from analog signals to digital signals via a microcontroller;

[0072] S1023: Use the serial port interrupt function to determine if there is data input. If there is data input, display it on the LCD screen. If there is no data input, return to the previous step.

[0073] S1024: Use a data error checking method to determine whether the data displayed on the LCD screen is valid. If the data is valid, perform serial communication; if the data is incorrect, return to the previous step.

[0074] S1025: The collected data is uploaded to the monitoring platform via serial port.

[0075] As one or more embodiments, in S103, a pre-trained SSA-BP neural network is used; the training steps include:

[0076] The first step is to construct a training set, a test set, and a validation set. The training set, test set, and validation set consist of batches of sinter with known FeO content and different masses. The weight data of the sinter in the first weighing area, the weight data of the sinter magnetized by neodymium iron boron magnets in the second weighing area, and the speed data of the conveyor belt rotating at a constant speed are obtained in real time through weighing and speed measuring devices installed at specific positions on the belt conveyor.

[0077] The second step is to construct a BP neural network, input the training data, and perform normalization processing.

[0078] The third step is to determine the topology of the BP neural network and initialize the weights and thresholds of the BP neural network.

[0079] The fourth step is to calculate the fitness of the population and the optimal individual based on the Sparrow Search Algorithm (SSA).

[0080] The fifth step involves adjusting an individual's location through foraging and anti-predation behaviors;

[0081] Step 6: When the termination criterion is met, output the optimal weight and threshold parameters;

[0082] The seventh step is to obtain the optimal parameters of the BP neural network, and the trained SSA-BP neural network is the pre-trained SSA-BP neural network.

[0083] The sparrow search algorithm, based on the sparrows' foraging and anti-predation behaviors, categorizes sparrows within a population into discoverers, joiners, and scouts. Discoverers find food for the entire population and provide foraging directions; joiners obtain food through discoverers; and scouts provide safety vigilance during foraging, signaling to foraging sparrows to move away from any approaching danger. Sparrows adjust their positions promptly based on their roles and the nature of the danger during foraging, adapting to changing circumstances.

[0084] The sparrow population is represented by an N×D matrix X, where D is the dimension of the variable to be optimized, and the number of sparrows is N. The initial positions of the sparrows are then...

[0085] The fitness value of a sparrow population is F = [f(X1)f(X2)...f(X...]. N )] T

[0086] Where f(X) is the fitness function; X c Let f(X) be the position of the c-th sparrow; c ) represents the fitness value of an individual sparrow.

[0087] Furthermore, the location update expression for the discoverer is:

[0088]

[0089] In the formula, t is the current iteration number, t max The maximum number of iterations; d = 1, 2, 3, ..., D, where D represents the dimension of the variable to be optimized; Let t be the position of the c-th sparrow in the d-th dimension after t iterations; max The maximum number of iterations is given; α is a random number in (0,1); R² and ST are the warning value and the safety value, respectively.

[0090] R2∈[0,1],ST∈[0.5,1]; Q is a random number that follows a normal distribution; L is a D-dimensional unit vector.

[0091] Furthermore, the position update expression for the joiner is:

[0092]

[0093] In the formula, X e X is the optimal position currently occupied by the discoverer. worst This is the worst position globally at present. Let A be the optimal position occupied by the discoverer in the (t+1)th iteration; let A be a 1×D matrix where each element is randomly assigned a value of 1 or -1, and A + =A T (AA T ) -1 N represents the number of sparrows.

[0094] Furthermore, the location update expression for the scout is:

[0095]

[0096] In the formula, X t best The current global optimal position; γ is the step size control parameter, a random number following a normal distribution with a mean of 0 and a variance of 1; h∈[-1,1] is a random number; f c f represents the fitness value of the current individual sparrow. g f is the current globally optimal fitness value. w σ represents the worst fitness value globally at present; σ is a constant to avoid zero in the denominator.

