Sensor data processing system and method for liquid packaging barrels

Abstract

The present invention relates to the technical fields of big data analysis and supply chain management, and in particular to a sensor data processing system and method for liquid packaging barrels. The method comprises the following steps: obtaining raw sensor data from a smart chip shared barrel; performing barrel status judgment on the raw sensor data to obtain status event data; locally caching and transmitting the status event data to obtain a real-time perception data stream; performing perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and barrel historical loading information; calculating liquid usage rate data based on the barrel historical loading information and optimized perception data; performing a comprehensive judgment on the container status based on the barrel historical loading information and optimized perception data to obtain currently loaded liquid type data; and generating comprehensive barrel status data based on the liquid type data and liquid usage rate data. The present invention significantly improves the efficiency and benefits of bulk liquid trade by integrating real-time perception data from smart containers.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis and supply chain management, and in particular to a sensor data processing system and method for a liquid packaging barrel. Background Art

[0002] Existing technologies primarily rely on regular manual inspections and the statistical analysis of historical sales data to manage containers and forecast demand. Manual inspections are infrequent, and critical information such as container liquid levels, locations, and status cannot be accurately monitored in real time. Historical sales data can only reflect past demand, making it difficult to accurately predict future trends. This results in a slow response to market changes and can easily lead to supply disruptions or waste of resources. Manually estimated liquid levels and usage patterns are subject to significant errors, and historical sales data also struggles to reflect the differentiated needs of individual customers. This results in low demand forecast accuracy, making it difficult to develop precise replenishment plans, and can easily lead to inventory backlogs or supply shortages. Traditional container management methods lack effective tracking and monitoring methods, making it difficult to monitor the real-time location and usage status of containers. This can easily lead to containers being lost, damaged, or left idle, increasing operating costs and reducing asset utilization.

[0003] In summary, existing technologies lack real-time, refined insights into container status and usage, making it difficult to dynamically predict customer demand and optimize replenishment timing, as well as inefficient container asset management, which needs to be addressed urgently. Summary of the Invention

[0004] Based on this, it is necessary to provide a sensor data processing system and method for liquid packaging barrels to solve at least one of the above technical problems.

[0005] To achieve the above-mentioned purpose, a method for processing sensor data of a liquid packaging barrel includes the following steps:

[0006] Step S1: Obtain the original sensor data of the smart chip shared bucket; perform bucket status judgment on the original sensor data to obtain status event data; locally cache and transmit the status event data to obtain a real-time perception data stream;

[0007] Step S2: Perform perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and historical barrel loading information; calculate liquid usage rate data based on the historical barrel loading information and the optimized perception data; perform comprehensive container status determination based on the historical barrel loading information and the optimized perception data to obtain currently loaded liquid type data; and generate comprehensive barrel status data based on the liquid type data and the liquid usage rate data.

[0008] Step S3: Perform customer usage and consumption analysis on the comprehensive bucket state data to obtain customer consumption characteristics; perform dynamic depletion time prediction based on the comprehensive bucket state data and customer consumption characteristics to obtain depletion prediction data; perform container location pattern analysis on the comprehensive bucket state data to obtain location pattern data; perform container life cycle assessment based on the location pattern data, depletion prediction data, and comprehensive bucket state data to obtain demand forecast and status assessment data;

[0009] Step S4: Perform barrel status warning and logistics optimization based on demand forecast and status assessment data to obtain an optimized path plan; perform trade scheduling and execution based on the optimized path plan to obtain optimized scheduling instructions.

[0010] The smart chip shared barrel used in the present invention is an intelligent container system that integrates multiple sensors. The barrel body is generally cylindrical, with a filling port and a vent valve on the top and a drain port on the bottom. Its core components include a multi-type sensor array and an intelligent main control chip built into the barrel structure. The sensor array consists of a pressure sensor MS5803-14BA, a MEMS three-axis tilt module SCL3300, an ADXL345 digital accelerometer, an embedded GNSS dual-mode receiver chip, and an RTC real-time clock chip DS3231. Among them, the pressure sensor is installed at the center of the barrel bottom to monitor the static pressure changes of the liquid in the barrel in real time; the tilt sensor is fixed at the center line of the barrel shell side wall to continuously collect the tilt angle of the barrel in multiple axes; the vibration sensor is located slightly above the inner wall of the barrel to capture the vibration frequency and acceleration data of the barrel in real time; the antenna of the GNSS positioning module is installed on the top shell of the barrel to ensure stable reception of the positioning signal; and the clock module provides accurate time stamps for all collected data. These sensors sample data at a preset frequency (usually 1 Hz) and transmit it to the main control chip via the I2C bus for preliminary processing, caching and standardization. It is then uploaded to the cloud server via the wireless communication module, forming a continuous real-time perception data stream.

