Automatic high-precision quantitative loading method for port bulk materials

Abstract

The invention discloses an automatic high-precision quantitative loading method for port bulk cargo, and relates to the technical field of transportation control, and the method comprises the steps: collecting bulk cargo data and vehicle data through a sensor, carrying out the preprocessing, and extracting features; an LSTM model is constructed and trained, the bulk cargo loading weight is predicted based on the extracted features, and the bulk cargo loading flow rate is optimized; starting bulk loading according to the predicted bulk loading weight and the predicted bulk loading flow rate, and after loading is completed, identifying the prediction precision to carry out LSTM model optimization; and generating a prediction record according to the bulk cargo loading weight, the loading speed and the prediction precision predicted each time, and storing the prediction record in a database. According to the method, the LSTM model is constructed to predict and control the loading weight of the bulk material, the loading capacity of the vehicle is accurately controlled, the working efficiency is improved, the loading speed of the bulk material is optimized by collecting the data of the bulk material, the loading precision of the bulk material is effectively improved, repeated unloading or loading caused by too much or too little loading of the material of the vehicle is avoided, and the working efficiency is improved. The vehicle operation period is shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of transportation control, and particularly to a method for automatically and highly accurately quantitatively loading bulk materials at a port. Background Art

[0002] In the bulk material industry, such as mines, coal, cement, steel mills, docks, etc., the loading operation is busy and complex. The traditional manual loading method not only takes time and effort, has low efficiency, but also has high danger and overspill risk. With the progress of technology, the automated loading technology has gradually matured. Especially the application of the intelligent automatic quantitative bulk material loading system has greatly reduced the manual intervention, saved the labor cost, and effectively avoided the overspill problem. To achieve intelligent automatic quantitative bulk material loading, it is necessary to accurately estimate the real-time value of the weight of the bulk material loaded onto the vehicle. Usually, equipment such as lidar is used on-site for three-dimensional reconstruction or a weighbridge is installed on-site for real-time weighing. However, the existing technology can only control the bulk material loading amount of the vehicle through equipment such as a weighbridge, cannot accurately control the vehicle loading amount, and often relies on a fixed bulk material loading speed, lacking dynamic optimization of the bulk material speed. In the actual bulk material loading environment, the accuracy and adaptability are relatively low, and it cannot meet the requirements of efficient port operation. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides a method for automatically and highly accurately quantitatively loading bulk materials at a port, which solves the problems of inability to accurately control the vehicle loading amount, often relying on a fixed bulk material loading speed, lacking dynamic optimization of the bulk material speed, having relatively low accuracy and adaptability in the actual bulk material loading environment, and being unable to meet the requirements of efficient port operation.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] In the first aspect, the present invention provides a method for automatically and highly accurately quantitatively loading bulk materials at a port, which includes

[0007] collecting bulk material data and vehicle data through sensors for preprocessing and extracting features;

[0008] constructing and training an LSTM model, predicting the bulk material loading weight based on the extracted features, and optimizing the bulk material loading flow rate;

[0009] starting the bulk material loading according to the predicted bulk material loading weight and the bulk material flow rate, and identifying the prediction accuracy after the loading is completed to optimize the LSTM model;

[0010] generating a prediction record for the bulk material loading weight, loading speed, and prediction accuracy predicted each time, and storing the prediction record in a database.

[0011] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: collecting bulk material data and vehicle data through sensors for preprocessing and extracting features means installing a high-frequency millimeter-wave radar level gauge and a laser flow velocity sensor at the bulk material discharging opening to obtain the bulk material height and bulk material flow velocity during the loading process, and collecting bulk material data and vehicle data including bulk material types, vehicle types, and vehicle load. Defining a collection time period, defining the bulk material height within the collection time period as a dynamic feature, the bulk material type as a category feature, and the vehicle type and vehicle load as static features.

