Train bulk material intelligent thawing control method and system based on multi-source data fusion

Abstract

The application discloses a train bulk material intelligent thawing control method and system based on multi-source data fusion and belongs to the technical field of bulk material railway transportation. The system comprises a train carriage, a train track, a heat preservation cover and a framework thereof, an end sealing curtain, a steam main pipe and a main pipe support, a main pipe steam regulating valve, steam branch pipes of multiple independent spraying sections along the length direction of the carriage and electric regulating valves of the sections, an infrared thermal imager, a near-infrared moisture meter, an RFID card reader, an environmental temperature and humidity sensor, a PLC and a self-learning optimization module. The PLC receives an infrared thermal imaging temperature cloud picture, identifies a frozen area and maps the frozen area to corresponding steam branch pipe sections; receives near-infrared moisture meter moisture content data to adjust steam flow; receives RFID historical thawing records and generates initial set values in combination with a self-learning module; and outputs control signals to each electric regulating valve. The application realizes more than 35% reduction of thawing energy consumption, 60% improvement of thawing uniformity, 99% thawing qualified rate and intelligent and unmanned operation.

Description

Technical Field

[0001] This invention belongs to the field of bulk material railway transportation technology, and particularly relates to an intelligent thawing control method and system for bulk materials transported by train based on multi-source data fusion. Background Technology

[0002] In cold winters, bulk materials transported by train freeze due to their moisture content, adhering firmly to the inner walls of the carriages, making unloading difficult. Various steam defrosting devices have been developed in the prior art, such as those using fixed insulation covers, laying steam pipes, and heating the outer walls of the carriages by injecting steam. However, these solutions have the following inherent drawbacks:

[0003] (1) Lack of perception of the frozen state: Regardless of the degree of freezing of the materials in the carriage, a fixed steam injection flow rate and time are used, resulting in local overheating or incomplete thawing.

[0004] (2) Inability to differentiate heating: Steam branch pipes are usually evenly arranged and controlled as a whole, which cannot be used to target and enhance heating in areas that are severely frozen in the carriage, resulting in energy waste and uneven thawing.

[0005] (3) Parameters are set entirely by human experience: parameters such as thawing time and steam volume depend on on-site observation and manual adjustment by operators. There are large differences in operation between different shifts and different personnel, and successful experience cannot be accumulated and reused.

[0006] (4) The change in the moisture content of the material itself is not considered: In actual production, the moisture content of different batches of the same mineral can fluctuate by 6%-14% (absolute percentage), but the existing system cannot detect the moisture content, resulting in steam waste when the moisture content is low and insufficient thawing when the moisture content is high.

[0007] Therefore, there is an urgent need for an intelligent steam defrosting system with multi-source sensing capabilities, the ability to adaptively adjust heating strategies, and the capacity for self-learning optimization, in order to solve the aforementioned technical problems. Summary of the Invention

[0008] This invention provides a method and system for intelligent thawing control of bulk materials on trains based on multi-source data fusion. The steam thawing system and method for bulk materials on trains, based on infrared thermal imaging, online detection of material moisture, RFID historical data mining and self-learning control, achieves precise, efficient, adaptive and self-optimizing thawing operations.

[0009] To achieve the above-mentioned objectives, the first objective of this invention is to provide an intelligent thawing control system for bulk cargo on trains based on multi-source data fusion, comprising: Insulating covers are used to contain train carriages that are about to thaw. The steam supply and regulation unit includes a main steam pipe and several independently controlled steam branch pipes. The steam branch pipes are divided into multiple independent injection sections along the length of the carriage inside the insulation cover, and each section is equipped with an electric regulating valve. Multi-source sensing units include an infrared thermal imager, a near-infrared moisture meter, an RFID reader, and an ambient temperature and humidity sensor; Intelligent control unit, including PLC and self-learning optimization module; The PLC is used for: Receive temperature cloud image data from an infrared thermal imager, identify frozen areas and map them to the corresponding steam branch pipe sections; Receive moisture content data from a near-infrared moisture meter and adjust the reference value of total steam flow rate according to the moisture content correction coefficient; Receive historical unfreezing records read by the RFID reader and combine them with the initial settings generated by the self-learning optimization module; Output control signals to the electric regulating valves of each steam branch section.

[0010] Preferably, the self-learning optimization module adopts support vector regression or random forest algorithm, with the goal of minimizing thawing time and steam consumption; when the energy consumption and time deviation of multiple consecutive thawing operations of the same carriage type, approximate moisture content and ambient temperature are all less than the set threshold, a new standard heating curve is automatically generated and stored in the database.