[0097] The training flowchart for the SSA-BP neural network is as follows: Figure 6As shown, in this invention, batches of sinter with known FeO content and different masses are conveyed at different speeds on a conveyor belt. The weight and speed data before and after the magnetic field increase are measured. All collected data are divided into training, validation, and test sets. The SSA-BP neural network is then trained sequentially using the training, validation, and test sets to obtain the optimal sinter FeO content prediction model. After obtaining the optimal sinter FeO content prediction model, the input information of the test sample is extracted, and the predicted output of the test sample can be calculated using the optimal sinter FeO content prediction model.

[0098] As one or more embodiments, in S103, a training set, a validation set, and a test set are constructed; the construction steps include:

[0099] The first step involves conveying batches of sinter with different masses and known FeO content at different speeds on a conveyor belt, and measuring the weight and speed data before and after the magnetic field increases the weight.

[0100] The second step is for the microcontroller to collect the weight and speed data before and after the magnetic field increase, and then upload them to the monitoring platform via serial communication.

[0101] The third step involves the monitoring platform randomly dividing all uploaded collected data into training, validation, and test sets at a ratio of 70%, 15%, and 15%, respectively.

[0102] Due to the numerical problems and solution requirements, data normalization is necessary before using the model. The operating mechanism of a BP neural network mainly consists of two parts: the forward propagation of network input information and the back propagation of network computational errors.

[0103] A simplified diagram of a neural network structure is shown below. Figure 7 As shown, the forward transmission of network input information (input layer — output layer):

[0104] The output of the i-th node in the input layer is: x i , i=(1,2,3). Where, x1, x2, x3 are respectively the weight data of sintered ore of different masses with known FeO content in the first weighing area of ​​the industrial site, the weight data of sintered ore magnetized by neodymium iron boron strong magnet in the second weighing area, and the speed data of the conveyor belt rotating at a constant speed, which are obtained in real time by weighing and speed measuring devices installed at specific positions of the belt conveyor.

[0105] The output of the j-th node in the hidden layer is: Among them, w ij It is the connection weight between the i-th node in the input layer and the j-th node in the hidden layer, b jf1() is the threshold at the j-th node of the hidden layer, and f1() is the activation function of the hidden layer node. The number of nodes in the hidden layer is mainly selected based on the following formula: j = log2(m + n), but the number of hidden layer nodes will be modified according to the effect, where j represents the number of nodes in the hidden layer, and m and n represent the number of nodes in the input layer and the output layer, respectively. The selection of the number of hidden layer nodes is not only different from the network structure, but more importantly, it will directly affect the performance of the entire network. In this invention, the number of hidden layer nodes is 5.

[0106] The output of the k-th node in the output layer is: In this invention, the output layer has only one output node, y1, which is the FeO content of the sinter predicted by the neural network algorithm; where w jk It is the connection weight between the j-th node in the hidden layer and the k-th node in the output layer, b k f2() is the threshold at the k-th node of the output layer, and f2() is the activation function of the output layer node;

[0107] Backpropagation of network computation error information (output layer — input layer):

[0108] The actual output value of the output layer is compared with the expected value to calculate the error E; the global error calculation formula is as follows: Wherein, the error e of the k-th node in the output layer k =O k -y k O k It is the expected result of the k-th node in the output layer, y k It is the output result of the kth node of the output layer, and the expected value is the known FeO content of the sinter.

[0109] Weight adjustment amount:

[0110] Where η is the learning rate of the network, and n is the number of iterations of the network;

[0111] Threshold correction amount:

[0112] Adjusted weights: w jk (n+1)=w jk (n)+Δw jk (n)

[0113] Adjusted threshold: b k (n+1)=b k (n)+Δb k (n)

[0114] As one or more embodiments, in S103, a pre-trained SSA-BP neural network is used; the training steps include:

[0115] The first step is to determine the network topology and initialize the BP neural network, including: determining the weight range, error function, computational precision, and number of training iterations;

[0116] The second step is to initialize the sparrow population-related parameters and define the maximum number of iterations.

[0117] The third step is to input the training data into the BP neural network, calculate the population fitness, and determine the optimal position of the current population.

[0118] The fourth step involves obtaining the optimal weights and thresholds using the Sparrow Search Algorithm (SSA) and assigning them to the BP neural network to calculate the inputs and outputs of each neuron in the hidden and output layers.

