Railway tank car intelligent checking system and method

Abstract

The invention relates to the technical field of railway transportation oil loading and unloading automation, and provides a railway tank car intelligent checking system and method.The method comprises the steps that after incoming oil related data is obtained, in-place state information of a tank car is obtained through recognition equipment, and a car number image of the tank car is collected according to image recognition equipment and converted into character information; comparing the character information with the car number information of the tank car of the corresponding batch in a refinery incoming oil information if the comparison is passed, starting metering data acquisition, and obtaining the metering data of the oil product in the tank car through a portable measuring instrument; performing comparative analysis on the acquired metering data, the theoretical data of the refinery incoming oil information base and the historical data of the railway tank car metering database by using a preset intelligent algorithm: if the comparison is passed, performing verification according to the car number, the in-place state and the metering data, and generating an automatic oil unloading instruction after the verification is passed; and the railway oil unloading automatic control system receives the automatic oil unloading instruction, and intelligent checking of the railway tank car is completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic loading and unloading of railway oil transportation, and in particular to an intelligent checking system and method for railway tank cars. BACKGROUND

[0002] In the loading and unloading operation of railway oil transportation in the petrochemical industry, the checking process of railway tank cars usually includes key links such as vehicle number checking, in-position state confirmation, metering data verification, and unloading operation control. In the traditional technology, the above processes mainly rely on manual operation and experience judgment: for example, the vehicle number is checked by manual visual inspection, which is easy to cause recognition errors due to factors such as light and angle; the in-position state confirmation relies on the on-site inspection of the relative position of the tank car and the loading and unloading port by the operator, which has subjective judgment errors; the metering data verification only collects data by a portable device once and compares it with the theoretical value simply, without trend analysis combined with historical data, so it is difficult to find equipment abnormalities or data fluctuation hidden dangers; the start and end state confirmation of the unloading operation relies on manual inspection, which has risks such as untimely response and safety interlock failure. In addition, the existing technology is independent in each link, lacks systematic intelligent integration, resulting in low checking efficiency, high human intervention cost, poor data traceability, and cannot meet the operation needs of modern refineries in automation and intelligence.

[0003] Although there are some improvement schemes for single links in the prior art (such as a separate vehicle number identification system or a metering data acquisition device), there is no technical scheme to organically combine image recognition, multi-source data fusion (refinery oil information library, railway tank car metering database), intelligent algorithm comparison and automatic control process, to form a complete intelligent checking system covering “in-position state recognition-vehicle number verification-intelligent analysis of metering data-unloading instruction generation-operation whole-process control”. In particular, the prior art does not involve multi-dimensional comparison and analysis of real-time metering data, theoretical data and historical data by preset intelligent algorithms, and does not disclose a multi-level verification mechanism and automatic instruction generation logic based on vehicle number, in-position state and metering data. Therefore, an innovative scheme is needed to improve the checking accuracy, automation level and operation safety. SUMMARY

[0004] The present application provides an intelligent checking system and method for railway tank cars, aiming to solve the problem that the prior art does not involve multi-dimensional comparison and analysis of real-time metering data, theoretical data and historical data by preset intelligent algorithms, and does not disclose a multi-level verification mechanism and automatic instruction generation logic based on vehicle number, in-position state and metering data.

[0005] In a first aspect, the present application provides an intelligent checking method for railway tank cars, comprising: After obtaining the oil-related data, the in-position state information of the tank car is obtained through the identification device, the vehicle number image of the tank car is collected by the image recognition device and converted into character information; The character information is compared with the tank car vehicle number information of the corresponding batch in the refinery oil information library: if the comparison is passed, the measurement data collection is started, and the tank car oil measurement data is obtained through the portable measuring instrument; The collected measurement data is compared and analyzed with the theoretical data of the refinery oil information library and the historical data of the railway tank car measurement database by using a preset intelligent algorithm: if the comparison is passed, the vehicle number, the in-position state and the measurement data are verified, and the automatic oil unloading instruction is generated after the verification is passed; The railway oil unloading automation control system receives the automatic oil unloading instruction, performs the unloading start state confirmation, starts the unloading operation after confirming that the device state is normal, performs the end state confirmation after the unloading is completed, generates the oil receiving and sending certificate information report, and completes the railway tank car intelligent verification.

[0006] In a second aspect, the present application provides a railway tank car intelligent verification system, comprising: A data acquisition unit is configured to obtain oil-related data, obtain in-position state information of a tank car through an identification device, and collect a vehicle number image of the tank car by an image recognition device and convert the vehicle number image into character information; An information comparison unit is configured to compare the character information with tank car vehicle number information of a corresponding batch in a refinery oil information library: if the comparison is passed, start measurement data collection, and obtain tank car oil measurement data through a portable measuring instrument; An instruction generation unit is configured to compare and analyze collected measurement data with theoretical data of a refinery oil information library and historical data of a railway tank car measurement database by using a preset intelligent algorithm: if the comparison is passed, verify the vehicle number, the in-position state and the measurement data, and generate an automatic oil unloading instruction after the verification is passed; An intelligent verification unit is configured to receive an automatic oil unloading instruction by a railway oil unloading automation control system, perform an unloading start state confirmation, start an unloading operation after confirming that a device state is normal, perform an end state confirmation after the unloading is completed, generate an oil receiving and sending certificate information report, and complete railway tank car intelligent verification.

[0007] In a third aspect, the present application further provides a computer device, comprising: A memory and a processor; The memory is configured to store a computer program; The processor is configured to execute the computer program and implement the steps of the railway tank car intelligent verification method of the first aspect when the computer program is executed.

[0008] In a fourth aspect, the present application also provides a computer readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement the steps of the intelligent checking method for railway tank cars according to the first aspect.

[0009] The railway tank car intelligent checking system and method provided by the embodiments of the present application automatically obtains the tank car positioning state and car number information through a recognition device, replaces manual visual inspection and manual input, reduces human operation errors, and improves work efficiency; compares and analyzes real-time measurement data, theoretical data of a refinery, and historical operation data through a preset intelligent algorithm, realizes automatic screening of abnormal data in combination with a logical verification rule, and improves the data verification accuracy compared with traditional single data comparison; multiple confirmations of the device state are performed before oil unloading operation, the state change is monitored in real time during oil unloading, and an unforgeable report record is automatically generated after the end, thereby constructing a safety protection system covering the whole operation cycle, reducing the risks of running, leaking, dripping, and electrostatic hazards caused by human negligence, and meeting the needs of digital management and compliance audit of a refining enterprise.

[0010] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present application. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0012] Figure 1 is a step schematic flowchart of a railway tank car intelligent checking method provided by an embodiment of the present application; Figure 2 is a logic schematic diagram of a railway tank car intelligent checking method provided by an embodiment of the present application; Figure 3 is a structure schematic diagram of a railway tank car intelligent checking system provided by an embodiment of the present application; Figure 4 is a structure schematic block diagram of a computer device provided by an embodiment of the present application.

[0013] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present application. DETAILED DESCRIPTION

[0014] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of the present application.

[0015] The flowcharts shown in the drawings are only illustrative, and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the described order. For example, some operations / steps can be further decomposed, combined or partially merged, so that the actual execution order can be changed according to actual conditions.

[0016] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, the terms "first", "second", etc. are used to distinguish the same or similar items with basically the same functions and effects. Those skilled in the art can understand that the terms "first", "second", etc. do not limit the quantity and execution order, and the terms "first", "second", etc. also do not necessarily mean different.