[0011] Preferably, the PLC has a built-in frozen area recognition algorithm that performs grayscale and threshold segmentation on the temperature cloud map collected by the infrared thermal imager to identify continuous areas with temperatures below freezing point, and maps the coordinate range of each area to the corresponding steam branch pipe section based on the coordinate system calibration of the outer wall of the carriage.

[0012] Preferably, the PLC adjusts the steam flow rate according to the real-time moisture content and the following rules: for every 1% increase in moisture content, the total steam flow rate increases by 5% to 8%; for every 1% decrease in moisture content, the total steam flow rate decreases by 5% to 8%.

[0013] Preferably, the steam branch pipe is divided into 4-6 independent sections along the length of the carriage, each section being 1.5m to 3.8m long; each section of the branch pipe is provided with injection holes facing the side wall and bottom of the carriage, with a hole diameter of 3mm to 6mm and a hole spacing of 80mm to 150mm.

[0014] The second objective of this invention is to provide a method for intelligent thawing control of bulk materials on trains based on multi-source data fusion, comprising: S1. Identify the car number to be thawed using an RFID reader, and retrieve the input parameters and output results of the car in historical thaw operations; at the same time, use a near-infrared moisture meter to detect the real-time moisture content of the materials in the car online. S2. Based on historical thawing records, the initial settings for this thawing are generated using a self-learning optimization module. The initial settings include the total steam volume, valve opening degree of each section, and pulse mode. The baseline value of the total steam volume is then corrected based on the real-time moisture content. S3. Before the thawing begins, use an infrared thermal imager to collect temperature cloud maps of the outer wall of the carriage and the surface of the materials. Use a frozen area identification algorithm to determine continuous areas with temperatures below freezing point and map the coordinate range of the frozen area to the corresponding steam branch pipe section inside the insulation cover. S4. Adjust the opening of the electric regulating valve of the corresponding steam branch section according to the mapping result, and combine it with pulse mode to control steam injection to achieve targeted heating for severely frozen areas. S5. During the thawing process, continuously monitor the temperature field changes and dynamically adjust the valve opening of each section; after thawing is completed, record the data of the entire operation and feed it back to the self-learning optimization module to update the thawing parameter model.

[0015] Preferably, the correction of the baseline value of total steam based on the real-time moisture content is specifically as follows: Set a moisture content correction factor. When the real-time moisture content increases by 1%, the benchmark value of total steam flow increases by 5% to 8%; when the real-time moisture content decreases by 1%, the benchmark value of total steam flow decreases by 5% to 8%.

[0016] Preferably, after acquiring temperature cloud images using an infrared thermal imager, the following image processing procedures are included: The temperature cloud map was converted to grayscale and segmented with 0℃ as the threshold to extract the low-temperature region; based on the coordinate system calibration of the outer wall of the carriage, the coordinate range of the low-temperature region in the length direction was calculated.

[0017] Preferably, the steam injection is controlled using a pulse mode, specifically as follows: Calculate the proportion of the frozen area to the total area of ​​the outer wall of the carriage; When the ratio is greater than 40%, continuous spray mode is used; When the ratio is between 20% and 40%, pulse jet mode is used, with a jet cycle of 3 minutes of spraying and 1 minute of stopping. When the ratio is less than 20%, a reduced injection mode is adopted, limiting the maximum valve opening to no more than 40%.

[0018] Preferably, the thawing is completed when infrared thermal imaging detects that the temperature of all areas of the outer wall of the carriage is above 2°C and remains above 2°C for more than 3 minutes.

[0019] The advantages and positive effects of this application are: This invention achieves precise steam thawing by integrating infrared and moisture meter sensing and driving zone control. Actual measurements show that thawing uniformity is improved by more than 60% and energy consumption is reduced by 35%.

[0020] This invention solves the problem of traditional systems being unable to adapt to fluctuations in moisture content by detecting moisture content in real time and automatically adjusting the total amount of steam.

[0021] This invention, through the synergy of RFID and a self-learning model, enables the system to start from its historical best mode and continuously optimize itself.

[0022] This invention automates the entire process from identification and decision-making to judgment, significantly reducing human intervention. Attached Figure Description

