Silo eccentric discharging monitoring and deviation rectifying method and system

Abstract

The invention relates to the technical field of mechanical measurement and sensing, in particular to a silo eccentric unloading monitoring and deviation rectifying method and system.The method comprises the steps that force sensors are installed below all supporting legs of a silo, original load signals are collected, noise is removed through band-pass filtering, and the real-time loads of the supporting legs are obtained; based on the supporting legs, the silo is divided into sectors, material level data of the sectors are obtained, the weight of the sector of the longitudinal gradient is constructed, and the weighted load is calculated; and constructing an eccentric vector based on the weighted load sequence, and mapping the weighted load into a two-dimensional eccentric vector component by using the angle coordinate of each supporting leg. According to the silo eccentric discharging monitoring and deviation rectifying method provided by the invention, the weighted load is mapped into the two-dimensional eccentric vector component based on the angle coordinates of the supporting legs, linear fusion is carried out by combining the two-dimensional eccentric vector component with the material level vector component, the combined eccentric coefficient is obtained, and the coefficient can quantify the eccentric degree and output the eccentric direction angle; and a clear physical basis is provided for subsequent correction decision.

Description

Technical Field

[0001] This invention relates to the field of mechanical measurement and sensing technology, specifically to a method and system for monitoring and correcting eccentric unloading of silos. Background Technology

[0002] Silos, as storage and unloading facilities for bulk materials, are widely used in industries such as grain, building materials, and chemicals. During actual unloading, due to factors such as material particle shape, moisture content, outlet layout, and operating methods, eccentric unloading often occurs, where material concentrates on one side, leading to uneven stress distribution within the silo. Eccentric unloading can cause excessive load on local support legs and foundations, resulting in structural tilting, crack propagation, or even overall instability, potentially leading to silo collapse in severe cases.

[0003] Existing technologies for silo eccentricity detection mostly rely on outrigger load detection, which makes it difficult to simultaneously obtain the dynamic relationship between the force on the silo outriggers and the internal material level distribution. This makes it impossible to accurately reflect the formation process of unloading eccentricity. Furthermore, existing technologies generally lack diagnostic analysis of the causes of eccentricity, and cannot implement targeted control based on different types of eccentricity such as local bridging, outlet blockage, or uneven flow, thus affecting the effectiveness of eccentricity correction. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method and system for monitoring and correcting eccentric unloading of silos.

[0005] This invention employs the following technical solution: a method for monitoring and correcting eccentric unloading in silos, comprising:

[0006] Force sensors are installed under each leg of the silo to collect raw load signals. Noise is removed by bandpass filtering to obtain the real-time load on the leg.

[0007] based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and vertical gradient;

[0008] The longitudinal gradient is constructed based on the average material level of the sector and the longitudinal gradient. The sector weight, based on the constructed... Calculate the weighted load based on the sector weights and the real-time load of the outriggers;

[0009] An eccentric vector is constructed based on the weighted load sequence, and the weighted load is mapped into two-dimensional eccentric vector components using the angular coordinates of each leg.

[0010] Based on sector material level data, material level vector components are constructed. The two-dimensional eccentricity vector components and material level vector components are linearly fused to obtain a joint eccentricity coefficient. The joint eccentricity coefficient is compared and analyzed with a preset eccentricity coefficient threshold to determine and generate a correction command.

[0011] As a further description of the above technical solution: the silo eccentric unloading monitoring and correction method also includes:

[0012] When a correction command is generated, the two-dimensional eccentricity vector components and the material level vector components are input into the pre-constructed eccentricity cause-effect diagnostic model for cause classification to obtain the cause type of silo eccentric unloading.

[0013] The orientation angle and eccentricity of the material's center of gravity offset are obtained based on two-dimensional eccentric vector components;

[0014] The obtained silo eccentricity unloading cause type, joint eccentricity coefficient, direction angle and eccentricity are input into the pre-built correction strategy prediction model, and the correction strategy is output.

[0015] As a further description of the above technical solution: the acquisition Methods for obtaining material level data for a single sector include:

[0016] K non-contact ranging sensors are installed at equal intervals along the height direction in the sector corresponding to each outrigger on the inner circumference of the cylinder. The time of each sensor is recorded. The instantaneous reading is expressed as ; For the first sector number The material level height measured by each sensor, among which, =1,2,…,K;

[0017] Based on the Calculate the average material level of a sector using K sensor sequences for each sector. With longitudinal gradient .