[0017] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clearly indicated by the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0018] It should also be understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0019] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0020] In the oil and chemical industry, the checking process of railway tank car usually includes the following key links: checking the car number, confirming the in-place state, checking the measurement data, and controlling the unloading operation. In the traditional technology, the above processes mainly rely on manual operation and experience judgment: for example, the car number is checked by manual visual inspection, which is easy to cause recognition errors due to factors such as light and angle; the in-place state confirmation relies on the on-site inspection of the relative position of the tank car and the loading and unloading port by the operator, which has subjective judgment errors; the measurement data checking only collects data by a portable device at a time and compares it with the theoretical value, without trend analysis combined with historical data, so it is difficult to find equipment abnormalities or data fluctuation risks; the start and end state confirmation of the unloading operation relies on manual inspection, which has the risk of delayed response and safety interlock failure. In addition, the existing technology lacks systematic intelligent integration of each link, resulting in low checking efficiency, high human intervention cost, poor data traceability, and inability to meet the modern refinery automation and intelligent operation requirements.

[0021] Although there are some improvement schemes for a single link in the prior art (such as a separate car number identification system or a measurement data acquisition device), there is no technical solution to organically combine image recognition, multi-source data fusion (refinery oil information library, railway tank car measurement database), intelligent algorithm comparison, and automatic control process to form a complete intelligent checking system covering "in-place state recognition-car number checking-intelligent analysis of measurement data-unloading instruction generation-operation whole process control". In particular, the prior art does not involve multi-dimensional comparison and analysis of real-time measurement data, theoretical data, and historical data by a preset intelligent algorithm, and does not disclose a multi-level verification mechanism and automatic instruction generation logic based on the car number, in-place state, and measurement data. Therefore, there is an urgent need for an innovative solution that can improve the checking accuracy, automation level, and operation safety.

[0022] To solve the above problems, please refer to Figure 1 , Figure 1 is a schematic flowchart of the railway tank car intelligent checking method provided by an embodiment of the present application. The railway tank car intelligent checking method can be implemented by a computer device, which can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0023] It should be noted that the acquisition of any information involved in the provided method is in compliance with relevant regulations and with the consent of the user, and does not infringe on the user's privacy or violate relevant laws and regulations.

[0024] Specifically, as shown in Figure 1 , the provided railway tank car intelligent checking method includes steps S101 to S104, which are described in detail as follows: Step S101. After obtaining the oil-related data, the in-position state information of the tank truck is obtained through the identification device. The vehicle number image of the tank truck is collected by the image recognition device and converted into character information.

[0025] Specifically, the in-position state and vehicle number information of the tank truck are collected by the multi-modal device to provide basic data for subsequent verification. Specifically, it includes: oil-related data acquisition: obtaining batch information in the refinery oil plan (such as oil type, planned vehicle number, expected weight / volume, loading / unloading port number, etc.), tank truck basic information (vehicle type, maximum load, historical measurement data index, etc.). In-position state information collection: identify whether the tank truck is accurately parked at the designated loading and unloading position, including position alignment (lateral / longitudinal deviation of the tank truck and the loading / unloading port), connection state (whether the loading / unloading arm / hose is aligned with the interface), safety state (whether the anti-slip device is in place, whether the static grounding is connected), etc. Vehicle number image recognition and character conversion: collect the vehicle number area image through the image recognition device, and convert it into machine-readable character information using optical character recognition (OCR) technology.

[0026] The data acquisition interface connects the refinery ERP system or the oil scheduling platform through API to synchronize the vehicle number list, loading / unloading port allocation, and theoretical measurement value of the oil batch in real time.

[0027] In-position state detection includes: hardware deployment: install laser ranging sensors (detect lateral / longitudinal position deviation), industrial cameras (take pictures of the alignment state of the loading / unloading port and the tank truck interface), pressure sensors (detect whether the loading / unloading arm connection is in place), and wireless sensors (monitor the anti-slip device and static grounding state) at the loading / unloading station. Data fusion: input multiple sensor data into the edge computing module, judge the in-position state (such as deviation <5 cm, loading / unloading arm pressure > threshold, static grounding signal normal) through pre-set rules, output "in-position qualified" or "in-position abnormal" state.

[0028] Vehicle number recognition includes: image collection: install high-definition cameras (with fill light, suitable for different lighting conditions) above the tank truck travel route, trigger the camera to take pictures after the tank truck is stationary, collect vehicle number area images (support side / top vehicle number recognition, cover different angles). Character conversion: use deep learning OCR model (such as CRNN+CTC architecture), identify characters after image preprocessing (denoising, perspective correction), output vehicle number string (such as "Tank A-12345"), and label confidence (such as ≥95% considered as valid recognition).

[0029] Step S102. Compare the character information with the tank truck vehicle number information of the corresponding batch in the refinery oil information library: if the comparison is successful, start measurement data collection, and obtain the tank truck oil measurement data through the portable measuring instrument.

[0030] Specifically, the identified car number is matched with the planned car number in the refinery incoming oil information library to verify the legality of the tank car identity. If the comparison passes, the metering data acquisition process is triggered to obtain real-time oil metering parameters.

[0031] The character information generated in step S101 is accurately matched with the car number list corresponding to the incoming oil batch, fuzzy matching (such as allowing case differences and space differences) and fault tolerance mechanisms (such as triggering manual review for three consecutive recognition failures) are supported.

[0032] Metering data acquisition is triggered only when the car number comparison passes, and portable measuring instruments (such as smart liquid level meters, temperature sensors, and density meters) are remotely activated to collect real-time metering data (liquid level, temperature, density, volume / weight, etc.) of the oil.

[0033] The car number comparison logic includes: data matching: accurately match the car number list (such as containing 10 car numbers) of the current batch in the incoming oil information library with the recognition result, supporting "one car one comparison" (checking each car number) or "batch full matching" (checking all arriving car numbers). Abnormal processing: if the recognized car number is not in the planned list (such as confidence ≥ 95% but no matching record), the system automatically marks "car number abnormal", triggers sound and light alarm and pushes manual review instruction (such as on-site operator confirms car number through handheld terminal).

[0034] Metering device linkage includes: device integration: portable measuring instruments are connected to on-site industrial computers through Bluetooth / Wi-Fi, and the industrial computers send instructions to activate device acquisition functions (such as automatically lowering the liquid level meter probe into the tank car) after receiving the car number comparison pass signal. Data acquisition items: at least collect oil temperature (for volume conversion), liquid level (to calculate current volume), density (to convert standard volume), oil temperature and tank wall temperature difference (to determine whether there is stratification), and data accuracy needs to meet industry standards (such as temperature ±0.5℃, liquid level ±1mm).

[0035] Step S103. The collected metering data is compared and analyzed with the theoretical data of the refinery incoming oil information library and the historical data of the railway tank car metering database using a pre-set intelligent algorithm: if the comparison passes, the car number, entry state and metering data are verified, and an automatic unloading instruction is generated after verification.

[0036] Specifically, through a pre-set intelligent algorithm, real-time metering data, refinery theoretical data (planned value), and tank car historical data (same model / historical metering records of the same car number) are analyzed in multiple dimensions to build a multi-level verification mechanism. If the verification passes, an automatic unloading instruction is generated.

[0037] Multi-source data fusion analysis includes: theoretical data comparison: comparison of real-time measurement value with theoretical value (such as planned volume, density range) in oil information library, calculation of absolute deviation and relative deviation (such as volume deviation rate ≤ ± 2%). Historical data trend analysis: call the measurement data of the car number in the railway tank car measurement database for the last 3 times of transportation, analyze the fluctuation trend of density and volume (such as continuous two times of density drop > 5% to trigger equipment abnormal early warning).

[0038] Intelligent algorithm verification uses statistical model (such as 3σ principle to identify abnormal value) or machine learning algorithm (such as isolated forest to detect data abnormal point), and comprehensively judges whether the measurement data is reliable.

[0039] Multi-level verification mechanism includes: car number legitimacy (step S102 passes), in-place state qualification (step S101 determines), and measurement data verification (analysis results of this step) must be met at the same time, and then the oil unloading instruction can be generated.

[0040] Data comparison rules include: theoretical value verification: the calculation formula is deviation rate = (real-time value-theoretical value) / theoretical value × 100%, for example, volume deviation rate ≤ ± 2%, density deviation rate ≤ ± 1% is considered qualified, and if it exceeds the range, it is marked as "measurement abnormality". Historical trend analysis: sliding average calculation is performed on the density data of the same tank car for the last 3 times of transportation, if the deviation of real-time density and historical average is > 3σ (standard deviation), and it appears continuously for two times, it is judged as "equipment may be malfunctioning" (such as sensor drift).