[0011] By deploying intelligent chips integrating pressure, tilt, vibration, positioning, and clock modules within shared buckets, the system enables high-frequency, automated collection of bucket status, eliminating the delays and errors associated with traditional manual inspections. Leveraging multi-sensor fusion technology, the system can identify key events such as tipping, collisions, and sudden drops in liquid levels in real time, providing timely insights into liquid usage dynamics and container physical risks. Local caching and resumable transmission mechanisms ensure that critical data is not lost even in unstable network environments, ensuring data continuity and integrity. The resulting real-time perception data stream provides a high-quality, structured foundation for subsequent data fusion and status analysis, enabling a stable mapping of physical-world status to the digital world. By standardizing the real-time perception data stream and jointly modeling historical loading data, the system achieves a deep fusion of bucket status and liquid properties. Liquid level fluctuations are corrected and compensated using container structural parameters and liquid physical properties, resulting in more accurate calculations of liquid usage rates that reflect actual consumption trends. Leveraging event rules and feature recognition algorithms, the system can automatically identify complex states such as transport status, liquid replacement, and abnormal tilt, accurately determining the type of liquid currently contained in the container. The resulting comprehensive barrel status data not only includes current status information but also incorporates historical loading and real-time behavior characteristics, providing multi-dimensional support for customer behavior modeling and subsequent forecasting. By correlating barrel status data with customer dimensions and leveraging algorithms such as frequency domain analysis, trend modeling, and cycle identification to extract liquid usage patterns, the system accurately analyzes customer consumption characteristics for various liquid types and implements differentiated modeling. Combining real-time trends with historical stability, the system dynamically predicts liquid depletion times, enhancing the foresight and accuracy of replenishment responses. Based on location clustering and path analysis, the system automatically identifies container dwell points and flow paths, and models cyclical behavior using autocorrelation functions to generate high-confidence location behavior predictions. Furthermore, by integrating liquid level, location, and status information, the system assesses container lifecycle stages (in use, empty, abnormal, etc.), providing a scientific basis for asset recovery, maintenance, and redistribution. Based on depletion predictions and lifecycle assessment results, the system automatically identifies containers requiring replenishment, recycling, or abnormal handling, generates an alert list, and integrates inventory, transportation, and packaging resources to establish an end-to-end response mechanism. Through task merging and path planning algorithms, replenishment and recycling routes and resource allocation are significantly optimized, reducing logistics costs and improving execution efficiency. Automatic generation of repackaging instructions and quality inspection processes ensures liquid dispensing quality and operational consistency, while abnormal event handling notifications enable rapid response to potential risks. Ultimately, by integrating all tasks through a unified scheduling instruction package, the system achieves closed-loop management across forecasting, scheduling, and execution, enhancing the intelligence and operational flexibility of the entire liquid supply chain.

[0012] Therefore, this invention provides a sensor data processing method for liquid packaging barrels. By integrating real-time sensor data (liquid level, location, and status) from intelligent containers and combining it with historical usage records and customer consumption analysis, a data-driven intelligent trade decision support system is constructed. This system can monitor container status in real time, dynamically predict customer demand, optimize replenishment timing and logistics routes, and achieve efficient container asset management, thus addressing the aforementioned drawbacks and significantly improving the efficiency and benefits of bulk liquid trade.

[0013] Preferably, the present invention further provides a sensor data processing system for a liquid packaging barrel, for executing the sensor data processing method for a liquid packaging barrel as described above, the sensor data processing system for a liquid packaging barrel comprising:

[0014] The bucket-side status perception module is used to obtain the original sensor data of the smart chip shared bucket; determine the bucket status of the original sensor data to obtain status event data; and locally cache and transmit the status event data to obtain a real-time perception data stream.

[0015] The comprehensive barrel state fusion module is used to perform perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and historical barrel loading information; calculate liquid usage rate data based on the historical barrel loading information and optimized perception data; perform comprehensive container status judgment based on the historical barrel loading information and optimized perception data to obtain the current loaded liquid type data; and generate comprehensive barrel state data based on the liquid type data and liquid usage rate data.

[0016] The cycle behavior prediction module is used to analyze customer usage and consumption based on the comprehensive bucket data to obtain customer consumption characteristics; dynamically predict depletion time based on the comprehensive bucket data and customer consumption characteristics to obtain depletion prediction data; analyze container location patterns based on the comprehensive bucket data to obtain location pattern data; and perform container life cycle assessment based on location pattern data, depletion prediction data, and comprehensive bucket data to obtain demand forecast and status assessment data.

[0017] The intelligent trade scheduling module is used to perform barrel status warning and logistics optimization based on demand forecast and status assessment data to obtain an optimized path plan; trade scheduling and execution are carried out according to the optimized path plan to obtain optimized scheduling instructions.

[0018] The sensor data processing system for liquid packaging barrels provided by the present invention has constructed a full-process closed-loop architecture from data acquisition, state fusion, behavior prediction to intelligent scheduling through modular design. The barrel-end state perception module realizes high-frequency acquisition and event recognition of key parameters such as container liquid level, position, and posture, providing the system with high-quality and continuous raw data input; the comprehensive barrel state fusion module deeply integrates real-time perception data with historical loading information, accurately calculates the liquid usage rate and determines the container status, and realizes dynamic identification of liquid type and container status; the periodic behavior prediction module constructs a multi-dimensional prediction model based on the mining of customer usage behavior and location trajectory, accurately predicts the exhaustion time and container life cycle, and improves the foresight and personalized ability of demand response; the intelligent trade scheduling module unifies the scheduling resources through automatic triggering and path optimization of replenishment, recycling, and exception handling tasks, and improves allocation efficiency and execution reliability. The overall system has greatly improved the data-driven intelligent decision-making capabilities in bulk liquid trade, and realized automated and precise management from container perception to trade execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The figure is a flow chart of the steps of a method for processing sensor data of a liquid packaging barrel.

[0020] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0022] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0023] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0024] To achieve this, please refer to Figure 1 The present invention provides a method for processing sensor data of a liquid packaging barrel, comprising the following steps:

[0025] Step S1: Obtain the original sensor data of the smart chip shared bucket; perform bucket status judgment on the original sensor data to obtain status event data; locally cache and transmit the status event data to obtain a real-time perception data stream;

[0026] In this embodiment of the present invention, the original sensor data is collected by the smart chip deployed on the shared bucket, including the static pressure value obtained by the pressure sensor, the attitude information collected by the three-axis tilt sensor, the acceleration data recorded by the vibration sensor, the longitude and latitude coordinates provided by the GNSS module, and the timestamp generated by the real-time clock module. The acquisition frequency is set to 1Hz, and all data is transmitted to the main control chip buffer via the I2C bus. The data is then normalized to convert the pressure value into a liquid level percentage and the vibration data into a normalized value. The system then converts the tilt data into Euler angle format and standardizes the units. The system then reads the bucket ID stored in the chip, constructs a data packet header, binds the bucket ID to a timestamp, and merges it with the standardized perception data. Based on predefined rules, the system identifies events such as tipping, collisions, and abnormal liquid levels, generates a status event marker, and appends it to the data packet. Finally, all status event data is locally cached and, depending on network conditions, uploaded to the cloud, generating a continuous, real-time perception data stream.

