Railway bulk loading weight control method and system based on self-learning adjustment

Abstract

This invention relates to the field of bulk material loading weight control technology, specifically to a railway bulk material loading weight control method and system based on self-learning adjustment. The method includes: real-time acquisition of the actual weight and operating condition data of the preceding car; calculation of dynamic flow control thresholds and dynamic weight compliance thresholds based on the operating condition data, preset threshold calculation formulas, and initialized formula parameters; execution of three-stage coordinated control of different flow rates in the preceding and following chutes based on the comparison results of the actual weight and the two thresholds; and after loading a single car, iterative optimization of iterable coefficients using a gradient descent algorithm based on loading errors and preset convergence conditions, for use in threshold calculation for the next car. This allows for dynamic adaptive adjustment of the thresholds, enabling precise control and fundamentally avoiding overloading problems. It not only provides stable weight control accuracy but also possesses a fully closed-loop self-learning capability, requiring no manual intervention.

Description

Technical Field

[0001] This invention relates to the field of bulk material loading and weight control technology, specifically to a railway bulk material loading and weight control method and system based on self-learning adjustment. Background Technology

[0002] In railway loading operations for bulk materials such as coal and ore, double-flute chutes enable parallel operations of "feeding the front car and laying the bottom / loading the rear car," significantly improving loading efficiency and are widely used in ports, power plants, and mines. However, existing double-flute chute loading systems have the following core technical defects in feeding and weight control:

[0003] The weight control logic has a fundamental flaw, and overweight cannot be corrected: When the car is overweight, the existing system attempts to correct the weight by adjusting the gate to discharge the residual material. However, the material has already entered the car and been measured by the track scale, so the adjustment operation is completely ineffective and instead affects the work efficiency and aggravates the wear of the actuator. At the same time, the weight of the material column in the air after the gate is closed is not taken into account, which can easily lead to the continuous falling of material into the car after the gate is closed, causing overweight. There is no root cause solution.

[0004] Fixed control thresholds result in extremely poor adaptability to working conditions: The relevant control thresholds of the existing system are all calculated using fixed coefficients, which cannot adapt to complex on-site working conditions such as belt speed fluctuations, changes in material bulk density, and chute tilt angle deviations. After the working conditions change, the weight control accuracy drops significantly, and even batches exceed the standard.

[0005] Lacking self-learning ability, serious error accumulation: The existing system cannot feed back loading errors to the threshold calculation stage. After long-term operation, errors continue to accumulate, requiring frequent on-site manual calibration, which is costly and inefficient, and cannot achieve fully automated continuous loading.

[0006] There is currently no effective solution to address the aforementioned shortcomings. Therefore, there is an urgent need for a weight control scheme for railway bulk material loading that is logically consistent, has high weight control accuracy, possesses self-learning capabilities, and can adapt to complex working conditions. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method and system for controlling the weight of bulk cargo loading on railways based on self-learning adjustment, so as to overcome the technical problems of low weight control accuracy, inability to correct overweight, poor adaptability to working conditions, and lack of self-learning ability in the prior art.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, this application provides a self-learning-based method for controlling the weight of bulk materials loaded on railway wagons, applied to a double-spindle chute loading system. In this system, the flow rate of the front chute is controlled by an angle-controllable gate, and the flow rate of the rear chute is controlled by a flow regulating valve. The method includes: The car in the front chute of the corresponding double-fork chute is designated as the front car, and the car in the rear chute of the corresponding double-fork chute is designated as the rear car. Real-time collection of actual weight and operating condition data of the front carriage; Based on operating condition data, preset threshold calculation formulas, and initialized formula parameters, calculate dynamic flow control thresholds and dynamic weight compliance thresholds; wherein, the initialized formula parameters in the preset threshold calculation formulas include target loading weight and iterable coefficients; Based on the comparison results between the actual weight and the dynamic flow control threshold and the dynamic weight compliance threshold, a three-stage coordinated control is implemented, including: When the actual weight is less than the dynamic flow control threshold, the control gate is fully opened so that the front chute can load the front car body with the maximum flow rate; and the flow rate of the rear chute is controlled within the first preset flow rate range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic flow control threshold and less than the dynamic weight compliance threshold, the target gate angle is calculated based on the angle adjustment formula to control the gate to open to the target gate angle, so that the flow rate of the front chute decreases linearly with the target gate angle; and the flow rate of the rear chute is controlled within the second preset flow range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic weight threshold, the control gate is fully closed to stop the front chute from loading the front car body; and the flow rate of the rear chute is controlled within the third preset flow rate range by the flow regulating valve. After loading is completed in a single car, the iterable coefficients are determined by iteratively optimizing the gradient descent algorithm based on the error between the final actual weight of the loaded car and the target loading weight, and the preset convergence conditions. The optimized iterable coefficients are then used to calculate the threshold for the next car.

[0009] Furthermore, in some embodiments of this application, the operating data includes: belt speed, unit load, material bulk density, and front chute angle.