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

[0024] Figure 1 This is a schematic diagram of the overall system structure of a preferred embodiment of the present invention; Figure 2 for Figure 1 A side sectional view showing the infrared thermal imager and the layout of the zoned steam branch pipes; Figure 3 A schematic diagram of temperature cloud map and zonal mapping of frozen areas identified by infrared thermal imaging; Explanation of markings in the diagram: 1. Train carriage; 2. Train track; 3. Main steam regulating valve; 4. Main steam pipe; 5. Main pipe support; 6. Steam branch pipe; 7. Insulation cover; 8. Insulation cover frame; 9. End sealing curtain; 10. Infrared thermal imager; 11. PLC; 12. Near-infrared moisture meter; 13. RFID reader; 14. Ambient temperature and humidity sensor; 15. Self-learning optimization module; 16. Steam branch pipe electric valve. Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1 to 3 A train bulk material intelligent thawing control system based on multi-source data fusion mainly includes: The heat insulation cover 7 spans across the train track 2 and is used to contain the train carriage 1 to be thawed. It has end sealing curtains 9 at both ends. The heat insulation cover 7 is installed on the heat insulation cover frame 8. The steam supply and regulation unit includes a steam main pipe 4, several sets of independently controlled steam branch pipes 6 and a main steam regulating valve 3. The steam main pipe 6 is supported by several steam supports 5. The steam branch pipes 6 are divided into multiple independent injection sections along the length of the carriage inside the insulation cover. Each section is equipped with a steam branch pipe electric valve 16. Multi-source sensing unit, including: Infrared thermal imager 10 is installed in the inner ceiling and side walls of the insulation cover 7 to collect real-time temperature distribution images of the outer wall of the carriage and the surface of the material. At least one near-infrared moisture meter 12 is installed above the entrance of the carriage for online detection of material moisture content; RFID reader 13, installed on one side of the track, is used to identify the electronic tags of the carriage and read the historical thawing records of the carriage (including input parameters and output results of multiple thawing). An ambient temperature and humidity sensor 14 is installed on the outside of the insulation cover to detect ambient temperature and relative humidity; The intelligent control unit includes a PLC11 and a self-learning optimization module 15 connected thereto; The PLC 11 is configured to: receive temperature cloud image data from the infrared thermal imager 10, determine the required steam volume for each spray section using a built-in frozen area identification algorithm; receive moisture content data from the near-infrared moisture meter 12, adjust the total steam flow rate using a moisture content correction coefficient; receive historical thawing records read by the RFID reader 13, and receive the initial thawing setting value generated by the self-learning optimization module 15 based on the historical thawing records; and receive environmental parameters from the ambient temperature and humidity sensor 14; the PLC 11 integrates the above information and outputs control signals to the steam branch electric valves 16 and main steam regulating valves 3 of each steam branch section. The self-learning optimization module 15 records the input parameters (material type, moisture content, ambient temperature, initial compartment temperature, and frozen area distribution) and output results (thawing time, steam consumption, thawing pass rate, and valve opening curves for each section) for each thawing operation, and uses regression analysis or neural network algorithms to iteratively update the thawing parameter model. Furthermore, during the thawing process, the infrared thermal imager 10 continuously monitors the temperature field changes, and the PLC 11 dynamically adjusts the valve opening of each section according to the real-time temperature field data until thawing is completed. After thawing is completed, the self-learning optimization module 15 records the data of the entire thawing process and updates the self-learning model to form a closed-loop iterative optimization.

[0026] The infrared thermal imager 10 is an online uncooled focal plane detector with a resolution of not less than 320×240 pixels and a temperature measurement range of -20℃ to 150℃. It transmits temperature cloud map data to the PLC via Ethernet communication.

[0027] The frozen area identification algorithm built into the PLC11 includes: converting the temperature cloud map collected by the infrared thermal imager to grayscale and threshold segmentation (threshold set to 0℃), identifying continuous areas with temperatures below freezing, and mapping the coordinate range of each area to the corresponding steam branch pipe section based on the coordinate system calibration of the outer wall of the carriage, automatically increasing the valve opening of the section or extending the pulse injection time of the section.

[0028] The steam branch pipe 6 is divided into 4-6 independent sections along the length of the carriage, each section being 1.5-3.8m long and equipped with an electric regulating ball valve to achieve differentiated spraying in each section; each section of the branch pipe is equipped with spray holes facing the side wall and bottom of the carriage, with a spray hole diameter of 3-6mm and a hole spacing of 80-150mm.

[0029] The near-infrared moisture meter 12 is a diffuse reflection type, with a measurement wavelength range of 1200-2500nm and a measurement accuracy of ±0.2%. The PLC adjusts the steam flow reference value according to the real-time moisture content according to the following rules: for every 1% increase in moisture content, the total steam flow increases by 5%-8%; for every 1% decrease in moisture content, the total steam flow decreases by 5%-8%.

[0030] The RFID reader 13 communicates with the factory production management system (ERP / MES) database to read the historical thawing records of the carriage (including operating parameters and result data of at least the last 3 thawings). The self-learning optimization module generates the initial settings for this thawing based on these historical data (including total steam volume, valve opening degree of each section, heating time, and pulse mode), rather than directly calling a single "optimal parameter".