[0018] As a further description of the above technical solution: the method for obtaining the two-dimensional eccentric vector components is as follows:

[0019] Establish a coordinate system with the center point of the silo as the center, and obtain the installation angle of each support leg in radians;

[0020] The weighted load is projected onto a plane with the center of the cylinder as the origin, and the X and Y components are calculated respectively to obtain the two-dimensional eccentric vector components.

[0021] The expression for the two-dimensional eccentric vector component is:

[0022] ;

[0023] ;

[0024] In the formula, For the first The installation angle of each outrigger relative to the center point of the silo For the first The X-component of each support leg For the first The Y-component of each leg The number of outriggers. For the first Weighted load of each outrigger.

[0025] As a further description of the above technical solution: the method for constructing material level vector components based on sector material level data includes:

[0026] ;

[0027] ;

[0028] In the formula, For the first The material position vector X component of each sector. For the first The Y component of the material position vector of each sector The number of sectors For the first Average material level of each sector.

[0029] As a further description of the above technical solution: the method for determining and generating correction instructions includes:

[0030] When the joint eccentricity coefficient is greater than the eccentricity coefficient threshold, a correction instruction is generated; when the joint eccentricity coefficient is less than or equal to the eccentricity coefficient threshold, no correction instruction is generated.

[0031] As a further description of the above technical solution: the training method of the eccentric causal diagnostic model includes:

[0032] F sets of training data were collected in advance, where F is a positive integer greater than 0. The training data included two-dimensional eccentricity vector components and material level vector components, as well as the corresponding silo eccentricity unloading cause type.

[0033] A lightweight multilayer perceptron is used as an eccentric causal diagnostic model;

[0034] The acquired training data is divided into training set, validation set and test set according to a preset ratio. The training set is used for model parameter learning and network weight optimization through gradient descent. The validation set is used for hyperparameter tuning and early stopping judgment to avoid model overfitting. The test set is used to evaluate the model's generalization ability to ensure that the model can still make accurate diagnoses in unseen working conditions. Stratified sampling is used when dividing the data to ensure that the proportion of the four causes of the three datasets is consistent.

[0035] The training set is input into the model in batches. The probability distribution of the four causes is calculated through forward propagation. The gradient is calculated based on the cross-entropy loss function. The network weights are updated through backpropagation. After each round of training, the classification accuracy and average cross-entropy loss are calculated using the validation set. The current optimal model parameters are recorded.

[0036] Set a stopping condition. If the stopping condition is triggered, training will stop immediately. Load and save the model parameters with the highest classification accuracy on the validation set. Then test the model on the test set. When the average diagnostic accuracy of the four causes on the test set is greater than or equal to the preset threshold, the model can be deployed and used.

[0037] As a further description of the above technical solution: the correction strategy includes the number of discharge valves opened and the valve angle, as well as the number of pneumatic pulse nozzles opened and the pneumatic pulse nozzle angle.

[0038] As a further description of the above technical solution: the training method of the correction strategy prediction model includes:

[0039] The initialization of the deviation correction strategy prediction model structure adopts a multi-layer feedforward network structure of MLP type, with four input layers, two hidden layers, and two output layers. The input layers include a first input layer, a second input layer, a third input layer, and a fourth input layer, which are used to input the silo deviation discharge cause type label, joint deviation coefficient, orientation angle, and deviation degree, respectively. The hidden layers include a first hidden layer and a second hidden layer. The first hidden layer has 128 nodes and uses the ReLU function as the activation function. The second hidden layer has 64 nodes and uses the ReLU function as the activation function. The output layers include a first output layer and a second output layer. The first output layer has two branches, which are used to output the number of open discharge valves and the valve angle, respectively. The second output layer has two branches, which are used to output the number of open pneumatic pulse nozzles and the pneumatic pulse nozzle angle, respectively.

[0040] After initializing the error correction strategy prediction model structure, the error correction strategy prediction model is trained using the initial data and the error correction strategy corresponding to the initial data. The initial data and the error correction strategy corresponding to the initial data are obtained from the database, which records each silo error correction record for optimization of the model.