[0010] Furthermore, in some embodiments of this application, the formula for calculating the dynamic flow control threshold is:

[0011] The formula for calculating the dynamic weight threshold is:

[0012] Wherein, S1 is the dynamic flow control threshold, and S2 is the dynamic weight compliance threshold; , and For the initial formula parameters, are iterable coefficients, and For the target loading weight, The maximum speed of the belt. For maximum unit load, For standard material bulk density, v, q, ρ and θ are operating data, where v is the belt speed, q is the unit load, ρ is the material bulk density, and θ is the angle of the front chute.

[0013] Furthermore, in some embodiments of this application, based on the error between the final actual weight of the loaded wagon and the target loading weight, and a preset convergence condition, iterable coefficients are determined through iterative optimization using the gradient descent algorithm, including: Calculate the error between the final actual weight of the currently loaded wagon and the target loading weight, determine the error level, and determine the iteration step size based on the error level; The objective function is the sum of the squared errors between the final actual weight of all loaded wagons and the target loading weight. Based on the iteration step size and the partial derivative of the objective function with respect to each iterable coefficient, the iterable coefficients are iteratively optimized, and it is determined whether the preset convergence condition has been met. If it has been met, the iterable coefficients are locked.

[0014] Furthermore, in some embodiments of this application, the formula for iteratively optimizing the iterable coefficients based on the iteration step size and the partial derivatives of the objective function with respect to each iterable coefficient is as follows:

[0015] in, This represents the iterable coefficients to be iteratively optimized. for The current value, for The value after iterative optimization For the objective function E, pair The partial derivatives of E; the formula for E is:

[0016] Where n is the total number of wagons that have been loaded. The final actual weight of the i-th section of the wagon after loading.

[0017] Furthermore, in some embodiments of this application, the error between the final actual weight of the currently loaded wagon and the target loading weight is calculated, and the error level is determined, and the iteration step size is determined based on the error level, including: An error less than or equal to 150 kg is defined as a micro error, and the corresponding iteration step size is set to 0.001. Errors greater than 150kg and less than or equal to 200kg are defined as minor errors, and the corresponding iteration step size is set to 0.005. Errors exceeding 200 kg are defined as out-of-range errors, and the corresponding iteration step size is set to 0.01.

[0018] Furthermore, in some embodiments of this application, the preset convergence condition is: The final actual weight of three consecutive cars was within a small margin of error compared to the target loading weight. The iteration count of the iterable coefficients may reach the maximum iteration count.

[0019] Furthermore, in some embodiments of this application, it also includes: After each iteration of optimizing the iterable coefficients, the iterable coefficients after the iteration are constrained and verified based on preset constraints.

[0020] Furthermore, in some embodiments of this application, it also includes: When the real-time collected operating condition data exceeds the preset normal range, a fault warning is triggered, and the threshold is calculated using the previously collected operating condition data.

[0021] Secondly, this application provides a railway bulk material loading weight control system based on self-learning and threshold adjustment, used to execute the above-mentioned method, including: A double-forked chute loading system, wherein the flow rate of the front chute of the double-forked chute is controlled by an angle-controllable gate, and the flow rate of the rear chute is controlled by a flow regulating valve. The weight detection subsystem is used to collect the actual weight of the front compartment in real time. The operating condition data acquisition subsystem is used to collect operating condition data of the front compartment in real time. The data processing and control subsystem is used for: Based on operating condition data, preset threshold calculation formulas, and initialized formula parameters, calculate dynamic flow control thresholds and dynamic weight compliance thresholds; wherein, the initialized formula parameters in the preset threshold calculation formulas include target loading weight and iterable coefficients; Based on the comparison results between the actual weight and the dynamic flow control threshold and the dynamic weight compliance threshold, a three-stage coordinated control is implemented, including: When the actual weight is less than the dynamic flow control threshold, the control gate is fully opened so that the front chute can load the front car body with the maximum flow rate; and the flow rate of the rear chute is controlled within the first preset flow rate range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic flow control threshold and less than the dynamic weight compliance threshold, the target gate angle is calculated based on the angle adjustment formula to control the gate to open to the target gate angle, so that the flow of the front chute decreases linearly; and the flow of the rear chute is controlled within the second preset flow range by the flow regulating valve. When the actual weight exceeds the dynamic weight threshold, the control gate is fully closed to stop the front chute from loading the front car body; and the flow rate of the rear chute is controlled within the third preset flow rate range by the flow regulating valve. After loading is completed in a single car, the iterable coefficients are determined by iteratively optimizing the gradient descent algorithm based on the error between the final actual weight of the loaded car and the target loading weight, and the preset convergence conditions. The optimized iterable coefficients are then used to calculate the threshold for the next car.