[0012] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: constructing and training an LSTM model to predict the bulk material loading weight based on the extracted features includes:

[0013] Constructing an initial LSTM model, including an input layer, an LSTM layer, a fully connected layer, a feature fusion layer, and an output layer, and defining the inputs of the input layer as dynamic features, category features, and static features;

[0014] Training the initial LSTM model based on training data, and using the mean square error as the loss function, and adopting an Adam optimizer to iteratively optimize the parameters of the initial LSTM model to obtain a trained LSTM model;

[0015] Inputting the collected dynamic features, category features, and static features into the trained LSTM model to obtain the predicted bulk material loading weight;

[0016] The LSTM layer is used to process dynamic features, and the fully connected layer includes two layers, which are respectively used to process category features and static features.

[0017] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: optimizing the bulk material loading flow velocity includes:

[0018] Calculating the maximum bulk material loading flow velocity :

[0019]

[0020] Where is the highest safe pressure of the bulk material loading pipeline, is the bulk material density;

[0021] Defining the bulk material flow velocity as , and calculating the bulk material flow rate Q according to the cross-sectional area of the bulk material loading pipeline:

[0022]

[0023] Where is the cross-sectional area of the bulk material loading pipeline;

[0024] and calculate the pressure drop :

[0025]

[0026] where is the friction coefficient, is the length of the bulk material loading pipeline, is the diameter of the bulk material loading pipeline;

[0027] Calculate the hydraulic energy consumption according to the bulk material flow rate and the pressure drop :

[0028]

[0029] Calculate the velocity gradient of the bulk material in the loading pipeline according to the bulk material flow velocity, and define the dissipation function :

[0030]

[0031] where is the velocity gradient, is the dynamic viscosity of the bulk material, and y is the height of the bulk material in the loading pipeline;

[0032] Integrate the dissipation function to calculate the dissipation energy consumption :

[0033]

[0034] By summing the hydraulic energy consumption and the dissipation energy consumption to obtain the total energy consumption P during the bulk material loading process, and construct the momentum equation according to the total energy consumption P and the bulk material flow velocity:

[0035]

[0036] where is the Reynolds number, is the velocity gradient;

[0037] Initialize the bulk material flow velocity, set the boundary conditions, discretize the momentum equation using the finite difference method, and use the alternating direction implicit method to iteratively update until the bulk material flow velocity converges and then stop the iteration, and output the bulk material flow velocity obtained from the last iteration ;

[0038] Perform bulk material loading according to the calculated bulk material flow velocity ​

[0039] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: the step of starting the bulk material loading according to the predicted bulk material loading weight and the bulk material flow rate means starting the bulk material loading according to the predicted bulk material loading weight and the bulk material flow rate, and in the process of bulk material loading, collecting the bulk material height and the vehicle load data in real time, and stopping the bulk material loading when the vehicle load reaches the predicted bulk material loading weight.

[0040] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: the step of identifying the prediction accuracy and optimizing the LSTM model after the loading is completed means setting the expected vehicle load according to the vehicle type, and collecting the vehicle load and the expected load for comparison and calculating the error after the loading is completed. If the error exceeds the error threshold, the LSTM model is retrained and iteratively optimized.

[0041] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: the step of generating a prediction record for each predicted bulk material loading weight, loading speed and prediction accuracy means forming a prediction record for each predicted bulk material loading weight, bulk material flow rate and prediction accuracy according to the time stamp, and judging whether to optimize the LSTM model according to the prediction accuracy. If optimization is required, the prediction record is added to the training group, and if no optimization is required, the prediction record is added to the verification group. The prediction records of the training group and the verification group are used to train and optimize the LSTM model.

[0042] As a preferred embodiment of the method for automatically and highly accurately quantitatively loading bulk materials at ports according to the present invention, wherein: the step of storing the prediction record in the database means storing the prediction record in the database after generating the prediction record. The database regularly verifies the integrity and security of the prediction record, and synchronously uploads the prediction record to the cloud for backup.

[0043] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and: when the computer program is executed by the processor, any step of the method for automatically and highly accurately quantitatively loading bulk materials at ports as described in the first aspect of the present invention is implemented.

[0044] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the method for automatically and highly accurately quantitatively loading bulk materials at ports as described in the first aspect of the present invention is implemented.