[0041] Intelligent algorithm implementation includes: basic statistical model: preset reasonable range of density and temperature of each oil type (such as gasoline density 700-780 kg / m 3 ), and real-time data exceeding the range directly triggers early warning.

[0042] Machine learning model: based on historical normal operation data to train abnormal detection model (such as One-Class SVM), real-time input measurement data vector (volume, density, temperature, oil temperature difference), output "normal" or "abnormal" classification result.

[0043] Instruction generation logic includes: only when the following conditions are met at the same time, the oil unloading instruction is generated: car number comparison state = "pass"; in-place state = "qualified"; measurement data verification result = "pass" (theoretical deviation is within the threshold and there is no abnormality in historical trend).

[0044] Step S104. The railway oil unloading automation control system receives the automatic oil unloading instruction, performs the unloading start state confirmation, starts the unloading operation after confirming that the equipment state is normal, performs the end state confirmation after the unloading is completed, generates the oil receiving and issuing certificate information report, and completes the railway tank car intelligent checking.

[0045] Specifically, the oil unloading instruction is sent to the automation control system to realize unmanned starting and ending confirmation of oil unloading operation, and a full-process data report is generated to ensure traceability of the operation.

[0046] The oil unloading starting confirmation is achieved by verifying whether the oil unloading pipeline valve state and safety interlocking devices (such as emergency shut-off valve, combustible gas alarm) are normal, to prevent the risk of starting under pressure or leakage.

[0047] The operation process monitoring is achieved by collecting oil unloading flow and pressure data in real time to determine whether there is an abnormal interruption (such as flow suddenly returning to zero for more than 30 seconds) or excessive unloading (cumulative amount exceeding 110% of the theoretical value).

[0048] The ending state confirmation and report generation are achieved by re-verifying the remaining amount of the tank car after oil unloading is completed (to prevent incomplete unloading), generating an electronic report containing the car number, time, measurement data, and equipment state, and interfacing with the refinery ERP system for archiving.

[0049] The automation control integration includes: hardware interface: connecting oil unloading valve group, flow meter, pressure sensor through Modbus / TCP protocol, instructions include "open valve X" "start pump Y" specific operations, supporting double confirmation mechanism (system instruction + field device feedback signal). Safety interlocking: must confirm that static grounding signal is normal, anti-slip device is locked, and combustible gas concentration is less than 20% of the lower explosion limit before starting, otherwise the instruction is refused and an alarm is given.

[0050] Process monitoring and abnormal handling include: real-time data acquisition: collect oil unloading flow and pipeline pressure at a frequency of 1 second per second, display the curve in real time through the SCADA system, set the flow fluctuation threshold (such as instantaneous flow mutation > 20% to trigger an early warning). Abnormal response: if a leakage alarm (combustible gas concentration > threshold) or pressure anomaly occurs, the system automatically triggers the emergency shut-off valve to stop the operation and pushes the fault code to the operation and maintenance platform.

[0051] Report generation and traceability include: electronic certificate generation: includes "Railway Tank Car Acceptance Sheet" and "Oil Product Receiving and Issuing Measurement Sheet", automatically fills in car number, operation time (start / end), measurement data (original issue quantity, actual received quantity, loss rate), and equipment number (such as sensor ID participating in measurement). Data archiving is achieved by encrypting the report and process data (images, sensor logs, instruction records) into a blockchain database or distributed file system, supporting quick retrieval by car number, time, and batch, meeting the audit requirements.

[0052] In some embodiments, the entry state information of the tank truck is obtained by identifying the device, including: deploying laser ranging sensors, pressure sensors and visual recognition devices at the preset position of the track in the railway unloading operation area, collecting the distance data of the tank truck and the loading and unloading port by the laser ranging sensor, collecting the pressure distribution data of the wheel and the track by the pressure sensor, and collecting the relative position image of the coupler and the track identification line by the visual recognition device; inputting the distance data, pressure distribution data and position image into the edge computing unit, and performing logical judgment based on the preset entry rule library, if the distance data is in the safe operation interval, the pressure distribution data meets the track bearing balance condition, and the position image deviation of the coupler and the identification line is less than the preset threshold, it is determined that the entry state of the tank truck is ready.

[0053] Through the fusion of multi-source data of laser ranging, pressure sensing and visual recognition, it is accurately judged whether the tank car is correctly positioned. The core includes: multi-dimensional data acquisition: use laser ranging sensor to obtain the spatial distance between tank car and loading and unloading port, pressure sensor to monitor the pressure distribution balance of wheel and track, visual recognition device to capture the relative position of coupler and track identification line, and build a three-dimensional space positioning model. Edge computing logical judgment: input three types of data into the edge computing unit, and perform logical verification based on the preset rule library (safe distance, pressure balance threshold, position deviation threshold), to ensure that the tank car parking position meets the safety and alignment accuracy of loading and unloading operation.

[0054] Hardware deployment includes: laser ranging sensor: 2 groups (1 group horizontally / vertically) are installed on both sides of the loading and unloading port, the horizontal sensor is perpendicular to the track, and the horizontal distance between the side of the tank car and the loading and unloading port is measured (accuracy ±2mm); the longitudinal sensor is parallel to the track, and the longitudinal distance between the end of the tank car and the loading and unloading port is measured (accuracy ±5mm). Pressure sensor: 4 groups of distributed pressure sensors are embedded under the track sleeper (2 groups for each side of the wheel), which can collect real-time single-wheel pressure value and pressure difference on both sides, and judge whether the wheel is completely stopped in the specified bearing area (pressure difference exceeding 10% triggers an exception). Visual recognition device: an industrial camera is installed in front of the track to capture the relative position of the coupler and the ground identification line (yellow positioning line), and the horizontal offset is calculated by image recognition algorithm (identification line deviation ≤3cm is qualified).

[0055] Edge computing processing includes: rule library configuration: the safe operation interval is set to 10-15cm horizontally and 5-10cm longitudinally; the pressure balance condition is that the pressure difference of the wheels on both sides is ≤5%; and the coupler deviation threshold is ≤3cm. Logical judgment process: the edge computing unit receives three types of data in real time, and only when the distance data is in the safe interval, the pressure difference is ≤5% and the coupler deviation is ≤3cm, the "entry ready" signal is output; if any condition is not met, it is marked as "entry deviation", and the specific fault type (such as "longitudinal distance out of limit") is displayed on the LED screen.

[0056] In some embodiments, the image recognition device collects the tank truck number image and converts it into character information, comprising: deploying multi-angle high-definition cameras on both sides of the oil unloading operation area, dynamically scanning the tank truck number area on the side and end face, and obtaining multiple frames of continuous images; performing noise reduction, distortion correction and image enhancement preprocessing on the multiple frames of images, using a convolutional neural network character recognition model to segment and recognize the number characters of the preprocessed images, and generating initial character information; performing string similarity matching on the initial character information and the candidate truck numbers of the same batch transportation plan in the refinery oil information library, and when the matching degree exceeds the preset threshold, outputting the final confirmed truck number character information.

[0057] Through multi-angle dynamic scanning, image preprocessing and intelligent matching algorithm, the accuracy and robustness of the truck number recognition are improved. The core includes: Multi-angle image acquisition: deploy high-definition cameras on both sides to cover the truck number area on the side and end face, solving the single view shielding problem. Multi-level recognition verification: first segment and recognize the characters through convolutional neural network (CNN), and then perform string similarity matching with the candidate truck numbers, double verification to ensure the reliability of the recognition result.