[0022] This invention relates to the field of bulk material loading weight control technology, specifically disclosing a railway bulk material loading weight control method and system based on self-learning adjustment. The method includes: determining the car corresponding to the front chute of a double-fork chute as the front car and the car corresponding to the rear chute as the rear car; real-time acquisition of the actual weight and operating condition data of the front car; calculating a dynamic flow control threshold and a dynamic weight compliance threshold based on the operating condition data, a preset threshold calculation formula, and initialized formula parameters; executing three-stage coordinated control based on the comparison result between the actual weight and the two thresholds: when the actual weight is less than the dynamic flow control threshold, the gate of the front chute is fully opened for loading; when the actual weight is between the two thresholds, the flow in the front chute is linearly reduced, and the rear chute is filled with material; when the actual weight is greater than the dynamic weight compliance threshold, the gate of the front chute is fully closed, reserving space for the air column, and the rear chute is loaded; and after loading is completed in a single car, iteratively optimizing the iterable coefficients using a gradient descent algorithm based on the final error and preset convergence conditions, for use in threshold calculation for the next car. This invention achieves dynamic adaptive adjustment of the threshold for control, thereby avoiding overweight problems at the source. It not only has stable weight control accuracy, but also has a fully closed-loop self-learning capability, requiring no manual intervention. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the weight control method for railway bulk cargo loading based on self-learning adjustment provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the self-learning adjustment-based railway bulk material loading weight control system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the principle of the railway bulk material loading weight control system based on self-learning adjustment provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the principle of a railway bulk material loading weight control method based on self-learning adjustment, provided in another embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0026] Figure 1 This is a flowchart illustrating the weight control method for railway bulk cargo loading based on self-learning adjustment provided in this embodiment of the invention. Please refer to [link / reference]. Figure 1 This embodiment may include the following steps: S101. Real-time collection of the actual weight and operating condition data of the front carriage.

[0027] Specifically, in this application, corresponding sensors are installed in the double-fork chute loading system to collect in real time the actual weight of the front car corresponding to the front chute in the double-fork chute, as well as the operating condition data of the double-fork chute loading system.

[0028] S102. Calculate the dynamic flow control threshold and dynamic weight compliance threshold based on the operating condition data, the preset threshold calculation formula, and the initialized formula parameters.

[0029] The initial formula parameters in the preset threshold calculation formula include the target loading weight and the iterable coefficient.

[0030] S103. Based on the comparison results between the actual weight and the dynamic flow control threshold and the dynamic weight compliance threshold, execute three-stage coordinated control.

[0031] Specifically, in the first stage: when the actual weight is less than the dynamic flow control threshold, the control gate is fully opened so that the front chute can load the front compartment with the maximum flow rate; and the flow rate of the rear chute is controlled within the first preset flow rate range through the flow regulating valve.

[0032] Second stage: When the actual weight is greater than or equal to the dynamic flow control threshold and less than the dynamic weight target threshold, the target gate angle is calculated based on the angle adjustment formula to control the gate to open to the target gate angle, so that the flow rate of the front chute decreases linearly with the target gate angle; and the flow rate of the rear chute is controlled within the second preset flow range by the flow regulating valve.

[0033] The third stage: When the actual weight exceeds the dynamic weight threshold, the control gate is fully closed to stop the front chute from loading the front car body; and the flow rate of the rear chute is controlled within the third preset flow rate range by the flow regulating valve.

[0034] S104. After loading is completed in a single car, based on the error between the final actual weight of the loaded car and the target loading weight and the preset convergence condition, determine the iterable coefficients through gradient descent algorithm for iterative optimization. The optimized iterable coefficients are used for threshold calculation of the next car.

[0035] Specifically, the iterable coefficients are iteratively optimized using the gradient descent algorithm. These optimized coefficients are then used to calculate the threshold for the next carriage, replacing the initial (or previously optimized) iterable coefficients for more precise control. Furthermore, a preset convergence condition is used to determine if convergence has occurred. If convergence is achieved, the iterable coefficients are locked, eliminating the need for further iterations (meaning subsequent threshold calculations use the optimized coefficients from this iteration). If convergence fails, iterative optimization continues. This approach enables dynamic adaptive adjustment and control of the threshold, fundamentally avoiding overweight issues. It not only provides stable weight control accuracy but also possesses a fully closed-loop self-learning capability, requiring no manual intervention.

[0036] Figure 2 This is a schematic diagram of the structure of the railway bulk material loading weight control system based on self-learning adjustment provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the principle of the railway bulk material loading weight control system based on self-learning adjustment provided in an embodiment of the present invention, as shown below. Figure 2 and Figure 3 As shown, based on the same inventive concept, this application also provides a railway bulk material loading weight control system based on self-learning adjustment, used to implement the above-mentioned railway bulk material loading weight control method based on self-learning adjustment. The following describes in detail the railway bulk material loading weight control method based on self-learning adjustment provided by this application, in conjunction with this system: In this application, the self-learning-adjusted railway bulk material loading control system includes a weight detection subsystem with signal connection (e.g., via industrial Ethernet), a working condition data acquisition subsystem, a data processing and control subsystem, and a double-straight chute loading system (including at least a double-straight chute actuator) connected to the data processing and control subsystem, as well as a power supply and protection subsystem that supplies power to the entire system.