[0045] The beneficial effects of the present invention are as follows: By constructing an LSTM model to predict and control the weight of bulk materials during loading, the present invention accurately controls the vehicle loading amount, improves work efficiency, and optimizes the bulk material loading speed by collecting bulk material data, effectively improving the accuracy of bulk material loading, avoiding repeated unloading or loading caused by excessive or insufficient material loading on the vehicle, and shortening the vehicle operation cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0047] Figure 1 It is a flowchart of the method for automatically and highly accurately quantitatively loading bulk materials at a port in Embodiment 1.

[0048] Figure 2 It is a schematic diagram of bulk material loading at a port in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0050] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0051] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that mutually excludes other embodiments.

[0052] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a method for automatically and highly accurately quantitatively loading bulk materials at a port, including the following steps:

[0053] S1. Collect bulk material data and vehicle data through sensors for preprocessing and extract features;

[0054] Specifically, collecting bulk material data and vehicle data through sensors, preprocessing them, and extracting features means installing a high-frequency millimeter-wave radar level gauge and a laser flow velocity sensor at the bulk material discharge opening to obtain the height and flow velocity of the bulk material during the loading process, and collecting bulk material data and vehicle data including the type of bulk material, vehicle type, and vehicle load. Define the acquisition time period, define the height of the bulk material within the acquisition time period as a dynamic feature, the type of bulk material as a categorical feature, and the vehicle type and vehicle load as static features.

[0055] Combining the millimeter-wave radar level gauge and the laser flow velocity sensor to collect the height and flow velocity of the bulk material respectively, constructing a complete dynamic feature of the bulk material supply process. By combining the height and flow velocity data, the dynamic feature description is more comprehensive. When the height of the bulk material drops rapidly while the flow velocity remains stable, it may indicate that the bulk material supply is restricted, and the model can use this information for precise adjustment. Through the reasonable classification of features, separating the dynamic features with strong time dependence, the categorical features representing the material characteristics, and the static features of the loading restriction conditions, providing a clear feature input structure for subsequent modeling. By setting the acquisition time period, clarifying the start and end ranges of the time series data, ensuring the time continuity and data quality of the dynamic features, avoiding the interference of redundant data on model training, and improving the efficiency of data processing.

[0056] S2. Construct an LSTM model and train it to predict the loading weight of bulk materials based on the extracted features and optimize the loading flow velocity of bulk materials;

[0057] Specifically, constructing an LSTM model and training it to predict the loading weight of bulk materials based on the extracted features includes:

[0058] Construct an initial LSTM model, including an input layer, an LSTM layer, a fully connected layer, a feature fusion layer, and an output layer. Define the inputs of the input layer as dynamic features, categorical features, and static features;

[0059] The input layer includes a dynamic feature input (time series data) with an input dimension of (batch_size, sequence_length, input_size_seq), where input_size_seq is the number of features at each time point (in this example, it is 3, representing the heights of three piles); a categorical feature input with an input dimension of (batch_size, num_material_types), which is the one-hot encoded categorical feature vector; a static feature input with an input dimension of (batch_size, num_static_features), representing the static features of each sample;

[0060] The input to the LSTM layer is time series data (batch_size, sequence_length, input_size_seq). This layer has num_layers LSTM cells, each cell has hidden_size hidden units, and the output is the final hidden state of the LSTM layer, with the output dimension being (batch_size, hidden_size).

[0061] The fully connected layer (for categorical feature processing) is fully connected layer 1. Its input is the one-hot encoded categorical features (batch_size, num_material_types). This layer contains a fully connected layer and a ReLU activation function, which converts the categorical features into a feature vector with the same dimension as the LSTM output. The output is the processed categorical feature vector, with the dimension (batch_size, hidden_size). By processing the one-hot encoded categorical features through the fully connected layer, the adaptability of the model to different materials can be significantly improved.

[0062] The fully connected layer (for static feature processing) is fully connected layer 2. The input is the static features (batch_size, num_static_features). This layer contains a fully connected layer and a ReLU activation function, which converts the static features into a feature vector with the same dimension as the LSTM output. The output is the processed static feature vector, with the dimension (batch_size, hidden_size). The vehicle type has a direct limit on the loading weight. The processed static features (such as the vehicle's load capacity) effectively enhance the constraint ability of the model and prevent the prediction of the loading weight from deviating from the vehicle's carrying range.