[0041] A silo eccentric discharge monitoring and correction system is used to implement the aforementioned silo eccentric discharge monitoring and correction method, including:

[0042] The load acquisition module installs force sensors under each leg of the silo to acquire raw load signals, and uses bandpass filtering to remove noise to obtain the real-time load of the legs.

[0043] Material level data acquisition module, based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and vertical gradient;

[0044] The weighted load calculation module constructs the longitudinal gradient based on the sector average material level and the longitudinal gradient. The sector weight, based on the constructed... Calculate the weighted load based on the sector weights and the real-time load of the outriggers;

[0045] The vector construction module constructs an eccentric vector based on the weighted load sequence, and uses the angular coordinates of each leg to map the weighted load into two-dimensional eccentric vector components.

[0046] The instruction generation module constructs material level vector components based on sector material level data, linearly fuses the two-dimensional eccentricity vector components and material level vector components to obtain a joint eccentricity coefficient, compares and analyzes the joint eccentricity coefficient with a preset eccentricity coefficient threshold, and determines and generates a correction instruction.

[0047] The eccentricity cause diagnosis module, when generating a correction command, inputs the two-dimensional eccentricity vector components and the material level vector components into the pre-constructed eccentricity cause diagnosis model for cause classification, and obtains the cause type of eccentric unloading in the silo.

[0048] The eccentricity parameter extraction module obtains the orientation angle and eccentricity of the material's center of gravity offset based on two-dimensional eccentricity vector components.

[0049] The deviation correction strategy generation module takes the acquired causes of silo eccentric unloading, joint eccentricity coefficient, orientation angle, and eccentricity degree as inputs into the pre-built deviation correction strategy prediction model, and outputs the deviation correction strategy.

[0050] The beneficial effects of this invention are as follows:

[0051] The present invention provides a method and system for monitoring and correcting eccentric unloading of silos. The system collects the load on the outriggers through force sensors under the outriggers, divides the silo into a corresponding number of sectors based on the number of outriggers, constructs a longitudinal gradient weight by combining the average material level and longitudinal gradient of each sector, and then calculates the weighted load on the outriggers. Through the vectorization processing of the weighted load sequence on the outriggers, the load distribution can be mapped into two-dimensional eccentricity vector components, and linearly fused with the material level vector components to obtain a joint eccentricity coefficient that reflects the uneven distribution of materials. This coefficient can quantitatively reflect the direction and degree of eccentricity, and realizes the accurate identification and quantitative analysis of eccentricity.

[0052] Furthermore, after detecting an eccentric state, by inputting the eccentric vector and material level vector components into a pre-constructed eccentricity causal diagnostic model, the eccentricity cause type can be automatically identified, providing a basis for the selection of subsequent correction strategies. The direction angle and eccentricity of the material's center of gravity offset can be directly obtained based on the two-dimensional eccentricity vector components. Then, the eccentricity cause type, joint eccentricity coefficient, direction angle, and eccentricity are used as inputs. After training, the optimal correction strategy can be automatically output, including the number and angle of valve opening and the number and angle of pneumatic pulse nozzle opening, to achieve targeted correction and dynamic adjustment. Attached Figure Description

[0053] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0054] Figure 1 This is a flowchart of the silo eccentric unloading monitoring and correction method provided in Embodiment 1 of the present invention;

[0055] Figure 2 This is a flowchart of the silo eccentric unloading monitoring and correction method provided in Embodiment 2 of the present invention;

[0056] Figure 3 This is a module connection diagram of the silo eccentric unloading monitoring and correction system provided in Embodiment 3 of the present invention;

[0057] Figure 4 A schematic diagram of silo sector division provided for this invention. Detailed Implementation

[0058] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0059] Example 1

[0060] Please see Figure 1 and Figure 4 This invention provides a technical solution: a method for monitoring and correcting eccentric unloading in silos, comprising:

[0061] Force sensors are installed under each leg of the silo to collect raw load signals. Noise is removed by bandpass filtering to obtain the real-time load on the leg. ,in, For the first Each support leg is in constant motion The original readings, =1,2,…, .