[0037] Specifically, the weight detection subsystem is used to collect the actual weight m of the loading car in real time (in practical applications, since the front car, which corresponds to the front chute of the double-fork chute, finishes loading first, the calculation and control must first ensure that the front car meets the requirements, so the weight m collected here is the actual weight m of the front car), and transmits the collected signal to the data processing and control subsystem.

[0038] like Figure 3As shown in some embodiments of this application, the weight detection subsystem includes a high-precision digital track scale 1 and a data acquisition terminal with signal filtering function connected in communication with it; wherein, the measurement error of the track scale 1 is ≤ ±0.1%, and the sampling frequency is 10Hz; the data acquisition terminal is used to filter vibration and impact interference signals in the track scale acquisition signal, and transmit the processed signal to the data processing and control subsystem.

[0039] The operating condition data acquisition subsystem is used to synchronously acquire operating condition data and transmit the acquired signals to the data processing and control subsystem. In this application, the operating condition data includes belt speed v, unit load q, material bulk density ρ, and front chute angle θ.

[0040] like Figure 3 As shown, in some embodiments of this application, the working condition data acquisition subsystem includes a belt speed sensor 2 (used to collect belt speed), a gamma-ray load detector 3 (used to collect unit load), a lidar and microwave bulk density sensor 4 (used to collect material bulk density), and a dual-axis tilt sensor 5 (used to collect the angle of the front chute, i.e., the front chute angle), all mounted on the conveyor belt (configured in the double-chute loading system or connected to the input end of the double-chute loading system). The sampling frequency of each of the above sensors is ≥5Hz, and the signal transmission period is ≤100ms.

[0041] The double-fork chute loading system includes a double-fork chute actuator, which specifically comprises a front chute 6 for replenishing the front car, a rear chute 7 for loading the rear car, a gate 8 with a variable angle located at the front chute 6, and a flow regulating valve located at the rear chute 7. The entire double-fork chute actuator receives control commands from the data processing and control subsystem to complete the precise actions of replenishing the front car, loading the rear car, and regulating the flow rate.

[0042] For example, in some embodiments of this application, in the double-fork chute actuator, the gate 8 for the front chute 6 is frequency-driven and the adjustment angle range is 0°-90° (fully open is 90°, so that the front chute 6 loads the front car at the maximum flow rate, and fully closed is 0°, so that the front chute 6 stops loading the front car), with an adjustment accuracy of ±1°; the flow regulating valve at the rear chute 7 can achieve continuous linear adjustment of the flow rate from 0% to 100%.

[0043] The data processing and control subsystem, as the core hub, has a built-in iterable coefficient optimization model and gradient descent algorithm module, which is used to perform the above-mentioned threshold calculation, comparison, three-stage collaborative control, and iterative optimization processing of iterable coefficients.

[0044] For example, in some embodiments of this application, the data processing and control subsystem includes a PLC controller and an edge computing server; wherein, the response time of the PLC controller is ≤50ms, and it is used for real-time data processing and control instruction output; the edge computing server is used to run the gradient descent algorithm, complete the iterative optimization of the iterable coefficients in the threshold, data storage, and can also complete the relevant diagnosis of abnormal working conditions.

[0045] The power supply and protection subsystem is used to provide a stable power supply for the entire system and to provide overload, collision protection, and emergency shutdown protection functions.

[0046] Based on this, in some embodiments of this application, the formula for calculating the dynamic flow control threshold is as follows:

[0047] The formula for calculating the dynamic weight threshold is:

[0048] Wherein, S1 is the dynamic flow control threshold, and S2 is the dynamic weight compliance threshold; , and For the initial formula parameters, are iterable coefficients, and For the target loading weight, The maximum speed of the belt. For maximum unit load, For standard material bulk density, v, q, ρ and θ are operating data, where v is the belt speed, q is the unit load, ρ is the material bulk density, and θ is the angle of the front chute.

[0049] Figure 4 This is a schematic diagram illustrating the principle of a railway bulk material loading weight control method based on self-learning adjustment, provided in another embodiment of the present invention. Figure 4 As shown, the specific process of the data processing and control subsystem cooperating with other subsystems and actuators to achieve the entire control process includes: First, initialize the parameters, including the formula parameters that need to be initialized (as mentioned above). , , , The initialization process includes setting constraints on the iterable coefficients (for verifying the iterable coefficients after iterative optimization) and convergence conditions (for determining whether further iterative optimization of the iterable coefficients is needed). It should be noted that, for the subsequent comparison steps to proceed smoothly, the initialization in this application should be based on a reasonable range, ensuring that the calculated S1 is less than S2 (for example, the threshold range of S1 can be set to...). ×95% - 98%, the S2 threshold range is ×99% - 99.5).

[0050] For example, in some embodiments, the initial value and constraint range of the iterable coefficient are: the initial value of k1 is 0.95, and the constraint range is 0.93 ≤ k1 ≤ 0.97; the initial value of k2 is 0.03, and the constraint range is 0.02 ≤ k2 ≤ 0.04; the initial value of k3 is 0.02, and the constraint range is 0.01 ≤ k3 ≤ 0.03; the initial value of k4 is 0.99, and the constraint range is 0.985 ≤ k4 ≤ 0.995; the initial value of k5 is 0.005, and the constraint range is 0.003 ≤ k5 ≤ 0.007; the initial value of k6 is 0.01, and the constraint range is 0.008 ≤ k6 ≤ 0.012.