[0119] The fifth step is to compare the actual output value of the output layer with the expected value and calculate the error E.

[0120] The sixth step is to determine whether the error E meets the requirements (an error within ±10% is sufficient). When the error E meets the requirements, there is no need to adjust the weights and thresholds. The optimization of the entire BP neural network model is then completed, and the optimal prediction model for FeO content in sintered ore is obtained.

[0121] When the error E does not meet the requirements, it is determined whether the number of learning iterations has reached the maximum number of learning iterations. When the number of learning iterations reaches the maximum number of learning iterations, the establishment of the sinter FeO content prediction model ends; when the number of learning iterations does not reach the maximum number of learning iterations, it returns to step four, iterates, and repeatedly adjusts the threshold of the weights to complete the optimization of the sinter FeO content prediction model.

[0122] A BP neural network can update its parameters once through one forward propagation and one backward propagation. The optimization of the sinter FeO content prediction model is to continuously perform forward and backward propagation to update the network parameters, so that the network can eventually approximate the true relationship.

[0123] This embodiment provides an online detection system for FeO content in sintered ore based on a BP neural network; a schematic diagram of the overall system is shown below. Figure 8 As shown;

[0124] The online detection system for FeO content in sintered ore based on a BP neural network is characterized by including:

[0125] Weighing and speed measuring devices are used to collect weight and speed data of sintered ore at specific locations on the conveyor belt in the industrial site; the acquired weight and speed data of sintered ore at specific locations on the conveyor belt in the industrial site are then transmitted to the monitoring platform.

[0126] The monitoring platform processes the weight and speed data of sinter at specific locations on the conveyor belt in the industrial site using its own pre-trained BP neural network; the pre-trained BP neural network outputs the results and determines the FeO content of the sinter and indicates whether the FeO content of the sinter meets the requirements.

[0127] The monitoring platform is also used to remind users whether the FeO content of the sinter meets the requirements (the FeO content of the sinter is 7%-12% which meets the requirements); if the FeO content of the sinter does not meet the requirements (the FeO content of the sinter is less than 7% or greater than 12%), the monitoring platform will issue an alarm.

[0128] Furthermore, the weighing device includes: a plurality of weighing sensors and a microcontroller. The plurality of weighing sensors are respectively connected to a junction box. The junction box is connected to a weighing force measuring instrument. The weighing force measuring instrument is connected to the microcontroller. The weight data measured by the weighing sensors is converted from an analog signal to a digital signal by the weighing force measuring instrument. The conversion result is collected by the microcontroller.

[0129] Furthermore, the speed measuring device includes: a measuring wheel, an analog-to-digital conversion module, and a microcontroller. The measuring wheel is installed on the return section of the conveyor belt and connected to a speed sensor. The speed sensor is connected to the microcontroller. The speed data measured by the speed sensor is converted from an analog signal to a digital signal through the digital-to-analog conversion module, and the conversion result is collected by the microcontroller.

[0130] Furthermore, the weighing device and the speed measuring device respectively collect weight data and speed data through a microcontroller, and then communicate with the monitoring platform through the microcontroller to upload the data to the monitoring platform.

[0131] The training steps for the SSA-BP neural network are exactly the same as those in Example 1, and will not be repeated here.

[0132] Figure 9 This is a comparison chart of the experimental results and the on-site test results in this embodiment. As can be seen from the chart, the detection method and the test results in this embodiment have similar trends. Figure 10 This is a graph showing the absolute error between the experimental results and the on-site test results in this embodiment. Figure 10 It can be seen that the detection results of the detection method in this embodiment meet the on-site process requirements.