[0027] Step S2: Perform perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and historical barrel loading information; calculate liquid usage rate data based on the historical barrel loading information and the optimized perception data; perform comprehensive container status determination based on the historical barrel loading information and the optimized perception data to obtain currently loaded liquid type data; and generate comprehensive barrel status data based on the liquid type data and the liquid usage rate data.

[0028] In an embodiment of the present invention, the real-time sensor data stream uploaded from the receiving bucket is first verified for structural integrity, timestamp validity, and data deduplication to ensure data validity and correct timing. Preprocessing operations such as filtering, interpolation, and unit unification are performed on the verified data to improve the stability and accuracy of the liquid level, inclination, and positioning data. Subsequently, the system queries the historical loading database based on the bucket ID to retrieve information such as the corresponding liquid type, physical properties, and loading time. Combining the optimized sensor data with historical liquid parameters, the system calculates the liquid level change rate and corrects the volume change using the container shape function. Liquid density and temperature compensation factors are further combined to complete the conversion of liquid usage rate. The container status is determined by inclination, vibration, position change, and liquid level consistency. Status tags such as in transit, delivered, and abnormal change are identified, and the type of liquid currently in the bucket is inferred. Finally, the liquid level, position, rate, liquid type, and status tags are integrated to generate structured, comprehensive bucket status data, which serves as the basis for subsequent analysis.

[0029] Step S3: Perform customer usage and consumption analysis on the comprehensive bucket state data to obtain customer consumption characteristics; perform dynamic depletion time prediction based on the comprehensive bucket state data and customer consumption characteristics to obtain depletion prediction data; perform container location pattern analysis on the comprehensive bucket state data to obtain location pattern data; perform container life cycle assessment based on the location pattern data, depletion prediction data, and comprehensive bucket state data to obtain demand forecast and status assessment data;

[0030] In an embodiment of the present invention, a mapping relationship is established between the bucket ID and the customer ID, and the comprehensive bucket status data is classified according to the customer and liquid type. The liquid level and rate series of the last 90 days are extracted for each bucket, the main period is extracted through Fourier transform, the average daily consumption and usage fluctuation are calculated, and the bucket-level usage pattern is constructed. All bucket-level patterns are aggregated to generate customer-level consumption features, including the average usage rate, periodicity, stability coefficient and liquid preference vector. Combined with the current state of the liquid level and rate, LOWESS trend fitting and periodic modulation are performed to generate a predicted rate series for the next 48 hours, and the expected depletion time node is estimated by the integration method. The system also performs DBSCAN clustering from the historical location sequence to identify the permanent location and path transfer pattern, and uses the autocorrelation function to extract the periodic law to predict the future residence time and migration time. The predicted liquid level depletion time, location behavior and current status are integrated to complete the container life cycle assessment and output structured demand forecast and status assessment data.

[0031] Step S4: Perform barrel status warning and logistics optimization based on demand forecast and status assessment data to obtain an optimized path plan; perform trade scheduling and execution based on the optimized path plan to obtain optimized scheduling instructions;

[0032] In this embodiment of the present invention, demand forecast and status assessment data is used to filter out all barrels with low liquid levels, near-depletion levels, abnormal conditions, and empty barrels awaiting recycling, generating a list of warning barrels. Based on the barrel ID and liquid type, the system queries the warehouse database for current liquid inventory, number of empty barrels, ISO-TANK capacity, and delivery vehicle resources to create a resource availability table. The system aggregates warning barrels by geographic location, generates replenishment orders, matches optimal warehouses with resources, calculates the replenishment volume per barrel and the replenishment time window, and simultaneously generates recycling tasks, assigning recycling points and warehouses. The replenishment and recycling tasks are jointly modeled as a time-windowed path optimization problem, employing a genetic algorithm to output the shortest path and optimal vehicle allocation. ISO-TANK repackaging instructions are generated based on the total liquid volume required for the order and the available tank capacity, with specified repackaging times and quality inspection requirements. A processing notification is generated for all abnormal barrels, indicating the event type, recommended handling method, and priority. The system integrates all orders, routes, and operation instructions to generate standardized scheduling instruction packages, which are distributed via interfaces to the logistics and field execution systems, achieving automated closed-loop scheduling.

[0033] Preferably, step S1 includes the following steps:

[0034] Step S11: acquiring raw sensor data through the sensor array built into the smart chip shared bucket, wherein the raw sensor data includes pressure data, vibration data, tilt data, positioning data, and clock data;

[0035] Step S12: performing data standardization processing on the original sensing data to obtain standardized sensing data;

[0036] Step S13: Bind the standardized perception data with the bucket identification information to obtain bucket-related data;

[0037] Step S14: Perform status event determination on the bucket associated data to obtain status event data.