[0051] On this basis, obtain the working condition data collected in real time by the above-mentioned subsystem, substitute the working condition data into the calculation formulas of S1 and S2, calculate S1 and S2, and then determine to output the control instruction of the three-stage collaborative control mentioned above to the double-fork chute actuator for the double-fork chute actuator to execute, perform three-stage collaborative control, and realize the parallel operation of feeding the front carriage and loading the rear carriage.

[0052] Specifically, in the three-stage collaborative control: The first stage, when m < S1: is the rapid feeding stage, control the gate to open fully to 90° to feed the front chute to the front vehicle at the maximum flow rate, quickly increase the weight of the front carriage, and shorten the loading time of the front carriage; the rear chute stands by at a first preset flow rate range (a relatively small flow rate), such as ≤ 10% of the maximum flow rate, to avoid empty discharging of materials.

[0053] The second stage, when S1 ≤ m < S2: calculate the target gate angle α, control the flow rate of the front chute to decrease linearly with α to avoid excessive growth of the weight of the front vehicle carriage resulting in overweight; at the same time, control the rear chute to lay the bottom for the rear carriage at a second preset flow rate range (a relatively large flow rate), such as 30% - 50% of the maximum flow rate, to prepare for the main body loading of the rear carriage and improve the overall loading efficiency. Among them, the calculation formula for the target gate angle α is: α = 90°×(M0 - m) / (M0 - S1) The third stage, when m ≥ S2: control the gate to close fully to 0°, cut off the feeding of the front chute, and reserve the weight margin of the air material column after the gate is closed in advance to avoid overweight caused by continuous falling of materials into the carriage from the source; at the same time, control the rear loading chute to load the main body of the rear vehicle at a third preset flow rate range (the maximum flow rate) to achieve seamless connection of the front and rear vehicle operations.

[0054] Furthermore, in this application, after loading a single car (i.e., a front car) is completed based on the above three-stage coordinated control, its final actual weight m is collected through the above weight detection subsystem (it can be understood that the final actual weight is also the weight of the car collected in real time, but it is collected at the moment after loading is completed, so it is also represented by m), the loading weight error Δm=m-M0 is calculated, and the iterable coefficients are iteratively optimized through gradient descent algorithm in combination with the preset convergence conditions, specifically including: Step 1: Calculate the error between the final actual weight of the currently loaded wagon and the target loading weight, determine the error level, and determine the iteration step size based on the error level.

[0055] Specifically, three levels of error can be set: micro-error, minor error, and excessive error. An error of 150 kg or less is defined as micro-error, with an iteration step size of 0.001; an error greater than 150 kg but less than or equal to 200 kg is defined as minor error, with an iteration step size of 0.005; and an error greater than 200 kg is defined as excessive error, with an iteration step size of 0.01. In practical applications, for each level, system warnings can be triggered, and abnormal operating condition checks can be initiated. It is understood that micro-error and minor error both represent achieving the accuracy target, while excessive error represents failing to meet the target.

[0056] Step 2: Use the sum of squared errors between the final actual weight of all loaded wagons and the target loading weight as the objective function.

[0057] Specifically, the objective function is expressed as:

[0058] Where E represents the objective function, which is the sum of squared errors between the final actual weight of all loaded wagons and the target loading weight; n is the number of all loaded wagons. The final actual weight of the i-th section of the wagon after loading.

[0059] Step 3: Iteratively optimize the iterable coefficients based on the iteration step size and the partial derivatives of the objective function with respect to each iterable coefficient.

[0060] Specifically, the formula for iterable coefficients in iterative optimization is:

[0061] in, This represents the iterable coefficients to be iteratively optimized. for The current value, for The value after iterative optimization For the objective function E, pair The partial derivatives of .

[0062] The formula for calculating the partial derivatives of the above iterable coefficients is as follows:

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] Step 4: Determine if the convergence condition is met, i.e., determine if convergence has occurred. For example, determine if the error between the final actual weight of three consecutive cars and the target loading weight is within a small error range, or if the number of iterations of the iterable coefficients has reached the maximum number of iterations (to avoid system oscillations caused by excessive iteration). If the conditions are met, convergence has occurred, and the iterable coefficients are locked (i.e., all subsequent calculations will use the current iterable coefficients). If the conditions are not met, convergence has not occurred, and the iterable coefficients are continuously iterated and optimized.

[0069] It should be noted that when iteratively optimizing the iterable coefficients, the front carriage that has been loaded is not adjusted to avoid unnecessary intervention. Its error data is only used for threshold calculation and optimization of the next carriage, i.e., the rear carriage.