[0058] Image acquisition and preprocessing includes: hardware deployment: install high-definition cameras with a pan-tilt head (frame rate 30 fps, resolution 2048x1536) at a height of 3m on both sides of the oil unloading operation area, support automatic tracking of tank truck movement, dynamically scan the truck number area (side "tank truck number", end face "truck head number"). Preprocessing process: median filter noise reduction on 10 frames of continuous images, distortion correction through perspective transformation (solve character distortion caused by shooting angle), then use histogram equalization to enhance contrast, output standardized character area image.

[0059] Character recognition and matching includes: CNN model application: use an improved ResNet+CTC architecture model, after inputting the preprocessed image, automatically segment individual characters (support mixed recognition of Chinese characters, letters and numbers), output character sequence and confidence (such as "Tank A-1234", confidence 98%). Similarity matching: calculate the edit distance between the candidate truck numbers in the oil information library (such as "Tank A-1234" "Tank B-5678") and the initial recognition result, and if the matching degree is ≥95% (i.e. at most 1 character difference), it is confirmed as a valid truck number; if the matching degree is <95%, automatically trigger adjacent camera retake, merge multiple view recognition results and re-verify.

[0060] In some embodiments, the start of the metering data collection, the acquisition of the tank truck oil metering data by the portable measuring instrument, comprises: the portable measuring instrument establishes a wireless communication connection with the liquid level meter, temperature sensor and pressure sensor of the tank truck, automatically synchronizes the clock and checks the equipment calibration time; sequentially collects the oil level height, temperature, density and volume data, and takes the mean value after removing the abnormal fluctuation value as the effective metering data; the collection time, equipment number and data content of the metering data are stored by the blockchain technology, and the non-tamperable metering data record is generated.

[0061] The measurement equipment is linked through wireless communication, multi-dimensional oil data is collected, and the blockchain technology is used to ensure that the data cannot be tampered with. The core includes: device interconnection and data calibration: the portable measuring instrument is wirelessly connected with the sensors built-in the tank truck, the clock is synchronized and the equipment calibration state is checked to ensure that the data timestamp and accuracy are reliable. Abnormal value processing and blockchain storage: remove abnormal fluctuation values in collected data, store key information through hash algorithm, and form tamper-proof records.

[0062] Device interconnection and data collection includes: communication protocol: the portable measuring instrument (such as an explosion-proof PDA) is connected with the tank truck liquid level meter (supporting RS485 interface), temperature sensor (Pt100 type) and pressure sensor (accuracy 0.1%FS) through Bluetooth 5.0 or LoRa, automatically synchronizes NTP clock (time error ≤10ms) when first connected, and checks the validity period of sensor calibration certificate (less than 12 months from the last calibration). Data collection strategy: continuously collect 10 sets of liquid level (unit: mm), temperature (℃) and density (kg / m 3 ) data, remove abnormal values exceeding 2σ (standard deviation), and take the mean value of the remaining data as the effective value (e.g. remove 1 abnormal liquid level value, take 9 mean values).

[0063] Blockchain storage includes: hash value generation: combine the collection time (accurate to seconds), equipment number (such as sensor SN code) and data content (liquid level 1234mm, temperature 25.3℃) into a string, generate a 256-bit hash value through SHA-256 algorithm, and associate the hash value of the previous block to form a chain structure. The storage method uploads the hash value and metadata to the alliance chain node (refinery, railway side, and supervisory side jointly maintain), the data cannot be modified once uploaded, and supports subsequent audit through hash value to quickly verify data integrity.

[0064] In some embodiments, the collected metering data is compared with refinery oil information library theoretical data, railway tank car metering database historical data, using a preset intelligent algorithm for comparison analysis, including: normalizing the theoretical data in the refinery oil information library, extracting the theoretical values and allowable error range of oil density and volume; retrieving historical metering data of the same type of oil in the tank car number within the last 12 months from the railway tank car metering database, calculating the mean value, standard deviation and trend of volume and density; using a dynamic time warping algorithm to calculate the similarity of the current metering data and the historical data sequence, and constructing a three-dimensional comparison model combined with the allowable error range of the theoretical data, if the similarity is higher than the preset threshold and the data falls within the theoretical error range, the comparison is determined to pass.

[0065] By normalizing the theoretical data, analyzing the trend of the historical data, and using the dynamic time warping algorithm, a three-dimensional comparison model is constructed to identify data anomalies. The core includes: data preprocessing: extracting the core indicators and error range of the theoretical data, and retrieving historical data to calculate statistical characteristics (mean value, standard deviation, trend). Similarity calculation and model determination: using the dynamic time warping (DTW) algorithm to match the similarity of the current data and the historical data sequence, and comprehensively determining in combination with the theoretical error range.

[0066] Data preprocessing includes: theoretical data processing: extracting the density theoretical value (e.g. 750 kg / m 3 ) and allowable error (±2%), volume theoretical value (e.g. 50 m 3 ) and allowable error (±1.5%) of the current batch of oil from the refinery oil information library, forming the theoretical data interval [735, 765] kg / m³, [49.25, 50.75] m 3 . Historical data retrieval: querying all transportation records of the same type of oil (e.g. gasoline) within the last 12 months from the railway tank car metering database, extracting the volume and density data each time, calculating the historical mean value (e.g. volume mean value 49.8 m³, density mean value 755 kg / m 3 ), standard deviation (volume σ = 0.3 m 3 , density σ = 8 kg / m 3 ) and trend (e.g. the last 3 times volume increases by 0.1 m 3 ).

[0067] The intelligent comparison algorithm includes: DTW similarity calculation: the current volume / density data sequence (such as [50.1, 752]) is matched with the historical data sequence (such as the last three times [49.8, 755], [49.9, 753], [50.0, 754]) in time, the Euclidean distance similarity (threshold set to be greater than or equal to 0.95) is calculated, and whether the data fluctuation mode conforms to the historical law is reflected. Three-dimensional comparison model: only when the following conditions are met at the same time, the comparison is determined to pass: ① Real-time data falls within the theoretical error interval; ② The deviation of real-time data from the historical mean is less than or equal to 2 sigma; and ③ The DTW similarity is greater than or equal to 0.95.

[0068] In some embodiments, the verification according to the vehicle number, the entry state and the metering data, and the generation of the automatic oil unloading instruction after the verification, include: establishing a verification rule engine, configuring vehicle number uniqueness verification, entry state stability verification and metering data logic verification rules, including: verifying whether the vehicle number matches the current operation plan in the refinery oil import information library and is not repeatedly activated, and then verifying whether the fluctuation amplitude of the entry state information during the metering data acquisition is less than a preset stability threshold, and verifying whether the volume value in the metering data is within a preset data range; when all the logic verification rules trigger the pass condition, an encrypted automatic oil unloading instruction containing a timestamp, a vehicle number and a device number is generated.

[0069] By establishing a multi-layer verification rule engine containing vehicle number, entry and metering data, the safety and uniqueness of the instruction generation are ensured. The core includes: rule engine configuration: define vehicle number uniqueness, entry stability, metering data logic three types of verification rules, form a three-dimensional verification system. Encrypted instruction generation: after verification, an encrypted instruction containing a timestamp and a device number is generated to prevent the instruction from being tampered with or repeatedly executed.

[0070] The rule engine configuration includes: vehicle number uniqueness verification: check whether the currently identified vehicle number is in the oil import operation plan list, and whether the oil unloading instruction of the vehicle number has been repeatedly activated (by querying the instruction generation log, to avoid repeated oil unloading of the same vehicle). Entry stability verification: call the entry state data (such as laser ranging value) during metering data acquisition (about 2 minutes), calculate the fluctuation amplitude (maximum value-minimum value), and the lateral / longitudinal distance fluctuation less than or equal to 5mm and the coupler deviation fluctuation less than or equal to 2cm are considered stable. Metering data logic verification: the volume value needs to meet "0 < real-time volume < nominal volume of tank truck" (such as nominal volume 55m 3 , then the real-time volume needs to be between 1-55m 3 ), and the density value needs to meet the physical properties of oil products (such as diesel density > gasoline).