[0063] The feature fusion layer is a concatenation layer. The input is the dynamic features (batch_size, hidden_size) output by the LSTM layer, the processed categorical features (batch_size, hidden_size), and the static features (batch_size, hidden_size). The output is to concatenate the above features in the feature dimension to form a vector containing all feature information, with the dimension (batch_size, hidden_size * 3). Feature fusion enhances the model's global understanding of the loading scenario. For example, for a specific combination of material types (such as wet coal) and vehicle types (such as short carriages), through the fused feature vector, the model can accurately predict the impact of the change in the bulk material flow rate on the loading weight.

[0064] The output layer consists of two fully connected layers. The input to the fully connected layer 3 is the fused feature vector (batch_size, hidden_size * 3). This layer contains a fully connected layer that reduces the fused feature vector to 64 dimensions and applies the ReLU activation function. The fully connected layer 4 has an input of a 64-dimensional vector. This layer contains a fully connected layer that converts its output into a single scalar value representing the final prediction result. The output is the final predicted value, with a dimension of (batch_size, 1). The output layer provides an efficient prediction result. For example, through multiple rounds of optimization, the error of the final predicted loading weight is controlled within an extremely low range (such as less than 1%), significantly improving the reliability of practical applications.

[0065] The initial LSTM model is trained based on the training data, and the mean squared error is used as the loss function. The Adam optimizer is used to iteratively optimize the parameters of the initial LSTM model to obtain the trained LSTM model.

[0066] The dynamically acquired dynamic features, categorical features, and static features are input into the trained LSTM model to obtain the predicted bulk material loading weight.

[0067] The dynamic features include the change in the height of the pile over time, which is the main time series input part of the model. Each input sequence contains multiple time steps, and each time step records the heights of three piles. The data input is obtained through material measurement. The categorical features are the categorical input part of the model, representing the types of materials loaded on the vehicle, which remain unchanged throughout the sequence. It is converted into a categorical vector through one-hot encoding. The material type data is obtained through the terminal operating system (TOS). The static features include the vehicle type and initial conditions, which do not change throughout the time series but are related to the prediction target. The vehicle type is obtained through the terminal operating system.

[0068] The LSTM layer is used to process the dynamic features, and the fully connected layer consists of two layers, which are used to process the categorical features and static features respectively.

[0069] By processing the dynamic features through the LSTM layer, the model can learn the patterns of the change in bulk material height. For example, when the bulk material height drops rapidly or changes slowly, the LSTM can accurately predict the next loading weight by memorizing historical data and paying attention to the current data.

[0070] Furthermore, optimizing the bulk material loading flow rate includes:

[0071] Calculating the maximum flow rate of bulk material loading :

[0072]

[0073] Among them is the maximum safe pressure of the bulk material loading pipeline, obtained through the characteristics of the loading pipeline, is the bulk material density. The maximum flow rate of bulk material loading is calculated through the maximum safe pressure and bulk material density to ensure avoiding overpressure risks while meeting the loading efficiency;

[0074] Define the bulk material flow rate as , and calculate the bulk material flow rate Q according to the cross-sectional area of the bulk material loading pipeline:

[0075]

[0076] Among them is the cross-sectional area of the bulk material loading pipeline;

[0077] And calculate the pressure drop :

[0078]

[0079] Among them is the friction coefficient, is the length of the bulk material loading pipeline, is the diameter of the bulk material loading pipeline;

[0080] Calculate the hydraulic energy consumption according to the bulk material flow rate and pressure drop :

[0081]

[0082] Introducing the pressure drop and hydraulic energy consumption into the optimization strategy of the bulk material loading process can systematically evaluate the relationship between the flow rate and energy consumption during the loading process, thereby realizing more efficient control of the loading flow rate. Compared with the method that simply relies on loading time or weight control, this calculation method can accurately locate the main sources of energy consumption and conduct targeted optimization;