[0070] Furthermore, in some embodiments of this application, after completing an iterative optimization (that is, replacing the previous iterable coefficient with the newly obtained iterable coefficient), the iterable coefficient after iterative optimization can also be constrained and verified based on preset constraints. Only after the iteratively optimized coefficient passes the verification can it be used as the iterable coefficient when calculating the threshold of the next carriage (if it fails, the iterable coefficient before iterative optimization is used).

[0071] This process continues until all carriages are loaded, storing the final iterable coefficients, which can be used as the initial values ​​for the iterable coefficients of subsequent loading operations of similar materials or the same vehicle model, and then ending the process.

[0072] Furthermore, in some embodiments of this application, a fault warning is triggered when the real-time collected operating data is detected to exceed a preset normal range. In this case, the currently collected operating data is discarded, and the threshold is calculated using the previously collected operating data. Additionally, the gate can be immediately closed and an emergency shutdown protection mechanism can be triggered when a sudden belt stop or material jamming is detected.

[0073] The following detailed description, with reference to specific embodiments, illustrates the self-learning adjustment-based railway bulk material loading weight control method provided by the present invention: This embodiment applies to a rapid loading scenario on a coal mine railway, using C70 type railway carriages, with a target loading weight. =70t, standard material bulk density =1.1t / m³, maximum belt speed =2m / s, maximum unit load =1.5t / m.

[0074] The specific configurations of each subsystem and actuator are as follows: The weight detection subsystem uses a high-precision digital rail scale (measurement error ±0.1%, sampling frequency 10Hz) and is equipped with a data acquisition terminal with digital filtering function.

[0075] The working condition data acquisition subsystem includes a belt speed sensor (measurement range 0-5m / s, accuracy ±0.01m / s), a gamma-ray material load detector (measurement range 0-3t / m, accuracy ±0.05t / m), a lidar combined with a microwave bulk density sensor (measurement range 0.8-1.5t / m³, accuracy ±0.02t / m³), and a dual-axis tilt sensor (measurement range 0-90°, accuracy ±0.1°). The sampling frequency of each sensor is 10Hz, and the signal transmission period is 100ms.

[0076] The double-fork chute actuator includes a front chute, a rear chute, a gate (adjustable angle 0°-90°, adjustment accuracy ±1°) and a flow regulating valve (0%-100% flow rate continuous adjustment).

[0077] The data processing and control subsystem includes a PLC controller (response time 40ms) and an edge computing server.

[0078] The power supply and protection subsystem adopts dual redundant power supply and UPS uninterruptible power supply, and is equipped with overload protection, collision protection and emergency shutdown modules.

[0079] Based on the above system, the specific implementation steps of the railway bulk material loading weight control method based on self-learning adjustment in this embodiment are as follows: Step 1: Parameter initialization, input target loading weight =70t, Enter =2m / s, =1.5t / m, =1.1t / m³, set the initial values ​​of the iterable coefficients: k1=0.95, k2=0.03, k3=0.02, k4=0.99, k5=0.005, k6=0.01, and the coefficient constraint range is the same as the interval range of the corresponding iterable coefficients mentioned in the above embodiment; set the algorithm convergence condition: the error of 3 consecutive carriages ≤150kg, or the maximum number of iterations of 20 is reached; the initial range of S1 is 66.5t-68.6t, and the initial range of S2 is 69.3t-69.65t.

[0080] Step 2: Real-time synchronous data acquisition. When the first car reaches the corresponding position of the front chute, the track scale collects the actual weight m of the car, and the working condition acquisition subsystem synchronously collects real-time working condition data: v = 2m / s ( =1), q=1.5t / m ( =1), ρ=1.1t / m³ =1), θ=30°, ( All data is transmitted to the PLC controller in real time.

[0081] Step 3: Dynamic threshold calculation. Using initial coefficients substituted into the formula, the calculation is as follows: ;

[0082] Step 4: Three-stage coordinated control, during the loading process of the first car: When m < 67.2t, the gate is fully opened to 90°, the front chute is fed with the maximum flow rate, and the rear chute is on standby with 10% flow rate; when 67.2t ≤ m < 69.18t, the target gate angle α is calculated in real time as 90° × (70 - m) / (70 - 67.2), the flow rate of the front chute decreases linearly with α, and the rear chute is used to lay the bottom of the rear vehicle with 40% flow rate; when m ≥ 69.18t, the gate is fully closed to 0°, the feeding of the front chute is cut off, and the rear chute is switched to the maximum flow rate for loading the main body of the rear vehicle.

[0083] Step 5: Determine the error level and coefficient self-learning iterative optimization. The final actual weight of the first car is detected to be m=70.3t, Δm=+300kg, |Δm|>200kg, which is an over-limit warning. The error level is recorded as an over-limit warning. The iteration step size is determined to be η=0.01, and the objective function is E=(70.3-70)²=0.09. The partial derivatives of each coefficient are calculated according to the formula, and the iterable coefficients are iteratively optimized. After optimization, k4=0.985, k6=0.012, and the remaining coefficients are slightly adjusted. All coefficients are within the constraint range and have not reached the convergence condition. The optimized coefficients are used as the threshold calculation benchmark for the second car. Substituting into the threshold calculation formula, the weight of the second car is S2≈68.72t. The door closing threshold is lowered in advance to avoid overweight.