[0038] In this embodiment of the present invention, a multi-dimensional data collection system for the environment and container status is performed using a multi-type sensor array pre-embedded within the intelligent chip sharing barrel. The sensor array is fixedly installed in the barrel's bottom and sidewall structures and connected to the I / O interface of the intelligent chip main control unit. The pressure sensor, model MS5803-14BA, is installed at the center of the barrel's bottom to collect static pressure data of the liquid within the barrel. The tilt sensor, a MEMS three-axis tilt module SCL3300, is installed at the centerline of the barrel's outer shell sidewall to collect inclination changes in the barrel's X and Y axes. The vibration sensor, an ADXL345 digital accelerometer module, is installed slightly above the barrel's inner wall to detect vibration frequency and acceleration changes. The positioning module is an embedded dual-mode GNSS receiver chip that supports Beidou B1 and GPS L1 signal reception. The antenna is installed on the top shell of the barrel to ensure the real-time and coverage of positioning data. The clock module, a DS3231 real-time clock chip, is used to timestamp all collected data with high precision. Each sensor samples synchronously at a frequency of 1 Hz, with an acquisition cycle of 60 seconds. The collected data is transmitted to the internal cache of the main control chip through the I2C bus to form the original sensor data set. The data structure is in the form of a five-tuple: (P_i, A_i, θ_i, L_i, T_i), where P_i represents the pressure value, A_i represents the vibration acceleration, θ_i represents the inclination angle, L_i represents the longitude and latitude coordinates, and T_i is the timestamp.

[0039] The raw sensor data collected above is standardized. First, the static pressure value P_i obtained by the pressure sensor is converted into liquid level according to the liquid density ρ and the gravity acceleration g. The liquid level height h_i is calculated as follows: h_i=P_i / (ρ×g), where ρ is the liquid density (unit: kg / m 3 ), g is the acceleration due to gravity (take 9.8m / s 2 ). If the liquid type is diesel, then ρ=830kg / m 3. The converted liquid level height h_i is converted into a percentage with the total height H of the barrel to obtain the liquid level percentage L%=(h_i / H)×100%. The inclination data θ_i is a three-axis inclination vector, which is uniformly converted into the Euler angle representation (α, β, γ) with the unit being degrees (°). The vibration acceleration A_i data is filtered through a sliding window (the window width is 5), and its peak value a_max is calculated. Then a_max is graded and divided into vibration levels 1 to 5. The positioning data L_i is converted using the WGS-84 coordinate system, and the output format is six digits after the decimal point of longitude and latitude. All processed data are uniformly encapsulated into a standardized perception data set with the data structure: (L%, α, β, γ, a_level, lat, lon, T_i), where a_level is the vibration level, indicating the discrete grading of vibration intensity, ranging from 1 to 5; lat is the latitude, indicating the latitude coordinate of the geographic location; lon is the longitude, indicating the longitude coordinate of the geographic location.

[0040] Standardized sensor data is bound to the bucket's unique identifier. The bucket's unique identifier, ID_i, is pre-programmed into the chip's internal EEPROM. After being read by the main control chip, it is merged with the standardized sensor data acquired during the current data sampling cycle. A data packet header is constructed, containing fields such as the bucket ID, timestamp, data version number, and acquisition device ID. The packet header and standardized sensor data fields are combined to form a bucket-associated data packet. The data structure is: (ID_i, T_i, L%, α, β, γ, a_level, lat, lon, ver, dev_id), where ver is the data version number (version), which identifies the data format version; dev_id is the device ID (device ID), which identifies the device ID of the data acquisition device. This data is encoded in JSON format to accommodate subsequent transmission protocols. A separate data packet is generated for each sampling cycle, allowing for subsequent status event identification and caching.

[0041] Perform status event judgment on bucket-related data. The system sets multiple event recognition rules and uses a rule engine to execute event judgment logic. For dumping events, the rule is: when any tilt angle component (α or β) ≥ 15 and the duration Δt ≥ 10s, the dumping event flag is triggered, and the flag field is "dumping = 1"; for collision events, according to the vibration level a_level, if a_level ≥ 4 in 3 consecutive samples, and the a_max change rate Δa / Δt> 1.5m / s 3, it is considered a high-intensity collision event and marked as "collision = 1." For abnormal consumption events, the system calculates the liquid level percentage difference ΔL% between two consecutive samplings. If ΔL% > 20% and the time interval Δt < 5 minutes, it is considered abnormal liquid consumption and marked as "abnormal consumption = 1." The event determination result is appended to the original barrel's associated data using a Boolean logic tag, forming a status event data structure: (ID_i, T_i, L%, α, β, γ, a_level, lat, lon, dump, collision, abnormal consumption). This data structure is used for subsequent caching strategy determination and priority transmission task scheduling. All status event data is written to the local cache after generation, awaiting the network connection detection module to determine whether to trigger upload or queue caching.

[0042] Preferably, the perceptual optimization and loading retrieval of the real-time perceptual data stream in step S2 includes:

[0043] Receive and verify the real-time perception data stream and obtain the verified data packet;

[0044] Performing perception data preprocessing on the verified data packets to obtain optimized perception data;

[0045] Historical liquid information is retrieved based on the optimized perception data to obtain the historical loading information of the barrel.

[0046] In this embodiment of the present invention, during the "Receive and Verify Real-Time Perception Data Stream" sub-step, the cloud-side receiving module establishes a data channel with the bucket-side smart chip via the MQTT protocol, monitoring data upload requests from each bucket ID in real time. Each data transmission utilizes a standard JSON structure, containing the following fields: bucket ID (ID_i), timestamp (T_i), liquid level percentage (L%), tilt angle (α, β, γ), vibration level (a_level), geographic coordinates (lat, lon), and event flags (such as tipping, collision, etc.). The receiving module first performs a structural integrity check, using JSON Schema to define standard field formats and data types. All fields must meet type matching and non-null checks. It then verifies the validity of the timestamp, requiring the difference ΔT between the current server time and T_i in the data packet to satisfy |ΔT| ≤ 600 seconds. Otherwise, the data is marked as delayed or retransmitted. The bucket ID is then checked for validity, matching the bucket ID against the bucket registry. If ID_i does not exist or is outside the current deployment range, the packet is discarded and an exception is logged. After the packet passes verification, it enters the deduplication process. Redis cache stores the hash digests of all bucket data from the past 10 minutes and compares them with the hash value of the current packet. If the hash already exists, it is considered a duplicate and not processed. Deduplicated packets are sorted by timestamp T_i to ensure temporal consistency during subsequent processing. This ultimately creates a verified packet list structured as [Data_1, Data_2, ..., Data_n], where each Data_k is a structured and time-valid single-bucket-aware data unit.