[0062] based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and longitudinal gradient; where n is the number of legs, which is also equal to the number of sectors.

[0063] It should be noted that the silos are divided into The specific method for dividing the silo into sectors is to divide the silo into a corresponding number of sectors based on the number of legs, and to ensure that the legs are located on the center line of the corresponding sectors.

[0064] Get Methods for obtaining material level data for a single sector include:

[0065] K non-contact ranging sensors are installed at equal intervals along the height direction in the sector corresponding to each outrigger on the inner circumference of the cylinder. The time of each sensor is recorded. The instantaneous reading is expressed as ; For the first sector number The material level height measured by each sensor, among which =1,2,…, ;

[0066] The average material level of each sector is calculated based on K sensor sequences for each sector. With longitudinal gradient ;

[0067] The formula for calculating the average material level is:

[0068] In the formula, This represents the average material level in the sector.

[0069] The formula for calculating the longitudinal gradient is:

[0070] ; For the vertical gradient, The height of the top layer of the sector. This represents the height of the bottom layer of the sector.

[0071] Based on average material level in sector With longitudinal gradient Constructing the longitudinal gradient Sector weight Based on the constructed first Weighted load is calculated based on sector weights and real-time outrigger loads. ;

[0072] The formula for calculating the vertical gradient weights is as follows: In the formula, For the vertical gradient sector weights, This is the sensitivity coefficient. To prevent zero-value items;

[0073] The construction-based first The method for calculating the weighted load based on sector weights and real-time leg loads is as follows:

[0074] In the formula, For the first Weighted load of each outrigger for Real-time load of each outrigger For the first Sector weights;

[0075] It should be noted that the physical meaning of constructing sector weights by combining longitudinal features is to eliminate the load deviation caused by material level fluctuations, and to ensure that the eccentric vector reflects the true offset of the center of gravity, rather than the false eccentricity caused by the fluctuation of the material surface. That is, when only the support leg load is used, the system can only identify that the load is large at a certain point, but cannot determine which direction the material body is offset. However, after introducing the material level features, a clear two-dimensional eccentric vector can be obtained through spatial weighted mapping.

[0076] An eccentric vector is constructed based on the weighted load sequence, and the weighted load is mapped into two-dimensional eccentric vector components using the angular coordinates of each leg.

[0077] The method for obtaining the two-dimensional eccentric vector components is as follows:

[0078] Establish a coordinate system with the center point of the silo as the center, and obtain the installation angle of each support leg in radians;

[0079] The weighted load is projected onto a plane with the center of the cylinder as the origin, and the X and Y components are calculated to obtain the two-dimensional eccentric vector components.

[0080] The expression for the two-dimensional eccentric vector component is:

[0081] ;

[0082] ;

[0083] In the formula, For the first The installation angle of each outrigger relative to the center point of the silo, in radians. For the first The X-component of each support leg For the first The Y-component of each leg The number of outriggers. For the first Weighted load of each outrigger.

[0084] Based on sector material level data, material level vector components are constructed. The two-dimensional eccentric vector and material level vector components are linearly fused to obtain a joint eccentricity coefficient. The joint eccentricity coefficient is compared and analyzed with a preset eccentricity coefficient threshold to determine and generate a correction command.

[0085] Methods for constructing material level vector components based on sector material level data include:

[0086] ;

[0087] ;

[0088] In the formula, For the first The material position vector X component of each sector. For the first The Y component of the material position vector of each sector The number of sectors For the first Average material level of each sector.

[0089] The formula for calculating the joint eccentricity coefficient is as follows:

[0090] ;

[0091] In the formula, For the joint eccentricity coefficient, This is the dimensionless load eccentricity amplitude, representing the relative magnitude of the force eccentricity. To normalize the level eccentricity amplitude, where, The sum of weighted loads on all outriggers; This is the sum of the average material levels across all sectors; These are the weighting coefficients.

[0092] It should be noted that, Let be the magnitude of the vector, representing the degree of eccentricity, and sum it with the total load. The normalized value can be used to determine the degree of imbalance in the current force distribution. If the value is close to 0, it indicates equilibrium; if it is close to 1, it indicates severe eccentricity.

[0093] The combined eccentricity coefficient is compared and analyzed with the preset eccentricity coefficient threshold to determine and generate a correction command.