[0084] Step 6: If it is determined that not all cars have been loaded, return to step 2 and repeat the above steps for the second car.

[0085] The operating data changes after the second car reaches the position of the front chute: ρ=1.0t / m³ ( ), θ=25° v=1.8m / s ), q=1.2t / m ( The iterable coefficients optimized based on the first car were used to calculate S1=67.27t and S2=68.66t. Loading was carried out according to the three-stage collaborative control, and the final actual weight m=69.8t, Δm=-200kg, which is a slight error. The iteration step size η=0.005. The iterable coefficients were optimized again, k4 was adjusted to 0.988, k6 was adjusted to 0.011, and the other coefficients were slightly adjusted for subsequent calculation and control of the third car.

[0086] This embodiment can continuously complete the loading of 50 carriages. Except for the first carriage, the error of the remaining 45 carriages is ≤150kg, and the error of the 4 carriages is between 150kg and 200kg, with high weight control accuracy. No manual calibration is required during the loading process, realizing fully automatic continuous operation, and the loading efficiency is significantly improved compared with the existing system.

[0087] The self-learning-based weight control method for railway bulk cargo loading provided in this application has the following advantages compared to existing technologies: It abandons the ineffective logic of "reverse adjustment after overloading" in existing technologies, and realizes the pre-reservation of the weight margin of the air column by using the weight threshold. With the three-stage collaborative control, the flow rate is accurately controlled and the gate is closed before the material falls into the car, thus avoiding the overloading problem at the source. At the same time, the judgment of the error of a single car is only used for the threshold optimization of the next car, without making ineffective interventions, which greatly improves the operation efficiency and reduces the wear of the actuator.

[0088] By incorporating operating condition data into the threshold calculation formula, the threshold can be dynamically adjusted according to changes in on-site operating conditions, completely solving the problem of poor adaptability of fixed threshold operating conditions. Through continuous loading verification under multiple on-site operating conditions, the weight control accuracy can be kept stable within ±200kg for a long time, and the error of more than 90% of the carriages can be controlled within ±150kg, which is far superior to the industry's conventional weight control level of ±500kg.

[0089] A closed-loop control logic was constructed, encompassing loading results, error analysis, coefficient optimization, threshold updates, and precise loading. A gradient descent algorithm was used to determine the iteration step size through error classification, enabling adaptive optimization of the threshold coefficient. Small errors were fine-tuned, and large errors were quickly corrected, balancing optimization accuracy and system stability. Frequent manual calibration was eliminated, significantly reducing labor costs and enabling continuous fully automated loading operations.

[0090] Strict constraints are set for the iterable coefficients to prevent excessive coefficient iteration from causing system out of control. At the same time, fault-tolerant logic and protection mechanisms are set for abnormal operating conditions to deal with on-site emergencies such as sensor failure, sudden belt stoppage, and material blockage, ensuring long-term stable operation of the system.

[0091] The hardware structure can fully utilize the relevant components of the existing double-fork chute loading system. Only the addition of sensors and optimization of algorithm logic are required. No large-scale equipment modification is needed, and the modification cost is low. It can be quickly promoted and applied in existing bulk material loading scenarios in mines, ports, and power plants.

[0092] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0093] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0094] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0095] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0096] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0097] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0098] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0099] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0100] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A weight control method for railway bulk material loading based on self-learning adjustment, applied to a double-straight chute loading system, characterized in that, In the double-fork chute loading system, the flow rate of the front chute is controlled by an angle-controllable gate, and the flow rate of the rear chute is controlled by a flow regulating valve. The method includes: The car in the front chute of the corresponding double-fork chute is designated as the front car, and the car in the rear chute of the corresponding double-fork chute is designated as the rear car. Real-time collection of actual weight and operating condition data of the front carriage; Based on operating condition data, preset threshold calculation formulas, and initialized formula parameters, calculate dynamic flow control thresholds and dynamic weight compliance thresholds; wherein, the initialized formula parameters in the preset threshold calculation formulas include target loading weight and iterable coefficients; Based on the comparison results between the actual weight and the dynamic flow control threshold and the dynamic weight compliance threshold, a three-stage coordinated control is implemented, including: When the actual weight is less than the dynamic flow control threshold, the control gate is fully opened so that the front chute can load the front car body with the maximum flow rate; and the flow rate of the rear chute is controlled within the first preset flow rate range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic flow control threshold and less than the dynamic weight compliance threshold, the target gate angle is calculated based on the angle adjustment formula to control the gate to open to the target gate angle, so that the flow rate of the front chute decreases linearly with the target gate angle; and the flow rate of the rear chute is controlled within the second preset flow range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic weight threshold, the control gate is fully closed to stop the front chute from loading the front car body; and the flow rate of the rear chute is controlled within the third preset flow rate range by the flow regulating valve. After loading is completed in a single car, the iterable coefficients are determined by iteratively optimizing the gradient descent algorithm based on the error between the final actual weight of the loaded car and the target loading weight, and the preset convergence conditions. The optimized iterable coefficients are then used to calculate the threshold for the next car.