[0047] Perform median filtering on the liquid level percentage sequence L%. Select a data point set with a sliding window length of 5 {L ,L ,L ,L ,L The median L_med is calculated as the smoothed liquid level value at the current moment to remove short-term sharp fluctuations. The inclination angle values ​​(α, β, γ) are smoothed using a first-order low-pass filter. The filter is discretely implemented in the form of a difference equation: θ_filtered(k)=θ_filtered(k−1)+(Δt / τ)×(θ_raw(k)−θ_filtered(k−1)), where τ is the time constant and θ represents any one of the three channels α, β, or γ. If the vibration level data a_level remains unchanged for five consecutive sampling periods, it is merged into a stable segment to simplify the subsequent collision recognition logic. The positioning data (lat, lon) are optimized using the Kalman filter algorithm. The observation noise covariance matrix R is set to the diagonal matrix diag(0.00001, 0.00001), and the state transition matrix F is set to a 2×2 unit matrix. Velocity estimation is introduced into the prediction process to improve positioning continuity. Each processed data packet contains the filtered level value L_filt%, smoothed tilt angles (α_filt, β_filt, γ_filt), stable vibration level a_stable, optimized geographic location (lat_opt, lon_opt), and the original timestamp T_i and bucket ID. The final output is a sequence of optimized perception data packets for subsequent fusion processing.

[0048] Using the bucket ID as the key index, execute SQL queries in the cloud-based relational database. The "liquid_history" table records all bucket loading history, including fields such as bucket ID (ID_i), loading time (T_load), liquid type (L_type), liquid density (ρ), viscosity (μ), loading volume (V), and unloading time (T_unload).

[0049] The query results return the most recent liquid loading information for the barrel. If the T_unload field is empty, it indicates that the loading has not yet ended, meaning that the current barrel is still in the liquid loading cycle. The system extracts the liquid type L_type, the liquid's physical properties (ρ, μ), the loading time T_load, and other information and encapsulates them into a data structure: {L_type_i, ρ_i, μ_i, T_load_i}. Combined with the current timestamp T_i in the optimized perception data, it calculates the duration Δt = T_i − T_load_i of the current barrel's loading cycle for subsequent liquid level change trend analysis. If the difference between the initial liquid level value in the optimized perception data and the residual liquid level at the last unloading in the historical record exceeds 10%, the system records a "suspected liquid change flag" to indicate an abnormal container status. All historical liquid information retrieval results are incorporated into the data input stack of the barrel state fusion module for use in subsequent liquid usage rate and status determination logic.

[0050] Preferably, calculating the liquid usage rate data according to the historical barrel loading information and the optimized perception data in step S2 includes:

[0051] Perform time window data screening on the optimized perception data to obtain the period liquid level data;

[0052] Calculate the liquid level change rate data based on the period liquid level data;

[0053] Performing container shape correction on the liquid level change rate data according to the container shape parameters in the barrel historical loading information to obtain the shape correction change rate;

[0054] Performing liquid characteristic compensation on the shape correction change rate to obtain the characteristic compensation change rate;

[0055] Calculate the basic usage speed based on the characteristic compensation change rate;

[0056] Identify abnormal pattern markers based on base usage velocity;

[0057] Generates fluid usage rate data based on base usage rates and abnormal pattern markers.

[0058] In this embodiment of the present invention, a fixed time window length Δt_window is set to 24 hours. The system filters out the historical liquid level data sequence corresponding to the current bucket ID from the optimized perception data. The data filtering condition is that the timestamp T_i satisfies: T_now−T_i≤Δt_window, where T_now is the current system time. The filtered liquid level data is arranged in ascending timestamp order to form a period liquid level data sequence: , where T_i is the timestamp of time i, and L_i is the corresponding liquid level percentage. To prevent sparse or uneven sampling from interfering with the results, the time interval ΔT_i between two adjacent time points must satisfy ΔT_i ≤ 30 minutes. Any data segment with an interval exceeding this limit is discarded. This ultimately forms a continuous, evenly spaced liquid level data segment as the input sequence.

[0059] The liquid level change rate is calculated using the adjacent time difference method, and the liquid level change rate R_i is defined as: R_i=(L_i-L_{i-1}) / (T_i-T_{i-1}), where L_i and L_{i-1} are the liquid level percentages of two consecutive sampling points, and T_i and T_{i-1} are the corresponding timestamps in hours. The unit of this change rate is % / h, which represents the liquid level change rate per hour. To improve stability, a sliding average process (window length is 3) is applied to the sequence composed of all R_i to generate a smooth change rate sequence R_smooth={ At the same time, the original change rate and smoothed change rate of each time period are recorded for subsequent model correction.

[0060] The current barrel's container parameters are extracted from the barrel's historical loading information, including the barrel's total height H, the barrel's cross-sectional structure type (cylindrical / elliptical / rectangular), and the corresponding parameters. For cylindrical containers, the volume change is linearly related to the liquid level. For elliptical or irregular shapes, a nonlinear conversion using the volume function V(h) is required. For an elliptical cylindrical container, for example, its volume function is: V(h) = π × a × b × h, where a is the major axis radius of the ellipse, b is the minor axis radius, and h is the liquid level height (unit: meters).