[0094] In this embodiment, the load on the outriggers is collected by force sensors under the outriggers. Based on the number of outriggers in the silo, the silo is divided into a corresponding number of sectors. By combining the average material level and longitudinal gradient of each sector, a longitudinal gradient weight is constructed, and then the weighted load on the outriggers is calculated. Through the vectorization processing of the weighted load sequence on the outriggers, the load distribution can be mapped into two-dimensional eccentricity vector components, and linearly fused with the material level vector components to obtain a joint eccentricity coefficient that reflects the uneven distribution of materials. This coefficient can quantitatively reflect the direction and degree of eccentricity, realizing the accurate identification and quantitative analysis of the eccentricity state.

[0095] Example 2

[0096] Please see Figure 2 and Figure 4 Based on Embodiment 1, this embodiment of the invention adds a method flow for obtaining the correction strategy when generating correction instructions:

[0097] The methods for determining the generation of correction instructions include:

[0098] When the joint eccentricity coefficient is greater than the eccentricity coefficient threshold, a correction instruction is generated; when the joint eccentricity coefficient is less than or equal to the eccentricity coefficient threshold, no correction instruction is generated.

[0099] When a correction command is generated, the two-dimensional eccentricity vector components and the material level vector components are input into the pre-constructed eccentricity cause-effect diagnostic model for cause classification to obtain the cause type of silo eccentric unloading.

[0100] It should be noted that the types of causes of eccentric unloading in silos include local bridging eccentricity, upper layer slippage eccentricity, material moisture and particle characteristics eccentricity, and operational error eccentricity.

[0101] Training methods for eccentric causal diagnostic models include:

[0102] F sets of training data were collected in advance, where F is a positive integer greater than 0. The training data included two-dimensional eccentricity vector components and material level vector components, as well as the corresponding silo eccentricity unloading cause type.

[0103] A lightweight multilayer perceptron is used as the eccentric causal diagnostic model, including:

[0104] The input layer has 5 nodes and corresponds to a 5-dimensional feature vector.

[0105] Hidden layers: Two fully connected hidden layers are set up, with 16 neurons in the first layer and 8 neurons in the second layer, both using the ReLU activation function;

[0106] Output layer: 4 nodes, corresponding to four types of eccentricity causes, using the Softmax activation function (outputting the probability distribution of each type of cause, which facilitates the judgment and diagnosis of confidence).

[0107] Regularization design: A Dropout layer is added after the first hidden layer to suppress overfitting; the weights are initialized using a He normal distribution to accelerate model convergence.

[0108] The acquired training data is divided into training set, validation set and test set according to a preset ratio; optionally, the preset ratio is 7:2:1.

[0109] The training set is used for learning model parameters and optimizing network weights through gradient descent. The validation set is used for hyperparameter tuning and early stopping judgment to avoid model overfitting. The test set is used to evaluate the model's generalization ability and ensure that the model can still make accurate diagnoses in unseen working conditions. Stratified sampling is used when dividing the datasets to ensure that the proportion of the four causes of the three datasets is consistent.

[0110] The training set is input into the model in batches. The probability distribution of the four causes is calculated through forward propagation. The gradient is calculated based on the cross-entropy loss function. The network weights are updated through backpropagation. After each round of training, the classification accuracy and average cross-entropy loss are calculated using the validation set. The current optimal model parameters are recorded.

[0111] Set the stopping condition as "the improvement in the classification accuracy of the validation set is less than 0.5% for 15 consecutive rounds". If this stopping condition is triggered, stop training immediately, load and save the model parameters with the highest classification accuracy on the validation set, and then use the test set for testing. When the average diagnostic accuracy of the four causes on the test set is greater than or equal to the preset threshold, it can be deployed and used. If the evaluation does not meet the standard, repeat the data collection and training process until all indicators meet the requirements.

[0112] The orientation angle and eccentricity of the material's center of gravity offset are obtained based on two-dimensional eccentric vector components;

[0113] The formula for calculating the direction angle is:

[0114] ;in, It is the direction angle;

[0115] It should be noted that obtaining the direction angle of the material's center of gravity offset, i.e., the angular range in which the material is concentrated, provides a spatial coordinate reference for subsequent correction control, such as the discharge valve and the position of the pneumatic pulse.