[0071] The instruction generation and encryption includes: trigger condition: when the three types of rules are passed (car number matching and no repetition, stable entry, and logical measurement data), the system automatically generates instructions. Instruction content: includes timestamp (accurate to milliseconds), car number, device number (such as loading and unloading port ID-001), instruction validity period (30 minutes), uses AES-256 encryption algorithm to encrypt the instructions, and adds a digital signature (generated based on the device private key), ensuring that the instructions cannot be tampered with and can only be executed once.

[0072] In some embodiments, the railway oil unloading automation control system receives an automatic oil unloading instruction, performs an unloading start state confirmation, and starts the unloading operation after confirming that the device state is normal, including: the railway oil unloading automation control system sends state query instructions to the unloading pipeline valve, centrifugal pump, and electrostatic grounding device in turn, and receives feedback signals from each device; analyzes the device operating parameters, fault codes, and safety interlocking states in the feedback signals, and if the valve opening feedback value is consistent with the initial state, the centrifugal pump motor temperature is lower than the warning threshold, and the electrostatic grounding resistance is less than the safety limit, then the device state is determined to be normal; after the state confirmation is passed, a segmented start instruction is sent to the field execution mechanism, the tank car bottom unloading valve is opened first, the centrifugal pump is started after a delay of 10 seconds, and the unloading start time and device start log are recorded synchronously.

[0073] Through device state query, safety interlocking verification, and segmented start strategy, the unloading operation is safely started. The core includes: device state full-link verification: the states of the valve, centrifugal pump, and electrostatic grounding device are checked in turn to confirm that there is no fault and the safety interlocking is effective. Segmented start control: a delayed start strategy of "valve opening first, pump starting later" is adopted to avoid safety risks caused by sudden changes in pipeline pressure.

[0074] Device state query includes: communication protocol: the automation control system sends a "state query" instruction to the unloading pipeline valve (electric ball valve) through the Modbus TCP protocol to obtain the valve opening (0-100%); sends an instruction to the centrifugal pump control cabinet to obtain the motor temperature (°C) and speed (rpm); sends an instruction to the electrostatic grounding device to obtain the grounding resistance (Ω, safety limit ≤10Ω).

[0075] Safety interlocking verification: the following conditions must be met: ① valve opening feedback value = 0% (initial state is closed); ② centrifugal pump motor temperature < 60°C (warning threshold); ③ electrostatic grounding resistance ≤ 10Ω; if any condition is not met, return "device abnormal", suspend the operation and alarm (such as "electrostatic grounding failure").

[0076] The execution steps include: ① Send instruction to open the tank truck bottom unloading valve (pneumatic valve), wait for 5 seconds, and read the valve feedback signal (opening degree 100% confirmation); ② After delaying for 10 seconds (to ensure that the pipeline is not under pressure), start the centrifugal pump, and record the starting time (accurate to seconds); Log record: record the sending time, device response time and feedback parameters of each instruction in real time to form an “unloading start log” for subsequent fault tracing (if the valve response time is greater than 15 seconds, mark it as a device fault).

[0077] In some embodiments, after the oil unloading is completed, an end state confirmation is performed, and an oil handling certificate information report is generated, including: the liquid level change is monitored in real time through the tank truck liquid level sensor, and when the liquid level data is continuously lower than the safe low threshold for 5 minutes, it is determined that the oil unloading operation is completed; the centrifugal pump and the unloading valve are automatically closed by the control system, and the state data after the device is reset is collected, including the valve closing feedback signal, the pipeline pressure zero value and the empty truck state image captured by the on-site video; the truck number, operation time, measurement data, device state record and block chain notarization hash value are integrated into the preset report template, and a PDF format oil handling certificate information report is automatically generated, which is pushed to the refinery information management system through an encrypted channel, and is simultaneously backed up to a distributed database.

[0078] The oil unloading is completed by continuous liquid level monitoring, the device is automatically reset, and an electronic report containing block chain notarization information is generated. The core includes: operation end intelligent judgment: the liquid level continuously lower than the safety threshold is used as the end condition to avoid manual misjudgment of residual amount. Whole process data archiving: integrate truck number, measurement data, device state, block chain hash value and other information to generate a standardized report and encrypt transmission.

[0079] The end state confirmation includes: liquid level monitoring: the data is collected in real time by the built-in liquid level sensor of the tank truck, and when the liquid level value is continuously less than 50mm (safety low threshold, corresponding to remaining volume <0.5m 3 ) for 5 minutes, it is determined that the oil unloading is completed (to prevent misjudgment due to liquid level fluctuation). Device reset: sequentially close the centrifugal pump (first stop the pump, delay for 5 seconds and then close the pipeline valve), collect the state after reset: valve feedback opening degree 0%, pipeline pressure 0MPa, electrostatic grounding device signal disconnected, and simultaneously trigger the on-site camera to capture the empty truck state (verify that there is no dripping at the bottom of the tank truck).

[0080] Report generation and transmission include: report content: contains truck number, operation time (start / end), measurement data (original 50m 3 , actual 49.8m 3The data includes: loss rate (0.4%), equipment number, blockchain notarized hash value (linked to metering data block), and any abnormal records. Generation and transmission: Data is automatically populated using a template engine (such as Freemarker), generating a PDF report and encrypting it (password generated from vehicle number + timestamp). The report is pushed to the refinery's ERP system via HTTPS, and the report hash value is simultaneously stored on the blockchain. A distributed database (such as HBase) backs up the entire dataset, supporting fast retrieval by vehicle number (retrieval latency ≤ 2 seconds).

[0081] In some embodiments, such as Figure 2 As shown, the business logic corresponding to the provided method includes: start, filling in oil arrival data registration, oil tanker entry identification, oil tanker number identification and comparison, comparison pass (with refinery oil arrival information database), pass (manual correction if not pass), manual confirmation, metering data collection (via portable measuring instrument), metering data comparison and analysis, comparison pass (with refinery oil arrival information database and railway tanker metering database), pass (abnormal situation registration if not pass), verification and confirmation, implementation of automatic oil unloading, confirmation of start and end status of oil unloading operation by railway oil unloading automation control system, automatic generation of report (oil receiving and dispatching certificate information report), end.

[0082] In some embodiments, a Long Short-Term Memory (LSTM) network is trained using historical equipment operating data to construct a fault prediction model for key equipment such as centrifugal pumps and valves, enabling preventative maintenance and reducing the risk of unplanned downtime. Core technologies include: Multi-dimensional feature engineering: extracting time-series features such as equipment vibration frequency, temperature change rate, and current fluctuations, and combining them with work order records to construct a fault label dataset. LSTM anomaly detection model: training a baseline model using normal equipment operating data, and identifying early fault signs through residual analysis of real-time data and predicted values.

[0083] Data Acquisition and Feature Construction: A triaxial accelerometer (sampling frequency 1000Hz) was installed in the centrifugal pump bearing housing, and a current transformer was installed in the motor junction box. Vibration waveforms (time-domain mean, kurtosis), RMS current values, and temperature gradients (ΔT / 10min) were collected every 5 minutes, representing 12 dimensions of features. Historical fault data (such as bearing wear and impeller corrosion) were labeled, and a balanced dataset with a positive-to-negative sample ratio of 1:3 was constructed. Data from the 72 hours prior to the fault was marked as early warning samples.

[0084] Model training and application include: Training process: A 2-layer LSTM network (128 neurons per layer) is used. The input is the feature sequence from the previous 24 hours, and the output is the probability of failure in the next 6 hours (threshold set to ≥85% to trigger an early warning). The Adam optimizer is used, with the loss function being binary cross-entropy, and the validation set accuracy is ≥92%. Real-time prediction: The control system inputs the current feature sequence to the model every 10 minutes. If the predicted failure probability is >80%, a maintenance work order (including the failure type and suggested repair time) is automatically generated, and a backup pump is scheduled for switching. Simultaneously, an early warning is pushed to the maintenance terminal (response latency ≤30 seconds).