[0061] The modified expression for converting the liquid level percentage change ΔL_i to the volume change ΔV_i is: ΔV_i=V(H×L_i / 100)−V(H×L_{i-1} / 100);

[0062] The final calculation is the rate of volume change per unit time: R_v(i) = ΔV_i / ΔT_i, in L / h. All ΔV_i and ΔT_i are measured in real physical units to ensure that the rate calculation reflects the actual liquid changes.

[0063] Consider the effects of the liquid's temperature expansion coefficient, density change, and viscosity on measurement accuracy. Assume the liquid density is ρ, the ambient temperature is T_env, and the liquid expansion coefficient is σ_T (unit: 1 / °C). The relationship between liquid density and temperature is: ρ_T = ρ_0 × (1-σ_T × (T_env-T_0)), where ρ_0 is the liquid's density at a reference temperature T_0 (e.g., 20°C).

[0064] The volume change rate after compensation is adjusted to: R_v (i)=R_v(i)×(ρ_T / ρ_0), this compensation is used to correct the liquid level measurement error caused by temperature change.

[0065] In addition, if the liquid is a high viscosity substance (such as lubricating oil), the system will introduce a hysteresis compensation factor λ_v=0.95~1.05, and the correction formula is: R_v (i)=R_v (i)×λ_v, and finally the rate of change sequence R_final={R ,R ,...,R_n }.

[0066] Select N valid data points in the past 24 hours and calculate the average liquid usage rate V_avg during this period: V_avg=(1 / N)×∑_ R_v (i), unit is L / h.

[0067] And calculate the standard deviation : = , if the standard deviation , indicating that the liquid usage rate is stable. Otherwise, further analysis of the abnormal fluctuations is required. This V_avg is used as the basic usage rate to determine the customer's current liquid consumption level.

[0068] The system sets a variety of abnormality judgment rules. The first type is a "sudden increase" event. If there is any of the three consecutive time points , marked as "accelerated consumption"; the second type is the "stagnation" event, if three consecutive time points =0 and the liquid level value remains unchanged, it is marked as "use stagnation"; the third type is the "reverse flow" event. <0, and the trend lasts for more than 2 hours, it is marked as "liquid level rises"; all abnormal marks are appended to the rate data sequence with Boolean values ​​to form the structure: {timestamp T_i, rate The system constructs a structured rate result package containing the following fields: bucket ID, current time, average usage rate V_avg during the calculation period, standard deviation of fluctuation σ_v, primary anomaly type (such as acceleration, stagnation, and rebound), anomaly occurrence time set, and unit volume consumption curve. This data serves as input for subsequent customer behavior modeling and exhaustion time prediction. All structured rate data is uploaded to the bucket data fusion center and stored in the "usage_rate_dataset" table for use by the periodic behavior analysis module.

[0069] Preferably, the comprehensive determination of the container status according to the historical loading information of the bucket and the optimized perception data in step S2 includes:

[0070] Extract the bucket's tilt and vibration data from the optimized perception data;

[0071] The bucket status is determined based on the inclination angle and vibration data to obtain the bucket status mark;

[0072] Perform position change analysis based on the optimized perception data to obtain position change data;

[0073] Determine the transport status based on the position change data and the barrel status mark to obtain the transport status mark;

[0074] Performing a liquid level consistency check on the optimized sensing data to obtain liquid level consistency data;

[0075] Perform liquid replacement detection based on the liquid level consistency data and the transportation status mark to obtain a liquid replacement mark;

[0076] Generate delivery status mark based on location change data;

[0077] Generate a container status determination result according to the barrel status mark, the transportation status mark, the liquid replacement mark, and the delivery status mark;

[0078] The type data of the liquid currently loaded is determined according to the container status determination result.

[0079] In the embodiment of the present invention, the tilt angle vector (α_i, β_i, γ_i) and the vibration level a_level_i at each sampling moment are extracted from the standardized optimized perception data. The tilt angle unit is °, and the vibration level is Tilt data comes from a three-axis tilt sensor, and vibration data comes from the stable values ​​of the acceleration sensor after sliding window processing. The system constructs a continuous tilt sequence Θ_t={α_t,β_t} and a vibration sequence A_t={a_level_t} for each bucket for state recognition and analysis.

[0080] Set fixed logic thresholds for state recognition: when |α| or |β| ≥ 15° at any moment and the state lasts for more than 10 minutes, mark the current bucket as "tilted state"; if a_level ≥ 4 in three consecutive sampling periods and the vibration acceleration change rate Δa / Δt > 1.5m / s 3 , it is marked as "collision state"; if both tilt and vibration are abnormal and last for more than 30 minutes, it is marked as "abnormal physics state". Bucket states are stored as Boolean flags in a structure with the format: {bucket ID, T_i, tilt = 1 / 0, collision = 1 / 0, abnormal physics = 1 / 0}.

[0081] From the continuous positioning data of each bucket ( , ) to extract the geographic trajectory sequence. The Haversine formula is used to calculate the spatial distance d_i between any two adjacent points: d_i=2×R×arcsin( ), where R is the radius of the Earth (6371 km), and In radians. If it occurs continuously at multiple sampling points, it is marked as a "position change event". The system calculates the total displacement per unit time With average moving speed ,like If the position is stable within ±50m of the same coordinate point for 3 consecutive days, it is determined to be in "transportation state". The position change data is output as a structure, including: {bucket ID, Δd series, , status = transport / stationary}.