[0116] The formula for calculating the eccentricity is: ; Eccentricity.

[0117] It should be noted that, The larger the value, the greater the degree of eccentricity.

[0118] The obtained silo eccentricity unloading cause type, joint eccentricity coefficient, direction angle and eccentricity are input into the pre-built correction strategy prediction model, and the correction strategy is output.

[0119] The correction strategy includes the number of discharge valves opened and the valve angle, as well as the number of pneumatic pulse nozzles opened and the pneumatic pulse nozzle angle.

[0120] The training method for the prediction model of the correction strategy includes:

[0121] The initialization of the deviation correction strategy prediction model structure adopts a multi-layer feedforward network structure of MLP type, with four input layers, two hidden layers, and two output layers. The input layers include a first input layer, a second input layer, a third input layer, and a fourth input layer, which are used to input the silo deviation discharge cause type label, joint deviation coefficient, orientation angle, and deviation degree, respectively. The hidden layers include a first hidden layer and a second hidden layer. The first hidden layer has 128 nodes and uses the ReLU function as the activation function. The second hidden layer has 64 nodes and uses the ReLU function as the activation function. The output layers include a first output layer and a second output layer. The first output layer has two branches, which are used to output the number of open discharge valves and the valve angle, respectively. The second output layer has two branches, which are used to output the number of open pneumatic pulse nozzles and the pneumatic pulse nozzle angle, respectively.

[0122] After initializing the error correction strategy prediction model structure, the error correction strategy prediction model is trained using the initial data and the error correction strategy corresponding to the initial data. The initial data and the error correction strategy corresponding to the initial data are obtained from the database. The database records each silo error correction record and is used to optimize the model. The optimizer is Adam, the loss function is the MSE loss function, the batch size is set to 32, and the number of iterations is 200 rounds. The training ends when the loss function converges, indicating that the training is complete.

[0123] It should be noted that for the input data, the silo eccentric discharge cause type label, joint eccentricity coefficient, orientation angle and eccentricity, the silo eccentric discharge cause type label needs to be one-hot encoded or embedded encoded, because the silo eccentric discharge cause type label is categorical data and cannot be directly input into the MLP.

[0124] In this embodiment, the established correction strategy prediction model adopts a multi-layer feedforward neural network structure. It takes the type of eccentricity, joint eccentricity coefficient, direction angle and eccentricity as input. After training, it can automatically output the optimal correction strategy, including the number and angle of valve opening and the number and angle of pneumatic pulse nozzle opening, so as to achieve targeted correction and dynamic adjustment, avoiding the lag and error caused by manual judgment.

[0125] Through the above-mentioned technical means, the present invention realizes the closed-loop linkage of eccentricity detection, cause diagnosis and correction control. It can trigger the correction command at the initial stage of eccentricity, automatically adjust the unloading direction and airflow distribution, and make the force on the silo tend to be balanced. This method not only improves the response speed and accuracy of eccentricity detection and correction, but also significantly reduces the load difference between the outriggers, extends the service life of the silo structure, and ensures the safety and stability of the unloading process.

[0126] Example 3

[0127] Please see Figure 3 and Figure 4 This invention provides a technical solution: a silo eccentric unloading monitoring and correction system, used to implement the aforementioned silo eccentric unloading monitoring and correction method, comprising:

[0128] The load acquisition module installs force sensors under each leg of the silo to acquire raw load signals, and uses bandpass filtering to remove noise to obtain the real-time load of the legs.

[0129] Material level data acquisition module, based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and vertical gradient;

[0130] The weighted load calculation module constructs the longitudinal gradient based on the sector average material level and the longitudinal gradient. The sector weight, based on the constructed... Calculate the weighted load based on the sector weights and the real-time load of the outriggers;

[0131] The vector construction module constructs an eccentric vector based on the weighted load sequence, and uses the angular coordinates of each leg to map the weighted load into two-dimensional eccentric vector components.

[0132] The instruction generation module constructs material level vector components based on sector material level data, linearly fuses the two-dimensional eccentricity vector components and material level vector components to obtain a joint eccentricity coefficient, compares and analyzes the joint eccentricity coefficient with a preset eccentricity coefficient threshold, and determines and generates a correction instruction.