[0085] In some embodiments, to address the challenge of license plate recognition under different lighting and dirt occlusion scenarios, transfer learning is employed to optimize the CNN model, combined with domain adaptation algorithms to improve generalization ability. Core technologies include: Cross-domain data augmentation: CycleGAN is used to generate simulated images such as those from rainy days, nighttime scenes, and oil stain coverage to expand the training dataset. Meta-learning fine-tuning mechanism: Based on the pre-trained model, the classification layer is rapidly fine-tuned for new scenarios to address the problem of decreased recognition accuracy in small sample scenarios.

[0086] Data Augmentation and Model Architecture: Simulated Data Generation: Utilizing CycleGAN to learn the mapping relationship between normal license plate images and images of adverse scenes, 50,000 augmented data images (e.g., adding Gaussian noise, rain streaks, and oil stain masks) were generated and mixed with real scene data (20,000 images) for training. Transfer Learning Framework: Based on ResNet50, the first four convolutional layers were frozen, and the last three layers were replaced with adaptive feature extraction layers, supporting feature normalization for different lighting / occlusion scenarios.

[0087] Real-time recognition optimization includes: Scene detection: Determining the current scene type (e.g., "low light at night" or "oil stain obstruction") based on prior features such as image brightness, contrast, and the proportion of stained areas, and automatically loading the corresponding fine-tuned model (switching latency ≤200ms). Dynamic fusion strategy: A voting decision is made based on the recognition results from multiple cameras. If the confidence level of a single camera in a certain scene is <90%, image fusion of adjacent cameras is triggered (e.g., reshooting from the left camera + image enhancement from the right camera), and the fused license plate number is output (recognition accuracy improved to 99.2%).

[0088] In some embodiments, a reinforcement learning model for oil unloading operation scheduling is constructed to dynamically optimize the tanker unloading sequence and equipment allocation strategy with the goal of minimizing the total operation time. Core technologies include: State-space modeling: Encoding parameters such as tanker type (light oil / heavy oil), equipment availability, and metering data integrity into state vectors. Action-space design: Defining 12 scheduling actions such as "allocate to loading / unloading port 1" and "prioritize unloading heavy oil tankers," and designing a reward function that combines waiting time and equipment load balancing.

[0089] Environment modeling and reward function: State vector: 20-dimensional parameters including the queue of vehicles to be unloaded (vehicle number, oil type, volume), the current state of each loading and unloading device (valve / pump availability, last maintenance time), and the capacity limit of the storage area. Reward function: +100 for completing the unloading of a vehicle, -50 for exceeding 80% of the device load, -10 for each 1-minute delay, and the policy network is trained using the Proximal Policy Optimization (PPO) algorithm.

[0090] Real-time scheduling application: Input interface: real-time access to the list of vehicles to be unloaded (including estimated arrival time and priority) from the refinery oil information library, triggering a scheduling re-optimization every time 3 new vehicles are added. Decision output: the model outputs the optimal scheduling sequence (e.g., "Vehicle A → Loading and Unloading Port 2, Vehicle B → Loading and Unloading Port 1 (priority for heavy oil)"), and generates a device preheating plan (starting the centrifugal pump preheating 5 minutes in advance to reduce idle loss). After manual confirmation, it is automatically executed, reducing the average scheduling time by 30%.

[0091] In some embodiments, by constructing a spatio-temporal graph neural network (ST-GNN) containing device status, personnel location, and video stream, complex abnormal behaviors such as personnel misoperation and device abnormal linkage in the unloading area are identified. The core technologies include: multi-modal data graph modeling: personnel trajectories detected by cameras, device sensor data, and job process nodes are abstracted as graph nodes, and edges represent spatio-temporal correlation. Abnormal score calculation: learn normal behavior patterns through graph convolutional neural network (GCN), and subgraph structures with a deviation greater than 3σ are determined as abnormal.

[0092] Graph structure construction and data fusion includes: node definition: personnel nodes (ID, location, action label), device nodes (number, state parameters), and process nodes (job steps, timestamps), with a total of 3 types and 15 node attributes.

[0093] Edge definition: an interaction edge is established when the distance between personnel and devices is less than 2m, a time sequence edge is established between device state changes and process nodes, forming a dynamic spatio-temporal graph (updated every 10 seconds).

[0094] Abnormal detection process includes: model training: use normal operation data to train ST-GNN to learn node state transition rules (e.g., "open valve → start pump after 10 seconds").

[0095] Real-time detection includes: when "unconfirmed entry starts pump" (personnel do not complete the confirmation process but trigger device start) or "multiple people enter dangerous area" (exceeding the safety number limit) occurs, the graph model calculates the abnormal score >0.85, immediately triggers sound and light alarms and freezes device control authority, and pushes the alarm information containing the abnormal subgraph to the safety management platform (response time ≤1 second).

[0096] In some embodiments, by addressing the data privacy protection needs of refineries and railway parties, a federated learning framework is used to collaboratively train a metering data verification model, improving comparison accuracy without sharing raw data. Core technologies include: horizontal federated learning architecture: refineries (theoretical data) and railways (historical metering data) as two data parties, collaboratively train neural networks, locally calculate gradients and upload encrypted aggregates. Differential privacy protection: add Laplace noise when aggregating gradients to ensure data privacy compliance (ε=0.5).

[0097] The federated learning process design includes: data preprocessing: the refinery party prepares the theoretical data (oil density, volume interval), the railway party prepares the historical metering data (car number, measured value, timestamp), and the data of both parties is aligned by car number ID, and 10% public ID is reserved for model verification. The model architecture uses a two-layer fully connected neural network (input layer 10D, hidden layer 32D, output layer 1D comparison result), the refinery party inputs the theoretical error range, the railway party inputs the historical mean / standard deviation, and both parties upload the gradient after local training.

[0098] Collaborative verification applications include: model updating: a federated training iteration is performed once a week, and the model comparison accuracy is stable at more than 98% after 5 or more aggregation times. Real-time verification: when new metering data is generated, the refinery inputs the theoretical parameters and the railway inputs the historical features, both parties call the federated model for encrypted comparison (intermediate results are protected by homomorphic encryption technology), and the verification result is returned within 10 seconds, protecting commercial data privacy and addressing the problem of insufficient generalization ability of single data source models.

[0099] In some embodiments, by building a digital twin of the unloading operation area, real-time data is used to drive virtual scene simulation, combined with physical models and machine learning to predict pressure fluctuations and leakage risks during unloading. Core technologies include: multi-physical field coupling modeling: establish a fusion simulation framework of Navier-Stokes equation for oil flow, equipment thermodynamic model, and sensor data. Twin data-driven: use real-time liquid level and pressure data to calibrate virtual model parameters to predict future 30-minute process states through simulation.

[0100] Digital twin construction includes: geometric modeling: use BIM technology to build a 3D model of the unloading area (accuracy ±5mm), integrate CAD drawings of tank cars, pipelines, and pump bodies, and define material properties (such as pipeline thermal conductivity, oil viscosity). Data interface: collect 100 sets of sensor data (liquid level, pressure, temperature) per second, drive virtual scene dynamic updates after fusion through Kalman filtering, with an error control within 1.5%.

[0101] The risk prediction application includes: pressure fluctuation prediction: when the rotating speed of the centrifugal pump changes, the twin model calculates the fluid mechanics (CFD) simulation pipeline pressure distribution, if the predicted pressure at a bend > design threshold (1.2 MPa), 5 minutes in advance warning "pipeline overload risk", automatically adjust the pump rotating speed to the safety interval (adjustment step ≤ 50 rpm). Leakage simulation: inject "valve sealing ring aging" fault in the virtual scene, simulate the leakage diffusion path, combine the wind direction, topography data to predict the impact range, guide the optimization of on-site emergency exercise scheme (such as determining a 30-meter safety warning radius), realize the "prediction-prevention-disposal" closed-loop management.