[0082] Combine the dumping / collision state with the high displacement state to construct a binary state matrix ,in Indicates the simultaneous presence of displacement and tilt, defined as a typical transport state; Indicates the presence of displacement but no tilt, defined as “slight movement”; It means there is no displacement but tilt, which is defined as "abnormal vibration"; It is in a static state. Use the logical judgment function: transport state = 1 if and only if =1. Other status values ​​are assigned as 0 or abnormal status. The transport status tag structure is: {bucket ID, , transport status = 1 / 0}.

[0083] From the level sequence Calculate the rate of change , and combined with the liquid type density recorded in the historical loading and viscosity If the current level change pattern deviates from the historical typical liquid consumption curve by more than ±30%, or the initial level value differs from the historical unloading residual value by more than 10%, a level abnormality flag will be generated. The system establishes a standard level change curve template library for each liquid type. Corresponding to a set of rate curves , calculate the current rate curve Mean square error with the template: When MSE > the threshold (set to 0.05), it is determined to be a liquid level anomaly. The output structure is: {bucket ID, liquid level consistency = 1 / 0, MSE value}.

[0084] A joint decision logic is used: if the following three conditions are met simultaneously, it is considered a liquid replacement event: (1) transport status = 1; (2) liquid level consistency = 0; (3) the liquid level value rises > 30% after transportation. The rule function F(transport, volume_rise, MSE) is used: the replacement flag = 1 when transport = 1 and ΔL > 30% and MSE > 0.05; otherwise, the replacement flag = 0. The final output structure is: {tank ID, replacement flag = 1 / 0, detection time T_i}.

[0085] Set the delivery judgment rule: If the current position ( ) and the customer registration location coordinates ( ) ≤0.1km, and the initial value of the liquid level If the liquid level is >90%, it is considered "delivered"; if the liquid level is below 10% and the location is stable for more than 7 days, it is marked as "pending collection"; if the liquid level is moderate and the location is not in a customer area, it is marked as "in transit". The delivery status is marked as an enumeration type: {delivered, pending collection, in transit}.

[0086] Constructing a four-dimensional state vector , each component takes a value of 0 or 1, and is classified by the state decision tree model. The model is based on the following rules: =1 and =1, the status is "replacement in transit"; if =1 and =0, it is "quiescent abnormality"; if =0 and = "delivered", then it is "normal use"; if = "to be recycled" and L% < 10%, then it is "empty barrel to be recycled"; if = 1 but not transported, it is "illegal replacement". The status judgment result is a single tag with the structure: {bucket ID, T_i, current status = }.

[0087] If the status tag contains "normal use" or "replacement during transportation", the system uses the most recent historical loading record to determine the liquid type; if the status is "illegal replacement" or "liquid level abnormality", the system re-determines the liquid type. The liquid type re-determination uses a liquid level change model and a matching method of the liquid's physical characteristics. The system extracts the typical usage rate range of each liquid from the liquid database. and viscosity Relationship model, calculate the current liquid level change rate t, find the Include Liquid collection If there is only one liquid in the set, assign a value directly; if there are multiple candidates, sort them according to the historical customer liquid preference priority. The final output liquid type data structure: {bucket ID, liquid type = , confidence = , judgment basis = history / inference}.

[0088] Preferably, performing customer usage and consumption analysis on the comprehensive bucket data in step S3 includes:

[0089] Classify and associate the comprehensive bucket data to obtain classified associated data;

[0090] Obtain historical usage data for each bucket; extract historical usage patterns from the historical usage data;

[0091] Analyze customer consumption characteristics based on historical usage patterns and classified association data.

[0092] In an embodiment of the present invention, a structured comprehensive bucket state data set is received, and each data contains the following fields: bucket ID (ID_i), timestamp (T_i), liquid level percentage (L_i), current liquid type (L_type_i), real-time usage rate (V_i), geographic location (lat_i, lon_i) and container status mark. The system first associates the bucket ID with the bucket-customer binding table in the customer database to obtain the customer code (CUST_j) corresponding to the bucket ID. It then creates a bucket-customer mapping table (MAP): MAP = {ID_i → CUST_j}. The bucket-state data is then aggregated by customer dimension, forming a customer-bucket dataset: D_j = {data(ID_i), ∀ID_i∈CUST_j}. During the aggregation process, the system groups the liquid type field L_type_i for each bucket to construct a customer-liquid type combination dimension. The final output is the categorical association data structure: {customer code CUST_j, liquid type L_type_k, bucket set {ID_1, ID_2, ..., ID_n}, and each bucket's corresponding liquid level and rate series}. This structure serves as the basic input for subsequent customer behavior modeling.

[0093] The liquid level percentage series L(t) and usage rate series V(t) of the past 90 days are retrieved from the bucket database by bucket ID, with a data time granularity of one record per hour. Data segments with a continuous missing time period of more than 6 hours are marked as "data unavailable" and are not included in the statistics. The periodic pattern of the usage data of each bucket is extracted, and the liquid level change cycle is analyzed using Fast Fourier Transform (FFT). Specifically: , perform a discrete Fourier transform: ,in is the frequency index, is the sequence length. Calculate the frequency domain amplitude spectrum , select the main frequency As the main use cycle, its countdown Indicates the average usage period (in days). At the same time, the average daily usage of the level sequence is calculated. , the unit is If the cycle amplitude ratio , identified as having periodic characteristics. The system records each bucket's usage period T_i, average daily consumption D_avg_i, and weekly standard deviation σ_i, forming a bucket-level usage pattern set: Pattern_i = {T_i, D_avg_i, σ_i}. All bucket usage patterns are attributed to the corresponding customer, forming a customer-level bucket pattern pool.