[0031] The self-learning optimization module 15 adopts support vector regression or random forest algorithm, with the optimization objective of minimizing thawing time and steam consumption. After each thawing operation is completed, the model parameters are updated. When the energy consumption and time deviation of 5 consecutive thawing operations under the same carriage type, approximate moisture content (deviation within ±0.5%) and ambient temperature (deviation within ±2℃) are all less than the threshold (energy consumption deviation ≤5%, time deviation ≤8%), a new standard heating curve is automatically generated and stored in the database.

[0032] The multi-source sensing unit specifically includes: Infrared thermal imagers (3 units installed on the top and side walls of the insulation cover, which can be increased or decreased according to the length of the insulation cover) acquire real-time temperature distribution cloud maps of the outer wall of the carriage and identify the location and area of ​​frozen areas.

[0033] Near-infrared moisture meter (installed at the material conveyor belt or car inlet) detects the moisture content of materials online with a measurement accuracy of ±0.2% (designed for a moisture content range of 6%-20%, effectively distinguishing fluctuations of 6%-14% in actual production).

[0034] The RFID reader reads the electronic tags in the carriage, connects to the factory database, and retrieves the historical unfreezing records of that carriage (including the input parameters and output results of the most recent unfreezing operations).

[0035] An ambient temperature and humidity sensor is used to compensate for the impact of the environment on the thawing process.

[0036] The steam supply and regulation unit has been improved to allow for independent zone control: the steam branch pipes inside the insulation hood are divided into multiple sections along the length of the carriage (e.g., 4-6 sections, each 1.5m to 3.8m in length). Each section is equipped with an electric regulating valve, which can independently adjust the steam flow rate of that section. Combined with the coordinates of the frozen area identified by infrared thermal imaging, the PLC can automatically increase the valve opening of severely frozen sections or extend the heating time of those sections, achieving "heating on demand".

[0037] The intelligent control unit includes a PLC and a self-learning optimization module. The PLC has a built-in frozen area recognition algorithm (image threshold segmentation, coordinate system mapping) and a moisture content correction algorithm. The self-learning optimization module uses machine learning algorithms (such as support vector regression or random forest) to continuously optimize the heating model using the inputs (material type, moisture content, ambient temperature, initial compartment temperature, frozen area distribution) and outputs (thawing time, total steam consumption, thawing pass rate, valve opening curves for each section) of each thawing operation as training samples. When the thawing effect is stable after multiple consecutive thawing operations under the same conditions, a standard heating curve is automatically generated and updated to the database.

[0038] The collaborative working mechanism of this invention achieves deep integration of multi-source sensing information and adaptive control through a five-order closed-loop structure of "perception—experience—decision—execution—self-learning". The components are not simply isolated modules, but rather exist in a collaborative relationship of mutual dependence and support: (a) Synergistic sensing of infrared thermal imaging and near-infrared moisture meter Infrared thermal imagers can identify the spatial distribution of temperature on the exterior walls of the vehicle compartment, but they cannot determine the freezing characteristics of the materials themselves—at the same temperature, materials with high moisture content have significantly higher freezing hardness and are more difficult to thaw than materials with low moisture content; conversely, near-infrared moisture meters can detect the moisture content of materials, but cannot locate the spatial position of frozen areas. The organic integration of these two technologies allows the system to simultaneously obtain dual criteria for judgment: "where is frozen" (spatial information) and "how difficult is it to thaw" (material property information). This complementary and collaborative sensing overcomes the limitation that the two types of sensors cannot independently achieve precise temperature control when used alone, forming the data foundation for subsequent precise heating.

[0039] (II) Collaborative Optimization of RFID Historical Data and Self-Learning Model RFID readers are not only used for identification within the train carriages, but their core function is to provide historical training samples for the self-learning optimization module. Based on the historical unfreezing records of the carriage read by RFID (including input parameters and output results from multiple unfreezing operations), the self-learning optimization module uses machine learning algorithms to generate the initial settings for the current unfreezing operation (rather than directly calling a historical best value). This closed-loop model of "experience reuse + real-time optimization" enables the system to continuously evolve—after each operation, the newly generated data becomes training samples for model updates, and system performance improves with the number of runs.

[0040] (III) Coordination between Frozen Area Identification and Zoned Spraying The PLC precisely maps the coordinates of the frozen areas identified by the infrared thermal imager to each steam branch pipe section to determine the initial values ​​of the zonal valve openings, and dynamically adjusts them according to changes in the temperature cloud map during the thawing process. This mapping relationship, where "sensing data directly drives the execution layer," achieves precise targeting from "where the freezing is severe" to "where to supply heat," avoiding the energy waste of traditional uniform pipe layout.

[0041] (iv) Overall coordination of the fifth-order closed loop The aforementioned three layers of collaboration do not operate independently, but are integrated into an organic whole through a feedback loop formed by the "perception layer → experience layer → decision-making layer → execution layer → self-learning layer". Data from the perception layer provides real-time input to the decision-making layer, historical data from the experience layer provides the initial model, and the control effect of the execution layer is fed back to the model for updates through the self-learning layer. Removing any component will render the closed-loop system incomplete, and intelligent unfreezing control will be impossible.