[0133] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An online detection method for FeO content in sintered ore, characterized in that: Includes the following steps, S1. Divide the conveyor belt used for transporting sinter into a first weighing area and a second weighing area. Weighing devices for weighing the passing sinter are set on the first and second weighing areas. The second weighing area is equipped with a magnetic device that applies magnetic force to the magnetic material in the sinter. A speed measuring device for obtaining the transmission speed is set on the conveyor belt. S2. When the sinter is being transported on the conveyor belt, the weight of the sinter passing through the first weighing area and the second weighing area, as well as the conveyor belt speed, are obtained in real time. S3. Based on the sinter weight data and conveyor belt motion data obtained in S2, the FeO content in the sinter is detected using a pre-trained SSA-BP neural network. S4. Determine whether the FeO content in the sinter is within the preset range. If it exceeds the preset range, an alarm will be triggered. The SSA-BP neural network includes an input layer, an output layer, and a hidden layer; The training steps for the SSA-BP neural network are as follows: S301. Initialize the SSA-BP neural network by setting parameters including weight range, error function, computational precision, and maximum number of learning iterations. S302. Initialize sparrow population parameters and define the maximum number of iterations; S303. Input the training data into the SSA-BP neural network, collect and distribute the population fitness, and determine the optimal position of the current population; wherein, the training data are the weight values ​​of the first weighing area, the weight values ​​of the second weighing area, and the transmission speed of the conveyor belt for each batch of sintered ore with different quality and known FeO content. S304. Obtain the optimal weights and thresholds using the sparrow search algorithm and assign them to the SSA-BP neural network. Calculate the inputs and outputs of each neuron in the hidden and output layers. S305. The actual output value of the output layer is compared with the expected value, and the error E is calculated. S306. When the error E is within the preset range, it is determined that the error E meets the requirements, and the current SSA-BP neural network is used as the final prediction model; otherwise, it is determined whether the current number of learning iterations has reached the maximum number of learning iterations. If it has reached the maximum number of learning iterations, the current SSA-BP neural network is used as the final prediction model. If it has not reached the maximum number of learning iterations, it returns to S304 for iterative iteration.

2. The online detection method for FeO content in sintered ore according to claim 1, characterized in that: S304 specifically refers to, The output of the i-th node in the input layer is , ; in, These represent the weight values ​​of the first weighing area, the weight values ​​of the second weighing area, and the conveyor speed of the conveyor belt for each batch of sintered ore with different masses and known FeO content. The output of the j-th node in the hidden layer is ; in, It is the connection weight between the i-th node in the input layer and the j-th node in the hidden layer. It is the threshold at the j-th node of the hidden layer. This refers to the activation functions of the hidden layer nodes; the number of nodes in the hidden layer is mainly selected based on the following formula. , where j represents the number of nodes in the hidden layer, and m and n represent the number of nodes in the input layer and output layer, respectively.

3. The online detection method for FeO content in sintered ore according to claim 1, characterized in that: In S305, the method for calculating the error E is as follows: The error of the kth node in the output layer , It is the expected result of the k-th node in the output layer. It is the output result of the kth node of the output layer; the expected value is the known FeO content of the sinter.

4. The online detection method for FeO content in sintered ore according to claim 1, characterized in that: The training data is randomly divided into training set, validation set and test set in proportions of 70%, 15% and 15% respectively.

5. The online detection method for FeO content in sintered ore according to claim 3, characterized in that: The preset range corresponding to error E is ±10%.

6. A sinter FeO content online detection system for implementing the online detection method for sinter FeO content according to any one of claims 1-5, characterized in that: include, Conveyor belt; the conveyor belt is equipped with a first weighing area and a second weighing area; Weighing device; It includes weighing sensors installed in the first and second weighing areas of the conveyor belt to obtain the weight of the sinter when it is conveyed to the first and second weighing areas. Magnetic device; Set in the second weighing area, it is used to apply magnetic force to the magnetic materials in the sinter that have passed through the second weighing area; Speed ​​measuring device; Used to obtain the conveyor belt's conveyor speed; The monitoring platform is electrically connected to the weighing and speed measuring devices. It includes a pre-trained SSA-BP neural network, which is used to determine the FeO content in the sinter based on the real-time acquired sinter weight data and the conveyor belt speed, and to issue an alarm when the FeO content in the sinter exceeds a preset range.

7. The online detection system for FeO content in sintered ore according to claim 6, characterized in that: The magnetic device is a neodymium iron boron magnet.

8. The online detection system for FeO content in sintered ore according to claim 6, characterized in that: The speed measuring device includes a speed measuring wheel installed on the return section of the conveyor belt.

Images