[0133] The eccentricity cause diagnosis module, when generating a correction command, inputs the two-dimensional eccentricity vector components and the material level vector components into the pre-constructed eccentricity cause diagnosis model for cause classification, and obtains the cause type of eccentric unloading in the silo.

[0134] The eccentricity parameter extraction module obtains the orientation angle and eccentricity of the material's center of gravity offset based on two-dimensional eccentricity vector components.

[0135] The correction strategy generation module inputs the acquired silo eccentricity unloading cause type, joint eccentricity coefficient, direction angle and eccentricity into the pre-built correction strategy prediction model and outputs the correction strategy.

[0136] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and correcting eccentric unloading in silos, characterized in that, include: Force sensors are installed under each leg of the silo to collect raw load signals. Noise is removed by bandpass filtering to obtain the real-time load on the leg. based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and vertical gradient; The longitudinal gradient is constructed based on the average material level of the sector and the longitudinal gradient. The sector weight, based on the constructed... Calculate the weighted load based on the sector weights and the real-time load of the outriggers; An eccentric vector is constructed based on the weighted load sequence, and the weighted load is mapped into two-dimensional eccentric vector components using the angular coordinates of each leg. Based on sector material level data, material level vector components are constructed. The two-dimensional eccentricity vector components and material level vector components are linearly fused to obtain a joint eccentricity coefficient. The joint eccentricity coefficient is compared and analyzed with a preset eccentricity coefficient threshold to determine and generate a correction command.

2. The method for monitoring and correcting eccentric unloading of silos according to claim 1, characterized in that, Also includes: When a correction command is generated, the two-dimensional eccentricity vector components and the material level vector components are input into the pre-constructed eccentricity cause-effect diagnostic model for cause classification to obtain the cause type of silo eccentric unloading. The orientation angle and eccentricity of the material's center of gravity offset are obtained based on two-dimensional eccentric vector components; The obtained silo eccentricity unloading cause type, joint eccentricity coefficient, direction angle and eccentricity are input into the pre-built correction strategy prediction model, and the correction strategy is output.

3. The method for monitoring and correcting eccentric unloading of silos according to claim 1, characterized in that, The acquisition Methods for obtaining material level data for a single sector include: K non-contact ranging sensors are installed at equal intervals along the height direction in the sector corresponding to each outrigger on the inner circumference of the cylinder. The time of each sensor is recorded. The instantaneous reading is expressed as ; For the first sector number The material level height measured by each sensor, among which, =1,2,…,K; Based on the Calculate the average material level of a sector using K sensor sequences for each sector. With longitudinal gradient .

4. The method for monitoring and correcting eccentric unloading of silos according to claim 1, characterized in that, The method for obtaining the two-dimensional eccentric vector components is as follows: Establish a coordinate system with the center point of the silo as the center, and obtain the installation angle of each support leg in radians; The weighted load is projected onto a plane with the center of the cylinder as the origin, and the X and Y components are calculated respectively to obtain the two-dimensional eccentric vector components. The expression for the two-dimensional eccentric vector component is: ； ； In the formula, For the first The installation angle of each outrigger relative to the center point of the silo For the first The X-component of each support leg For the first The Y-component of each leg The number of outriggers. For the first Weighted load of each outrigger.

5. The method for monitoring and correcting eccentric unloading of silos according to claim 1, characterized in that, The method for constructing material level vector components based on sector material level data includes: ； ； In the formula, For the first The material position vector X component of each sector. For the first The Y component of the material position vector of each sector The number of sectors For the first Average material level of each sector.

6. The method for monitoring and correcting eccentric unloading of silos according to claim 1, characterized in that, The method for determining and generating correction instructions includes: When the joint eccentricity coefficient is greater than the eccentricity coefficient threshold, a correction instruction is generated; when the joint eccentricity coefficient is less than or equal to the eccentricity coefficient threshold, no correction instruction is generated.