[0042] The technical effects produced by the aforementioned synergistic relationship are superior to the simple sum of the effects of each individual technical feature working independently. Specifically: infrared thermal imaging, when used alone, can only provide temperature monitoring and cannot adjust the steam volume according to material characteristics; near-infrared moisture analyzer, when used alone, can only detect moisture content and cannot locate frozen areas; RFID, when used alone, can only identify the carriage and cannot achieve adaptive control; while zoned spraying, when used alone, can achieve zoned heating, without frozen area identification information, the zone opening degree can only rely on manual experience to set, making precise alignment impossible; the self-learning optimization module, without training data provided by RFID, cannot build a personalized model for the carriage. This invention achieves technical effects that cannot be achieved by each element working alone through the organic synergy of these five elements.

[0043] A method for intelligent thawing control of bulk materials on trains based on multi-source data fusion includes the following steps: S1: Identify the car number to be thawed using RFID reader 13, retrieve the historical thaw records of the car from the database, and generate the initial settings for this thaw based on the historical data using the self-learning optimization module. S2: Before or during the entry of the carriage into the insulation cover, the real-time moisture content of the material is detected by the near-infrared moisture meter 12; the current ambient temperature and humidity are collected by the ambient temperature and humidity sensor 14, and the reference steam flow rate is adjusted according to the moisture content correction rule. S3: Pull the car to be thawed into the insulation cover 7 and stop the car; seal both ends with the end sealing curtain 9. S4: The infrared thermal imager (10) collects temperature cloud maps of the outer wall of the carriage and the surface of the material. The PLC11 executes the frozen area identification algorithm to identify continuous areas with temperatures below 0℃ and calculates the coordinate range of each area on the outer wall of the carriage, maps it to each steam branch pipe section, and determines the initial value of the valve opening of each section. S5: The PLC adjusts the total steam flow and pulse injection parameters (continuous injection, pulse injection or reduced injection) according to the moisture content, ambient temperature and frozen area ratio, and outputs control signals to open the main steam regulating valve and the electric regulating valve of each section to start heating and defrosting. S6: During the thawing process, the infrared thermal imager continuously monitors the temperature field changes, and the PLC dynamically adjusts the valve opening of each section until the temperature of all areas of the outer wall of the carriage is higher than 2°C and maintained for more than 3 minutes. S7: Close the steam valve, raise or remove the insulation cover, and pull out the thawed car; at the same time, the self-learning optimization module records the data of the entire thawing process and updates the historical database and self-learning model of the car.

[0044] The pulse spraying parameters described in S5 are adaptively adjusted according to the degree of freezing: when the frozen area identified by infrared thermal imaging is greater than 40% of the outer wall area of ​​the carriage, a continuous spraying mode is adopted; when the frozen area is between 20% and 40%, a pulse spraying mode is adopted (spray for 3 minutes and stop for 1 minute); when the frozen area is less than 20%, a reduced spraying mode is adopted (valve opening is not greater than 40%).

[0045] In this embodiment, the method includes the following core steps: Step 1 (Start): The system powers on and performs a self-test, waiting for the carriage to enter; Step 2 (RFID Identification and Historical Data Retrieval): The RFID reader reads the electronic tag number of the carriage to be thawed, communicates with the factory production management system database, reads the historical thaw records of the carriage, and the self-learning optimization module generates the initial setting values ​​for this thaw. Step 3 (Moisture content and environmental parameter detection): Near-infrared moisture meter detects the moisture content of the material, ambient temperature and humidity sensor detects the ambient temperature and humidity, and PLC adjusts the reference steam flow rate according to the moisture content correction rule; Step 4 (Infrared thermal imaging frozen area identification): The infrared thermal imager collects temperature cloud images of the outer wall of the carriage, and the PLC executes the frozen area identification algorithm to identify severely frozen high-altitude and cold areas. Step 5 (Initial opening setting of zoned valves): The PLC maps the coordinate range of the frozen area to each steam branch pipe section, and generates the initial opening setting value of each electric regulating valve accordingly. Step 6 (Steam injection heating): The PLC outputs control signals to open the main steam regulating valve and the electric regulating valve of each section based on the total steam flow reference value after moisture content correction, the initial value of the zone opening degree and the pulse mode parameters, so as to start heating and defrosting. Step 7 (Determination of Thawing Completion): The infrared thermal imager continuously monitors the temperature field changes, and the PLC dynamically adjusts the valve opening of each section. When the temperature of all areas on the outer wall of the carriage is higher than 2°C and maintained for more than 3 minutes, the thawing is determined to be complete (Yes), and the process proceeds to Step 8; if the completion condition is not met (No), the process returns to Step 6 to continue heating. Step 8 (Record data for this cycle): The PLC transmits the data of the entire unfreezing process (input parameters, output results) to the self-learning optimization module; Step 9 (Updating and storing the self-learning model): The self-learning optimization module updates the self-learning model based on the new data and stores the updated model parameters in the historical database for retrieval in the next assignment; then it returns to wait for the next assignment.