[0083] Calculate the flow rate gradient of the bulk material in the loading pipeline according to the bulk material flow rate, and define the dissipation function :

[0084]

[0085] Among them is the flow rate gradient, is the dynamic viscosity of the bulk material, determined through experiments, and y is the height of the bulk material in the loading pipeline;

[0086] Integrate the dissipation function to calculate the dissipated energy consumption :

[0087]

[0088] By introducing a calculation method for dissipative energy consumption, the energy loss caused by internal molecular friction and turbulent effects during the flow of bulk materials in pipelines can be accurately evaluated, thus further improving the comprehensive analysis of loading energy consumption. In traditional loading systems, this energy consumption source of viscous dissipation is usually ignored, resulting in a large deviation between the actual energy consumption estimation and the predicted value;

[0089] By taking the hydraulic energy consumption and the dissipative energy consumption and summing them to obtain the total energy consumption P during the bulk material loading process, a momentum equation is constructed based on the total energy consumption P and the bulk material flow velocity:

[0090]

[0091] where is the Reynolds number, is the velocity gradient;

[0092] Initialize the bulk material flow velocity, set the boundary conditions, discretize the momentum equation using the finite difference method, and use the alternating direction implicit method to iteratively update , and stop the iteration until the bulk material flow velocity converges, and output the bulk material flow velocity obtained from the last iteration ;

[0093] By introducing the total energy consumption constraint into the momentum equation and dynamically solving for the optimal loading flow velocity, it provides a scientific basis for flow velocity regulation in bulk material loading. This method not only solves the problem of lack of scientific basis in flow velocity design in traditional loading methods but also achieves a dynamic balance between efficiency and energy consumption. By using an iterative method to gradually approach the optimal flow velocity, it can accurately adapt to the actual loading scenario;

[0094] According to the calculated bulk material flow velocity carry out bulk material loading.

[0095] Through the comprehensive analysis and optimization of integrating pressure drop, hydraulic energy consumption, and dissipative energy consumption, the loading process reaches a new balance point in terms of energy consumption control and efficiency improvement, which can significantly reduce the total energy consumption during bulk material loading. Especially in large-scale bulk material transportation scenarios, its energy-saving effect is particularly prominent, and it can flexibly adapt to different bulk material types, pipeline specifications, and loading environments.

[0096] S3. Start bulk material loading according to the predicted bulk material loading weight and bulk material flow velocity, and identify the prediction accuracy after loading is completed to optimize the LSTM model;

[0097] Specifically, starting the bulk material loading according to the predicted bulk material loading weight and flow rate means starting the bulk material loading based on the predicted bulk material loading weight and flow rate. During the bulk material loading process, the bulk material height and vehicle load data are collected in real time, and the bulk material loading is stopped when the vehicle load reaches the predicted bulk material loading weight.

[0098] In the initial stage of loading, the loading strategy is planned in advance based on the predicted bulk material loading weight and flow rate values to achieve pre-control of precise loading. The target loading weight and flow rate parameters are calculated in advance to avoid pauses caused by frequent adjustments during the loading process. The predicted results are used to guide the flow rate to ensure that the bulk material supply quantity is consistent with the dynamic demand of the loading target. Through the combined monitoring of height and weight, the loading status can be comprehensively evaluated, and the bulk material supply strategy can be adjusted in real time to improve the loading uniformity and safety. Through automatic control, when the vehicle load reaches the target value, the bulk material supply is precisely stopped, which not only eliminates the errors caused by manual intervention but also ensures the stability and consistency of the loading weight. By comparing with the predicted value, the loading error is controlled within a smaller range (such as less than 1%), significantly improving the loading accuracy. By comprehensively optimizing the flow rate and loading target in real time based on the data, the adaptability and robustness of the loading process can be significantly improved.

[0099] Further, identifying the prediction accuracy after loading to optimize the LSTM model means setting the expected vehicle load according to the vehicle type, and collecting the vehicle load and the expected load for comparison and error calculation after loading. If the error exceeds the error threshold, the LSTM model is retrained and iteratively optimized.