[0102] Please refer to Figure 3 as shown, Figure 3 is a structural schematic diagram of a railway tank car intelligent checking system 200 provided by the embodiments of the present application. The railway tank car intelligent checking system 200 is used to execute the steps of the railway tank car intelligent checking method shown in each of the above embodiments. The railway tank car intelligent checking system 200 can be a single server or a server cluster, or the railway tank car intelligent checking system 200 can be a terminal, which can be a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0103] As Figure 3 shown, the railway tank car intelligent checking system 200 includes: a data acquisition unit 201, configured to acquire oil-related data, acquire the entry state information of the oil tank car through an identification device, collect the car number image of the oil tank car according to the image recognition device and convert it into character information; an information comparison unit 202, configured to compare the character information with the tank car number information of the corresponding batch in the refinery oil information library: if the comparison is passed, start the measurement data acquisition, and acquire the measurement data of the oil product in the tank car through a portable measuring instrument; an instruction generation unit 203, configured to compare and analyze the collected measurement data with the theoretical data of the refinery oil information library and the historical data of the railway tank car measurement database by using a preset intelligent algorithm: if the comparison is passed, verify according to the car number, the entry state and the measurement data, and generate an automatic oil unloading instruction after the verification is passed; an intelligent checking unit 204, configured to receive the automatic oil unloading instruction by the railway oil unloading automatic control system, perform the unloading start state confirmation, start the unloading operation after confirming that the equipment state is normal, perform the end state confirmation after the unloading is completed, generate an oil handling certificate information report, and complete the railway tank car intelligent checking.

[0104] It should be noted that, for the convenience and brevity of description, the specific working processes of the railway tank car intelligent checking system and each module described above can refer to the corresponding processes in the railway tank car intelligent checking method embodiments described above, which will not be described herein.

[0105] The railway tank car intelligent checking method described above can be implemented in the form of a computer program, which can run on a system as shown in the figure. Figure 3

[0106] Please refer to Figure 4 , Figure 4 is a structural schematic block diagram of a computer device provided by the embodiment of the present application. The computer device comprises a processor, a memory and a network interface connected through a device bus, wherein the memory can comprise a storage medium and an internal memory.

[0107] The storage medium can store an operating device and a computer program. The computer program comprises program instructions, which, when executed, can cause the processor to execute any kind of railway tank car intelligent checking method.

[0108] The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

[0109] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, can cause the processor to execute any kind of railway tank car intelligent checking method.

[0110] The network interface is used for network communication, such as sending assigned tasks. Those skilled in the art can understand that Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the terminal to which the scheme of the present application is applied. The specific computer device can comprise more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0111] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0112] Among them, in one embodiment, the processor is used to run the computer program stored in the memory to implement the following steps: ​After obtaining the oil-related data, the in-position state information of the tank truck is obtained through the identification device, and the vehicle number image of the tank truck is collected by the image recognition device and converted into character information; The character information is compared with the tank truck vehicle number information of the corresponding batch in the refinery oil information library: if the comparison is passed, the measurement data collection is started, and the measurement data of the oil in the tank truck is obtained by the portable measuring instrument; The collected measurement data is compared and analyzed with the theoretical data of the refinery oil information library and the historical data of the railway tank car measurement database using a preset intelligent algorithm: if the comparison is passed, the vehicle number, in-position state and measurement data are verified, and the automatic oil unloading instruction is generated after the verification is passed; The railway oil unloading automatic control system receives the automatic oil unloading instruction, confirms the unloading start state, starts the unloading operation after confirming that the device state is normal, confirms the end state after the unloading is completed, generates the oil receiving and sending certificate information report, and completes the intelligent verification of the railway tank car.

[0113] In some embodiments, the in-position state information of the tank truck is obtained by the identification device, including: deploying laser ranging sensors, pressure sensors and visual recognition devices at preset positions of the track in the railway oil unloading operation area, collecting distance data of the tank truck and the loading and unloading port by the laser ranging sensors, collecting pressure distribution data of the wheel on the track by the pressure sensors, and collecting the relative position image of the coupler and the track identification line by the visual recognition device; input the distance data, pressure distribution data and position image into an edge computing unit, and perform logical judgment based on a preset in-position rule library: if the distance data is in a safe operation interval, the pressure distribution data meets the track bearing balance condition, and the coupler and the identification line in the position image deviate by less than a preset threshold, it is determined that the tank truck is in the ready state.

[0114] In some embodiments, the vehicle number image of the tank truck is collected by the image recognition device and converted into character information, including: deploying multi-angle high-definition cameras on both sides of the oil unloading operation area, dynamically scanning the vehicle number area on the side and end surface of the tank truck, and obtaining multiple frames of continuous images; denoising, distortion correction and image enhancement preprocessing are performed on the multiple frames of images, a convolutional neural network character recognition model is used to perform vehicle number character segmentation and recognition on the preprocessed images, and initial character information is generated; the initial character information is matched with the candidate vehicle number of the same batch transportation plan in the refinery oil information library, and when the matching degree exceeds a preset threshold, the final confirmed vehicle number character information is output.

[0115] In some embodiments, the start of the metering data acquisition, the acquisition of the tank truck oil metering data by the portable measuring instrument, comprises: the portable measuring instrument establishes a wireless communication connection with the liquid level meter, temperature sensor and pressure sensor of the tank truck, automatically synchronizes the clock and checks the equipment calibration time; sequentially collects the oil level height, temperature, density and volume data, and takes the mean value after removing the abnormal fluctuation value as the effective metering data; the acquisition time, equipment number and data content of the metering data are stored by hash value through blockchain technology, and the metering data record which cannot be tampered is generated.

[0116] In some embodiments, the collected metering data is compared and analyzed with the theoretical data of the refinery incoming oil information library and the historical data of the railway tank car metering database by using a preset intelligent algorithm, which comprises: the theoretical data in the refinery incoming oil information library is normalized, and the theoretical values and allowable error range of oil density and volume are extracted; the historical metering data of the same type of oil in the tank car number in the past 12 months is called from the railway tank car metering database, and the mean value, standard deviation and change trend of volume and density are calculated; the similarity of the current metering data and the historical data sequence is calculated by using the dynamic time warping algorithm, a three-dimensional comparison model is constructed by combining the allowable error range of the theoretical data, and if the similarity is higher than the preset threshold and the data falls within the theoretical error range, the comparison is determined to pass.

[0117] In some embodiments, the verification is performed according to the car number, the in-place state and the metering data, and the automatic unloading instruction is generated after the verification passes, which comprises: establishing a verification rule engine, configuring car number uniqueness verification, in-place state stability verification and metering data logic verification rules, including: verifying whether the car number matches the current operation plan in the refinery incoming oil information library and has not been repeatedly activated, and then verifying whether the fluctuation amplitude of the in-place state information during the metering data acquisition period is less than the preset stability threshold, and verifying whether the volume value in the metering data is within the preset data range; when all the logic verification rules trigger the pass condition, an encrypted automatic unloading instruction containing the timestamp, car number and equipment number is generated.

[0118] In some embodiments, the railway unloading automation control system receives the automatic unloading instruction, confirms the unloading start state, and starts the unloading operation after confirming that the device state is normal, which comprises: the railway unloading automation control system sends state query instructions to the unloading pipeline valve, centrifugal pump and electrostatic grounding device in turn, and receives the feedback signals of each device; analyzes the device operating parameters, fault codes and safety interlocking state in the feedback signals, if the valve opening feedback value is consistent with the initial state, the centrifugal pump motor temperature is lower than the warning threshold, and the electrostatic grounding resistance is less than the safety limit, then the device state is determined to be normal; after the state confirmation passes, a segmented start instruction is sent to the field execution mechanism, the tank truck bottom unloading valve is opened first, the centrifugal pump is started after a delay of 10 seconds, and the unloading start time and equipment start log are recorded synchronously.