[0046] Implementation Case 1 An intelligent steam defrosting system of this invention was constructed at the sintering yard of a steel company in northern China. A 30m long straight section of the train track was equipped with an insulation cover (15m long, sufficient to cover one C70 train car). Inside the insulation cover, the main steam pipe (with regulating valves) branches into branch pipes. These branch pipes are divided into four independent sections along their length, each approximately 3.5m long, and each is equipped with an electrically adjustable ball valve.

[0047] Multi-source sensing unit configuration: Infrared thermal imager: FLIR A70 model, installed in the center of the ceiling and on both sides of the insulation cover at a height of 2.5m from the rail surface, a total of 3 units, connected to the PLC via GigE interface.

[0048] Near-infrared moisture meter: installed above the material conveyor belt, with a measurement wavelength range of 1400-2200nm and a measurement accuracy of ±0.2%.

[0049] RFID reader: Installed 20m in front of the entrance of the insulation cover next to the track, with a reading and writing distance of 0-2m.

[0050] Ambient temperature and humidity sensor: installed on the outer column of the insulation cover.

[0051] Intelligent control unit: The PLC adopts Siemens S7-1500 and is equipped with an image processing module; the self-learning optimization module adopts Advantech industrial computer, deploys a model based on random forest algorithm, and communicates with the PLC via Ethernet.

[0052] A typical operation process fully demonstrates the fifth-order closed-loop cooperative mechanism of this invention: S1: RFID Identification and Historical Data Retrieval (Empirical Level) A train loaded with iron ore concentrate enters the factory area. An RFID reader identifies the carriage number "C70-1234" and retrieves the complete records of the previous five thawing operations for that carriage from the database (including ambient temperature, moisture content, valve opening curves for each section, thawing time, total steam volume, and thawing pass rate). Based on this historical data, the self-learning optimization module uses a random forest model to predict the initial settings for this thawing operation: total steam volume 1.2 tons, heating time 22 minutes, and valve opening baseline values ​​for each section: Section 1 (front end) 65%, Section 2 80%, Section 3 80%, and Section 4 60%.

[0053] This step demonstrates the synergy between RFID and the self-learning model: historical thawing records are used as training samples to predict the initial settings for this time, rather than being set manually based on experience.

[0054] S2: Moisture content and environmental parameter detection (sensing layer) The near-infrared moisture analyzer detected a material moisture content of 9.3% in real time, which is close to the historical average of 9.5%. Based on the moisture content correction rule (±1% moisture content corresponds to ±6% steam flow), the total steam flow rate baseline was adjusted to 1.19 tons. The ambient temperature and humidity sensor measured an ambient temperature of -12℃ and a relative humidity of 65%.

[0055] This step demonstrates the synergy between the near-infrared moisture analyzer and the PLC: the moisture content detection results are directly used to correct the total steam flow rate reference value, achieving adaptive adjustment of material characteristics.

[0056] S3: The carriage is enclosed by an insulation cover and sealed. Once the thawed carriage is towed into the insulation cover and stopped, the end sealing curtains automatically close both ends.

[0057] S4: Identification of frozen areas by infrared thermal imaging and determination of initial opening degree of partitions (perception layer → decision layer) An infrared thermal imager acquires temperature cloud images of the outer wall of the carriage. The PLC executes a frozen area identification algorithm: the temperature cloud image is converted to grayscale, thresholded (threshold set to 0℃), continuous areas with temperatures below 0℃ are extracted, the coordinate range of each area on the outer wall of the carriage is calculated, and mapped to each steam branch pipe section. The identification results show that there are large frozen areas (temperatures from -8℃ to -3℃) in the rear of the carriage (sections 3 and 4), while the temperature in the front (sections 1 and 2) is relatively higher (-2℃ to 2℃). Based on the proportion of frozen area (65% in section 3 and 58% in section 4), the PLC automatically increases the initial valve opening of sections 3 and 4 from the baseline value of 80% to 85%, keeps section 2 at 70%, and keeps section 1 at 50%.

[0058] This step demonstrates the synergy between infrared thermal imaging and zoned injection: the coordinates of the frozen area are precisely mapped to each steam branch pipe section, enabling precise spatial control of on-demand heating.