2. The method according to claim 1, characterized in that, The operating data includes: belt speed, unit load, material bulk density, and front chute angle.

3. The method according to claim 2, characterized in that, The formula for calculating the dynamic flow control threshold is: The formula for calculating the dynamic weight threshold is: Wherein, S1 is the dynamic flow control threshold, and S2 is the dynamic weight compliance threshold; , and For the initial formula parameters, are iterable coefficients, and For the target loading weight, This is the maximum speed of the belt. For maximum unit load, For standard material bulk density, v, q, ρ and θ are operating data, where v is the belt speed, q is the unit load, ρ is the material bulk density, and θ is the angle of the front chute.

4. The method according to claim 1, characterized in that, Based on the error between the final actual weight of the loaded wagon and the target loading weight, and the preset convergence condition, the iterable coefficients for iterative optimization using the gradient descent algorithm are determined, including: Calculate the error between the final actual weight of the currently loaded wagon and the target loading weight, determine the error level, and determine the iteration step size based on the error level; The objective function is the sum of the squared errors between the final actual weight of all loaded wagons and the target loading weight. Based on the iteration step size and the partial derivative of the objective function with respect to each iterable coefficient, the iterable coefficients are iteratively optimized, and it is determined whether the preset convergence condition has been met. If it has been met, the iterable coefficients are locked.

5. The method according to claim 4, characterized in that, Based on the iteration step size and the partial derivatives of the objective function with respect to each iterable coefficient, the formula for iteratively optimizing the iterable coefficients is as follows: in, This represents the iterable coefficients to be iteratively optimized. for The current value, for The value after iterative optimization For the objective function E, pair The partial derivatives of E; the formula for E is: Where n is the total number of wagons that have been loaded. The final actual weight of the i-th section of the wagon after loading.

6. The method according to claim 5, characterized in that, Calculate the error between the final actual weight of the currently loaded wagon and the target loading weight, determine the error level, and determine the iteration step size based on the error level, including: An error less than or equal to 150 kg is defined as a micro error, and the corresponding iteration step size is set to 0.

001. Errors greater than 150kg and less than or equal to 200kg are defined as minor errors, and the corresponding iteration step size is set to 0.

005. Errors exceeding 200 kg are defined as out-of-range errors, and the corresponding iteration step size is set to 0.

01.

7. The method according to claim 6, characterized in that, The preset convergence condition is: The final actual weight of three consecutive cars was within a small margin of error compared to the target loading weight. The iteration count of the iterable coefficients may reach the maximum iteration count.

8. The method according to claim 4, characterized in that, Also includes: After each iteration of optimizing the iterable coefficients, the iterable coefficients after the iteration are constrained and verified based on preset constraints.

9. The method according to claim 1, characterized in that, Also includes: When the real-time collected operating condition data exceeds the preset normal range, a fault warning is triggered, and the threshold is calculated using the previously collected operating condition data.

10. A railway bulk material loading weight control system based on self-learning and threshold adjustment, used to perform the method as described in any one of claims 1-9, characterized in that, include: A double-forked chute loading system, wherein the flow rate of the front chute of the double-forked chute is controlled by an angle-controllable gate, and the flow rate of the rear chute is controlled by a flow regulating valve. The weight detection subsystem is used to collect the actual weight of the front compartment in real time. The operating condition data acquisition subsystem is used to collect operating condition data of the front compartment in real time. The data processing and control subsystem is used for: Based on operating condition data, preset threshold calculation formulas, and initialized formula parameters, calculate dynamic flow control thresholds and dynamic weight compliance thresholds; wherein, the initialized formula parameters in the preset threshold calculation formulas include target loading weight and iterable coefficients; Based on the comparison results between the actual weight and the dynamic flow control threshold and the dynamic weight compliance threshold, a three-stage coordinated control is implemented, including: When the actual weight is less than the dynamic flow control threshold, the control gate is fully opened so that the front chute can load the front car body with the maximum flow rate; and the flow rate of the rear chute is controlled within the first preset flow rate range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic flow control threshold and less than the dynamic weight compliance threshold, the target gate angle is calculated based on the angle adjustment formula to control the gate to open to the target gate angle, so that the flow rate of the front chute decreases linearly with the target gate angle; and the flow rate of the rear chute is controlled within the second preset flow range by the flow regulating valve. When the actual weight is greater than or equal to the dynamic weight threshold, the control gate is fully closed to stop the front chute from loading the front car body; and the flow rate of the rear chute is controlled within the third preset flow rate range by the flow regulating valve. After loading is completed in a single car, the iterable coefficients are determined by iteratively optimizing the gradient descent algorithm based on the error between the final actual weight of the loaded car and the target loading weight, and the preset convergence conditions. The optimized iterable coefficients are then used to calculate the threshold for the next car.

Images