7. The method for monitoring and correcting eccentric unloading of silos according to claim 2, characterized in that, The training method for the eccentric causal diagnostic model includes: F sets of training data were collected in advance, where F is a positive integer greater than 0. The training data included two-dimensional eccentricity vector components and material level vector components, as well as the corresponding silo eccentricity unloading cause type. A lightweight multilayer perceptron is used as an eccentric causal diagnostic model; The acquired training data is divided into training set, validation set and test set according to a preset ratio. The training set is used for model parameter learning and network weight optimization through gradient descent. The validation set is used for hyperparameter tuning and early stopping judgment to avoid model overfitting. The test set is used to evaluate the model's generalization ability. Stratified sampling is used when dividing the data to ensure that the proportion of the four causes of the three datasets is consistent. The training set is input into the model in batches. The probability distribution of the four causes is calculated through forward propagation. The gradient is calculated based on the cross-entropy loss function. The network weights are updated through backpropagation. After each round of training, the classification accuracy and average cross-entropy loss are calculated using the validation set. The current optimal model parameters are recorded. Set a stopping condition. If the stopping condition is triggered, training will stop immediately. Load and save the model parameters with the highest classification accuracy on the validation set. Then test the model on the test set. When the average diagnostic accuracy of the four causes on the test set is greater than or equal to the preset threshold, the model can be deployed and used.

8. The method for monitoring and correcting eccentric unloading of silos according to claim 2, characterized in that, The correction strategy includes the number and angle of the discharge valves, as well as the number and angle of the pneumatic pulse nozzles.

9. The method for monitoring and correcting eccentric unloading of silos according to claim 8, characterized in that, The training method for the prediction model of the correction strategy includes: The initialization of the deviation correction strategy prediction model structure adopts a multi-layer feedforward network structure of MLP type, with four input layers, two hidden layers, and two output layers. The input layers include a first input layer, a second input layer, a third input layer, and a fourth input layer, which are used to input the silo deviation discharge cause type label, joint deviation coefficient, orientation angle, and deviation degree, respectively. The hidden layers include a first hidden layer and a second hidden layer. The first hidden layer has 128 nodes and uses the ReLU function as the activation function. The second hidden layer has 64 nodes and uses the ReLU function as the activation function. The output layers include a first output layer and a second output layer. The first output layer has two branches, which are used to output the number of open discharge valves and the valve angle, respectively. The second output layer has two branches, which are used to output the number of open pneumatic pulse nozzles and the pneumatic pulse nozzle angle, respectively. After initializing the error correction strategy prediction model structure, the error correction strategy prediction model is trained using the initial data and the error correction strategy corresponding to the initial data. The initial data and the error correction strategy corresponding to the initial data are obtained from the database, which records each silo error correction record for optimization of the model.

10. A silo eccentric discharge monitoring and correction system, used to implement the silo eccentric discharge monitoring and correction method according to any one of claims 1-9, characterized in that, include: The load acquisition module installs force sensors under each leg of the silo to acquire raw load signals, and uses bandpass filtering to remove noise to obtain the real-time load of the legs. Material level data acquisition module, based on Each support leg divides the silo into Each sector, obtain The sector material level data includes average material level and vertical gradient; The weighted load calculation module constructs the longitudinal gradient based on the sector average material level and the longitudinal gradient. The sector weight, based on the constructed... Calculate the weighted load based on the sector weights and the real-time load of the outriggers; The vector construction module constructs an eccentric vector based on the weighted load sequence, and uses the angular coordinates of each leg to map the weighted load into two-dimensional eccentric vector components. The instruction generation module constructs material level vector components based on sector material level data, linearly fuses the two-dimensional eccentricity vector components and material level vector components to obtain a joint eccentricity coefficient, compares and analyzes the joint eccentricity coefficient with a preset eccentricity coefficient threshold, and determines and generates a correction instruction. The eccentricity cause diagnosis module, when generating a correction command, inputs the two-dimensional eccentricity vector components and the material level vector components into the pre-constructed eccentricity cause diagnosis model for cause classification, and obtains the cause type of eccentric unloading in the silo. The eccentricity parameter extraction module obtains the orientation angle and eccentricity of the material's center of gravity offset based on two-dimensional eccentricity vector components. The correction strategy generation module inputs the acquired silo eccentricity unloading cause type, joint eccentricity coefficient, direction angle and eccentricity into the pre-built correction strategy prediction model and outputs the correction strategy.

Images