[0119] In some embodiments, after the oil unloading is completed, an end state confirmation is performed, and an oil receiving and sending certificate information report is generated, including: real-time monitoring of liquid level changes through tank truck liquid level sensors, and determining that the oil unloading operation is completed when the liquid level data is continuously lower than the safe low threshold for 5 minutes; the control system automatically closes the centrifugal pump and the oil unloading valve, collects state data after the equipment is reset, including valve closing feedback signals, pipeline pressure zero value, and empty truck state image captured by the on-site video; the truck number, operation time, measurement data, equipment state record, and block chain notarization hash value are integrated into a preset report template, and an oil receiving and sending certificate information report in PDF format is automatically generated, pushed to the refinery information management system through an encrypted channel, and simultaneously backed up to a distributed database.

[0120] In the embodiments of the present application, a computer readable storage medium is also provided, which stores a computer program including program instructions. The processor executes the program instructions to implement the steps of the railway tank car intelligent checking method provided by the above embodiments of the present application.

[0121] The computer readable storage medium can be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0122] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for intelligent checking of a railway tank car, characterized in that, The method comprises the following steps: After obtaining the oil-related data, the in-position state information of the tank truck is obtained through the identification device, the vehicle number image of the tank truck is collected by the image recognition device, and the vehicle number image is converted into character information; The character information is compared with the tank truck vehicle number information of the corresponding batch in the refinery oil information library: if the comparison is passed, the measurement data collection is started, and the oil product measurement data in the tank truck is obtained through the portable measuring instrument; The collected measurement data is compared and analyzed with the theoretical data of the refinery oil information library and the historical data of the railway tank car measurement database by using a preset intelligent algorithm: if the comparison is passed, the vehicle number, the in-position state and the measurement data are verified, and the automatic unloading instruction is generated after the verification is passed; The railway unloading automatic control system receives the automatic unloading instruction, performs unloading start state confirmation, starts the unloading operation after confirming that the equipment state is normal, performs end state confirmation after the unloading is completed, generates oil receiving and sending certificate information report, and completes the intelligent checking of the railway tank car.

2. The method of claim 1, wherein, The in-position state information of the tank truck is obtained through the identification device, comprising: Laser ranging sensors, pressure sensors and visual recognition devices are deployed at preset positions of the track in the railway unloading operation area, distance data of the tank truck and the loading and unloading port is collected by the laser ranging sensors, pressure distribution data of the wheels on the track is collected by the pressure sensors, and relative position images of the coupler and the track identification line are collected by the visual recognition devices; The distance data, pressure distribution data and position images are input into an edge computing unit, logical judgment is performed based on a preset in-position rule library, if the distance data is in a safe operation interval, the pressure distribution data meets the track bearing balance condition, and the deviation of the coupler and the identification line in the position image is less than a preset threshold, it is determined that the in-position state of the tank car is ready.

3. The method of claim 1, wherein, The vehicle number image of the tank truck is collected by the image recognition device and converted into character information, comprising: Multi-angle high-definition cameras are deployed on both sides of the unloading operation area, the vehicle number area on the side and end surface of the tank truck is dynamically scanned, and multiple continuous images are obtained; The multiple images are preprocessed by noise reduction, distortion correction and image enhancement, a convolutional neural network character recognition model is used to segment and recognize the vehicle number characters of the preprocessed images, and initial character information is generated; The initial character information is matched with the candidate vehicle number of the same batch transportation plan in the refinery oil information library, and when the matching degree exceeds a preset threshold, the final confirmed vehicle number character information is output.

4. The method of claim 1, wherein, The portable measuring instrument is wirelessly connected with the liquid level meter, temperature sensor and pressure sensor of the tank car, automatically synchronizes the clock and verifies the equipment calibration time; the oil product liquid level height, temperature, density and volume data are sequentially collected, and the mean value after removing the abnormal fluctuation value is taken as the effective measurement data; The collection time, equipment number and data content of the measurement data are stored by hash value through the blockchain technology, and an unalterable measurement data record is generated. The collected measurement data is compared and analyzed with the theoretical data of the refinery oil information library and the historical data of the railway tank car measurement database by using a preset intelligent algorithm, comprising:

5. The method of claim 1, wherein, ​ The theoretical data in the refinery oil information base is normalized, and the theoretical values and allowable error range of oil density and volume are extracted; the historical measurement data of the same type of oil in the tank car measurement database within the last 12 months is called, and the mean value, standard deviation and variation trend of the volume and density are calculated; The similarity between the current measurement data and the historical data sequence is calculated using the dynamic time warping algorithm, and a three-dimensional comparison model is constructed by combining the allowable error range of the theoretical data. If the similarity is higher than the preset threshold and the data falls within the theoretical error range, the comparison is determined to pass.

6. The method of claim 1, wherein, The verification is performed according to the car number, entry state and measurement data, and an automatic unloading instruction is generated after the verification passes, including: A verification rule engine is established, and car number uniqueness verification, entry state stability verification and measurement data logic verification rules are configured, including: verifying whether the car number matches the current operation plan in the refinery oil information base and has not been repeatedly activated, and then verifying whether the fluctuation amplitude of the entry state information during the measurement data acquisition period is less than a preset stability threshold, and verifying whether the volume value in the measurement data is within a preset data range; When all the logic verification rules trigger the pass condition, an encrypted automatic unloading instruction containing the timestamp, car number and equipment number is generated.

7. The method of claim 1, wherein, The railway unloading automatic control system receives the automatic unloading instruction, performs unloading start state confirmation, and starts the unloading operation after confirming that the equipment state is normal, including: The railway unloading automatic control system sends state query instructions to the unloading pipeline valve, centrifugal pump and electrostatic grounding device in turn, and receives feedback signals from each device; The device operating parameters, fault codes and safety interlocking states in the feedback signals are analyzed. If the valve opening feedback value is consistent with the initial state, the centrifugal pump motor temperature is lower than the warning threshold, and the electrostatic grounding resistance is less than the safety limit, the device state is determined to be normal; After the state confirmation passes, a segmented start instruction is sent to the field actuator, the tank car bottom unloading valve is opened first, the centrifugal pump is started after a delay of 10 seconds, and the unloading start time and equipment start log are recorded synchronously.

8. The method of claim 1, wherein, After the unloading is completed, the end state is confirmed, and an oil handling certificate information report is generated, including: The liquid level sensor is used to monitor the liquid level change in real time. When the liquid level data is continuously lower than the safety low threshold for 5 minutes, it is determined that the unloading operation is completed; The control system automatically closes the centrifugal pump and unloading valve, and collects the state data after the equipment is reset, including the valve closed feedback signal, pipeline pressure zero value and empty car state image captured by the field video; The car number, operation time, measurement data, equipment state record and blockchain notarization hash value are integrated into the preset report template, and an oil handling certificate information report in PDF format is automatically generated, which is pushed to the refinery information management system through an encrypted channel, and is also backed up to a distributed database.

9. A railway tank car intelligent checking system, characterized in that, including: The data acquisition unit is used to acquire oil-related data, identify the entry state information of the oil tank car through the equipment, and collect the car number image of the oil tank car through the image recognition equipment and convert it into character information; The information comparison unit is used for comparing the character information with the tank car number information of the corresponding batch in the refinery incoming oil information library. If the comparison is passed, the metering data acquisition is started, and the metering data of the oil in the tank car are acquired by the portable measuring instrument. The instruction generation unit is used for comparing and analyzing the acquired metering data with the theoretical data of the refinery incoming oil information library and the historical data of the railway tank car metering database by using a preset intelligent algorithm. If the comparison is passed, the automatic oil unloading instruction is generated according to the car number, the entering state and the metering data. The intelligent checking unit is used for receiving the automatic oil unloading instruction by the railway oil unloading automation control system, confirming the starting state of the oil unloading, starting the oil unloading operation after confirming that the equipment state is normal, confirming the ending state after the oil unloading is completed, generating the oil receiving and sending certificate information report, and completing the intelligent checking of the railway tank car.

10. A computer device, comprising: The computer device comprises a memory and a processor; The memory is used for storing a computer program; The processor is used for executing the computer program and realizing the method in any one of claims 1 to 8 when the computer program is executed.

Images