[0059] S5: Initiate steam injection heating (Decision-making level → Execution level) The PLC calculates that the total frozen area accounts for 42% of the total area of ​​the outer wall of the carriage, which is greater than 40%. Therefore, the continuous injection mode (non-pulse) is automatically selected. The main steam regulating valve and the electric regulating valves of each section are opened, and heating begins at the set opening degree.

[0060] S6: Real-time temperature field feedback and dynamic adjustment (execution layer, feedback, self-learning layer) During the thawing process, the infrared thermal imager updates the temperature cloud map every 2 minutes. When the outer wall temperature of sections 3 and 4 rises above 0°C, the PLC gradually reduces its valve opening to 60%; when the outer wall temperature of the entire line is above 2°C and remains so for 3 minutes, the PLC determines that thawing is complete.

[0061] S7: Stop heating and record data (self-learning layer closed loop) Close the steam valves, raise the insulation cover, and pull out the thawed carriages. The total thawing time was 20 minutes, the total steam consumption was 1.15 tons, and the thawing pass rate was 100%. The self-learning optimization module recorded the entire process data of this operation (input parameters, output results), updated the historical records of "C70-1234" in the database, and retrained the random forest model. Model prediction: Under the same working conditions next time, the initial opening degree of sections 3 and 4 can be set to 82%, and the thawing time is expected to be shortened to 19 minutes.

[0062] This step completes the self-learning feedback of the fifth-order closed loop: the data from this assignment becomes the training sample for updating the self-learning model, enabling the system performance to continuously improve with the number of runs.

[0063] After a winter of continuous operation, the system thawed over 800 carriages, reducing the average thawing time from 25 minutes to 17 minutes, and the average steam consumption from 1.4 tons / carriage to 1.05 tons / carriage. The thawing pass rate was 99.5%, with no unloading failures due to incomplete thawing. There was no water accumulation or ice formation in the tracks or insulation covers (condensate was collected and recycled to the boiler room). The number of operators was reduced from 3 per shift to 1 for inspection.

[0064] Implementation Case 2 For another material used by the company: red mud powder (with a large fluctuation in moisture content, measured in the range of 12%-15%), the moisture content adaptive function was activated. When the near-infrared moisture meter detected a moisture content of 13.8%, the PLC automatically increased the total steam flow rate benchmark by 18% (calculated based on an increase of 6% steam volume for every 1% increase in moisture content). Simultaneously, due to the extremely high hardness of the red mud powder after freezing, the infrared thermal imaging identified a frozen area covering 65% of the surface. The system automatically selected the continuous spraying mode and, based on historical data from the self-learning model for "high moisture content red mud powder working conditions," preset the heating time to 35 minutes. The final thawing was successful, with no overheating. The self-learning module stored the parameters for this operation (moisture content 13.8%, frozen area 65%, continuous spraying, actual thawing time 34 minutes, steam consumption 1.58 tons) in an independent model, which can be directly referenced for subsequent similar working conditions, gradually stabilizing the thawing time to 30-33 minutes.

[0065] Abnormal operating condition handling When infrared thermal imaging detects a local temperature exceeding 70°C, the PLC determines it as local overheating, automatically shuts off the electric regulating valve in the corresponding section, and issues an alarm. When the steam pipeline pressure drops below 0.2 MPa, heating is automatically stopped, and a prompt to check the steam source is issued. The above safety interlock functions were verified in both Examples 1 and 2, and no safety incidents occurred.

[0066] This invention upgrades traditional steam thawing into a closed-loop intelligent system of "sensing-decision-execution-learning" by introducing infrared thermal imaging, near-infrared online moisture detection, RFID historical data mining, and machine learning self-optimization. This effectively solves the technical problems of existing mechanical and conventional temperature-controlled steam thawing systems, which cannot provide heat on demand, adapt to material changes, or self-evolve. The system has a clear structure, is easy to implement, and can significantly reduce the thawing costs of railway transportation in winter for steel enterprises, improve unloading efficiency, and has broad market promotion value.

[0067] Abnormal operating condition handling: When the infrared thermal imaging detects that the local temperature exceeds 70℃, the PLC determines that it is a local overheating, automatically closes the electric regulating valve of the corresponding section and issues an alarm; when the steam pipeline pressure is lower than 0.2MPa, it automatically stops heating and prompts to check the steam source.