[0100] By associating the expected load value with the vehicle type, the system can provide personalized loading weight control for different vehicles, improving the applicability and intelligence of the system. By collecting the actual load data in real time and comparing it with the expected load, a closed-loop feedback system is constructed, enabling the prediction error to be adjusted in a timely manner. Through the setting of the error threshold, the model optimization can dynamically respond to changes in the actual working conditions, such as fluctuations in bulk material density or vehicle type adjustment, significantly improving the pertinence of the optimization. Through iterative optimization, the model can dynamically adapt to changes in the actual working conditions. For example, when the increase in bulk material humidity causes an increase in the prediction deviation of the loading weight, the retrained model can quickly learn the new characteristics.

[0101] S4. Generate a prediction record for the predicted bulk material loading weight, loading speed, and prediction accuracy each time, and store the prediction record in the database;

[0102] Specifically, generating a prediction record for the weight, loading speed, and prediction accuracy of each predicted bulk material loading means forming a prediction record for the weight, flow rate, and prediction accuracy of each predicted bulk material loading according to the time stamp, and determining whether to optimize the LSTM model based on the prediction accuracy. If optimization is required, the prediction record is added to the training group; if no optimization is needed, the prediction record is added to the validation group. The prediction records of the training group and the validation group are used to train and optimize the LSTM model.

[0103] By systematically recording the prediction data for each time, not only can the source of prediction errors be traced, but also comprehensive data support can be provided for the long-term optimization of the model. Taking the prediction accuracy as a trigger condition, it can dynamically determine whether to optimize, avoiding resource waste and ensuring that the model continuously meets the actual requirements. By clearly dividing the training group and the validation group, the performance of the model can be verified in real time during the optimization process, ensuring that the optimized model is more stable and accurate. By introducing real-time prediction records and using them for model training and optimization, the LSTM model can continuously learn new characteristics of bulk materials, thus maintaining a high prediction accuracy in the long term. Through the division of the training group and the validation group, the optimized model can maintain a stable and efficient performance in the actual scenario. Continuous training and optimization enable the model to dynamically adapt to the characteristics of bulk materials and environmental changes, ensuring the high accuracy and reliability of the long-term operation of the system.

[0104] Furthermore, storing the prediction record in the database means storing the prediction record in the database after it is generated. The database regularly verifies the integrity and security of the prediction record and synchronously uploads the prediction record to the cloud for backup.

[0105] This embodiment also provides a computer device applicable to the method for automatic high-precision quantitative loading of bulk materials at ports, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for automatic high-precision quantitative loading of bulk materials at ports as proposed in the above embodiment.

[0106] The computer device may be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads provided on the housing of the computer device. It can also be an external keyboard, touchpad, or mouse, etc.

[0107] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for automatically and accurately quantitatively loading bulk materials at ports as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, magnetic disks, or optical discs.

[0108] In summary, the present invention predicts and controls the weight of bulk material loading by constructing an LSTM model, accurately controls the vehicle loading volume, improves work efficiency, and optimizes the bulk material loading speed by collecting bulk material data, effectively improving the accuracy of bulk material loading, avoiding repeated unloading or loading caused by excessive or insufficient material loading in the vehicle, and shortening the vehicle operation cycle.

[0109] 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 preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for automatic high-precision quantitative loading of bulk materials at a port, characterized by: include, Collect bulk material data and vehicle data through sensors for preprocessing and feature extraction; Build and train the LSTM model to predict bulk material loading weight based on extracted features and optimize bulk material loading flow rate; Start bulk material loading based on the predicted bulk material loading weight and bulk material flow rate, and identify the prediction accuracy after loading is completed to optimize the LSTM model; A prediction record is generated for each predicted bulk material loading weight, loading speed and prediction accuracy, and the prediction record is stored in a database.

2. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 1, characterized in that: The method of collecting bulk material data and vehicle data through sensors for preprocessing and extracting features refers to installing a high-frequency millimeter-wave radar level meter and a laser flow rate sensor at the bulk material discharge port to obtain the bulk material height and bulk material flow rate during the loading process, and collecting bulk material data and vehicle data including bulk material type, vehicle type, and vehicle load, defining a collection time period, defining the bulk material height within the collection time period as a dynamic feature, defining the bulk material type as a category feature, and defining the vehicle type and vehicle load as static features.

3. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 2, characterized in that: The construction and training of the LSTM model to predict the bulk material loading weight based on the extracted features includes: Construct the initial LSTM model, including the input layer, LSTM layer, fully connected layer, feature fusion layer and output layer, and define the input layer input as dynamic features, category features and static features; The initial LSTM model is trained based on the training data, and the mean square error is used as the loss function. The Adam optimizer is used to iteratively optimize the initial LSTM model parameters to obtain the trained LSTM model. Input the collected dynamic features, category features, and static features into the trained LSTM model to obtain the predicted bulk material loading weight; The LSTM layer is used to process dynamic features, and the fully connected layer includes two layers, which are used to process category features and static features respectively.

4. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 3, characterized in that: The optimization of bulk material loading flow rate includes: Calculate the maximum flow rate for bulk material loading : ; in It is the maximum safety pressure of bulk material loading pipeline. is the bulk density; The bulk material flow rate is defined as , calculate the bulk material flow rate Q according to the cross-sectional area of ​​the bulk material loading pipeline: ; in The cross-sectional area of ​​the bulk material loading pipeline; And calculate the pressure drop : ; in is the friction coefficient, is the length of the bulk loading pipeline, is the diameter of the bulk material loading pipe; Calculation of hydraulic energy consumption based on bulk material flow and pressure drop : ; Calculate the flow velocity gradient of bulk materials in the loading pipeline according to the bulk material flow velocity and define the dissipation function : ; in is the flow velocity gradient, is the dynamic viscosity of bulk material, y is the height of bulk material in the loading pipeline; The dissipation function Integrate to calculate dissipated energy : ; By increasing the hydraulic energy consumption and dissipated energy The total energy consumption P in the bulk material loading process is obtained by summing up, and the momentum equation is constructed based on the total energy consumption P and the bulk material flow rate: ; in is the Reynolds number, is the velocity gradient; Initialize the bulk flow rate, set boundary conditions, discretize the momentum equation using the finite difference method, and iteratively update it using the alternating direction implicit method , until the bulk material flow rate converges, the iteration is stopped, and the bulk material flow rate obtained by the last iteration is output ; The calculated bulk material flow rate Loading bulk materials.

5. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 4, characterized in that: Starting bulk material loading according to predicted bulk material loading weight and bulk material flow rate means starting bulk material loading according to predicted bulk material loading weight and bulk material flow rate, collecting bulk material height and vehicle load data in real time during bulk material loading, and stopping bulk material loading when the vehicle load reaches the predicted bulk material loading weight.

6. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 5, characterized in that: The optimization of the LSTM model based on the identification prediction accuracy after loading is completed refers to setting the expected vehicle load according to the vehicle type, and collecting the vehicle load and the expected load after loading is completed to compare and calculate the error. If the error exceeds the error threshold, the LSTM model is retrained and iteratively optimized.

7. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 6, characterized in that: The generating prediction record of each predicted bulk material loading weight, loading speed and prediction accuracy refers to forming a prediction record of each predicted bulk material loading weight, bulk material flow rate and prediction accuracy according to the timestamp, and judging whether to perform LSTM model optimization according to the prediction accuracy. If optimization is performed, the prediction record is added to the training group. If not optimized, the prediction record is added to the verification group. The prediction records of the training group and the verification group are used to train and optimize the LSTM model.

8. The method for automatic high-precision quantitative loading of bulk materials at a port as claimed in claim 7, characterized in that: Storing the prediction record in the database means storing the prediction record in the database after the prediction record is generated, the database regularly verifies the integrity and security of the prediction record, and simultaneously uploads the prediction record to the cloud for backup.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for automatic high-precision quantitative loading of bulk materials at a port as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for automatic high-precision quantitative loading of bulk materials at a port as described in any one of claims 1 to 8 are implemented.

Images