[0068] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A train bulk material intelligent thawing control system based on multi-source data fusion, characterized in that, include: Insulating covers are used to contain train carriages that are about to thaw. The steam supply and regulation unit includes a main steam pipe and several independently controlled steam branch pipes. The steam branch pipes are divided into multiple independent injection sections along the length of the carriage inside the insulation cover, and each section is equipped with an electric regulating valve. Multi-source sensing units include an infrared thermal imager, a near-infrared moisture meter, an RFID reader, and an ambient temperature and humidity sensor; Intelligent control unit, including PLC and self-learning optimization module; The PLC is used for: Receive temperature cloud image data from an infrared thermal imager, identify frozen areas and map them to the corresponding steam branch pipe sections; Receive moisture content data from a near-infrared moisture meter and adjust the reference value of total steam flow rate according to the moisture content correction coefficient; Receive historical unfreezing records read by the RFID reader and combine them with the initial settings generated by the self-learning optimization module; Output control signals to the electric regulating valves of each steam branch section.

2. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 1, characterized in that, The self-learning optimization module uses support vector regression or random forest algorithm to optimize the shortest thawing time and the lowest steam consumption. When the energy consumption and time deviation of multiple consecutive thawing operations of the same carriage type, approximate moisture content and ambient temperature are all less than the set threshold, a new standard heating curve is automatically generated and stored in the database.

3. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 1, characterized in that, The PLC has a built-in frozen area recognition algorithm that performs grayscale and threshold segmentation on the temperature cloud map collected by the infrared thermal imager to identify continuous areas with temperatures below freezing. Based on the coordinate system calibration of the outer wall of the carriage, the coordinate range of each area is mapped to the corresponding steam branch pipe section.

4. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 1, characterized in that, The PLC adjusts the steam flow rate according to the real-time moisture content and the following rules: for every 1% increase in moisture content, the total steam flow rate increases by 5% to 8%; for every 1% decrease in moisture content, the total steam flow rate decreases by 5% to 8%.

5. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 1, characterized in that, The steam branch pipe is divided into 4-6 independent sections along the length of the carriage, each section being 1.5m to 3.8m long; each section of the branch pipe is provided with injection holes facing the side wall and bottom of the carriage, with a hole diameter of 3mm to 6mm and a hole spacing of 80mm to 150mm.

6. A method for intelligent thawing control of bulk materials on trains based on multi-source data fusion, characterized in that, include: S1. Identify the car number to be thawed using an RFID reader, and retrieve the input parameters and output results of the car in historical thaw operations; at the same time, use a near-infrared moisture meter to detect the real-time moisture content of the materials in the car online. S2. Based on historical thawing records, the initial settings for this thawing are generated using a self-learning optimization module. The initial settings include the total steam volume, valve opening degree of each section, and pulse mode. The baseline value of the total steam volume is then corrected based on the real-time moisture content. S3. Before the thawing begins, use an infrared thermal imager to collect temperature cloud maps of the outer wall of the carriage and the surface of the materials. Use a frozen area identification algorithm to determine continuous areas with temperatures below freezing point and map the coordinate range of the frozen area to the corresponding steam branch pipe section inside the insulation cover. S4. Adjust the opening of the electric regulating valve of the corresponding steam branch section according to the mapping result, and combine it with pulse mode to control steam injection to achieve targeted heating for severely frozen areas. S5. During the thawing process, continuously monitor the temperature field changes and dynamically adjust the valve opening of each section; after thawing is completed, record the data of the entire operation and feed it back to the self-learning optimization module to update the thawing parameter model.

7. The intelligent thawing control method for bulk cargo on trains based on multi-source data fusion according to claim 6, characterized in that, The baseline value of total steam quantity is corrected based on the real-time moisture content as follows: Set a moisture content correction factor. When the real-time moisture content increases by 1%, the benchmark value of total steam flow increases by 5% to 8%; when the real-time moisture content decreases by 1%, the benchmark value of total steam flow decreases by 5% to 8%.

8. The intelligent thawing control method for bulk cargo on trains based on multi-source data fusion according to claim 6, characterized in that, After acquiring temperature cloud images using an infrared thermal imager, the following image processing steps are included: The temperature cloud map was converted to grayscale and segmented with 0℃ as the threshold to extract the low-temperature region; based on the coordinate system calibration of the outer wall of the carriage, the coordinate range of the low-temperature region in the length direction was calculated.

9. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 6, characterized in that, The specific steps for controlling steam injection using pulse mode are as follows: Calculate the proportion of the frozen area to the total area of ​​the outer wall of the carriage; When the ratio is greater than 40%, continuous spray mode is used; When the ratio is between 20% and 40%, pulse jet mode is used, with a jet cycle of 3 minutes of spraying and 1 minute of stopping. When the ratio is less than 20%, a reduced injection mode is adopted, limiting the maximum valve opening to no more than 40%.

10. The intelligent thawing control system for bulk cargo on trains based on multi-source data fusion according to claim 6, characterized in that, The conditions for successful thawing are: infrared thermal imaging shows that the temperature of all areas on the outer wall of the carriage is above 2°C and remains above 2°C for more than 3 minutes.

Images