[0094] Aggregate the usage patterns of all buckets using customer-liquid type as the index dimension. Use liquid type The corresponding bucket set is , then the average consumption rate of the liquid by the customer is The calculation formula is as follows: ,in For the Average daily consumption of barrels. Customer consumption stability Defined as: ,in represents the standard deviation of the consumption of each barrel, is the average consumption; , the larger the value, the more stable the consumption pattern. and If the change range is within ±1 day, it is judged as a "high stability customer"; if , is determined to be a "medium stability customer"; if , then it is a "low stability customer". The system also calculates the customer's liquid type usage preference vector ,in For customers who have purchased liquid types in the last 90 days of total usage, meeting The preference vector is used to assist in subsequent liquid type inference and replenishment planning.

[0095] The final output customer consumption feature data structure is: , liquid type , average consumption rate , consumption cycle , consumption stability , liquid usage preference }, and stored in " ” table for subsequent dynamic exhaustion time prediction and scheduling optimization module to call.

[0096] Preferably, performing dynamic exhaustion time prediction based on comprehensive bucket data and customer consumption characteristics in step S3 includes:

[0097] Extract current liquid level data and usage speed sequence from comprehensive barrel status data;

[0098] Perform short-term trend analysis on the speed series to obtain short-term trend data;

[0099] Perform customer feature matching on the current liquid level data and the customer consumption features to obtain customer matching features;

[0100] Adjust the cycle factors based on short-term trend data and customer matching characteristics to obtain the cycle adjustment speed;

[0101] Perform comprehensive rate prediction based on the periodic adjustment speed to obtain a predicted rate sequence;

[0102] Calculate the exhaustion time based on the predicted rate sequence and the current liquid level data to obtain the time node prediction data;

[0103] The current liquid level data, predicted rate series and time node prediction data are integrated to obtain the depletion prediction data.

[0104] In the embodiment of the present invention, the latest comprehensive data packet of the current target bucket is first read and the liquid level percentage field is extracted. (Unit: %) and historical usage rate time series ,in The hourly liquid volume consumption rate in the last 72 hours, in L / h. The data structure is defined as: , the corresponding time point is , where n≤72. To ensure data continuity, the system removes data segments with missing points exceeding 3 hours and performs linear interpolation on discontinuous data. The interpolation function is: ,After the processing is completed, the complete speed sequence and the current ,liquid level value are obtained as the input for short-term trend ,analysis.

[0105] The locally weighted regression (LOWESS) algorithm was used to Perform trend fitting. Assume that the fitting window length is k = 12 hours, and the weight function is defined as: ,in is the current time, is the window standard deviation. The speed trend line is fitted using the weighted least squares method. , calculate its first-order derivative Indicates the speed change trend. , it is determined to be consumption acceleration; if , it is determined to be consumption slowdown; if , it is a steady consumption. The system records the trend slope and fluctuation range , output as short-term trend data. The structure is: .

[0106] Extract the average usage rate R_avg_cust (in L / h) of the customer for the current liquid type from the customer consumption profile, and the stability index , periodic parameters (Unit is day). The current usage rate of the bucket and Perform ratio calculations: , and calculate the trend slope matching coefficient at the same time: ,like and , indicating that the current consumption status of the bucket is highly consistent with the customer's historical consumption characteristics, and the matching level is "high"; if , then it is "medium"; otherwise it is "low". The system outputs the customer matching feature structure: { Match Level}.

[0107] If the matching level is "high" or "medium", the historical period is used Make period adjustments. Construct a period modulation function: ,in is the fluctuation amplitude coefficient, set to , is the phase shift. and Perform point-by-point multiplication to obtain the periodically adjusted speed sequence: If the match level is "low", it will not be applied , use directly As .

[0108] Generates a series of forecast rates for the next 48 hours , Hours. The forecast method uses the exponential weighting method: ,in is the weight coefficient, ranging from [0.6, 0.9]. Automatic adjustment, the higher the stability, The larger the value, the higher the predicted speed sequence is: , the unit is L / h.

[0109] The current liquid level percentage Convert to Volume , calculated as: ,in is the total capacity of the bucket (in L). Then integrate the predicted rate sequence to calculate the cumulative consumption ,in =1h. Find the minimum satisfy ,but Estimated time to run out (Unit is hours), the exhaustion time node is: .

[0110] The information of each part is structured and output to form a depletion prediction data packet with the following structure: {bucket ID, current liquid level percentage , the remaining volume , the predicted rate series for the next 48 hours , estimated time to run out t, prediction model parameter , prediction confidence level (high / medium / low)}, the data is stored in " ” table and pushed to the scheduling system for generating replenishment and recovery plans in advance.

[0111] Preferably, the container position pattern analysis of the comprehensive bucket state data in step S3 includes:

[0112] Extracting position history sequences from comprehensive bucket data;

[0113] Perform location cluster analysis on the location history sequence to obtain a location cluster set;

[0114] Calculate the length of stay based on the location cluster set;

[0115] Extracting movement path patterns from position history sequences;

[0116] Identify periodic patterns based on movement path patterns and dwell time to obtain position periodic patterns;

[0117] Predict the dwell time according to the position cycle law to obtain the dwell prediction result;

[0118] Generate location pattern data based on the dwell prediction results.

[0119] In the embodiment of the present invention, the field is extracted from the comprehensive bucket state data of the specified bucket ID , forming a position history sequence: ,in and is the longitude and latitude in the WGS-84 coordinate system, in degrees. The time stamp is in Unix time (seconds). The system requires that the time interval between each data point does not exceed 1 hour. If the interval exceeds, a linear interpolation point is inserted. To improve the position accuracy, the system applies a Kalman filter to each position point for trajectory smoothing. The state transition model is set to a two-dimensional position + velocity state vector The prediction-update cycle is 1 hour, and the filtering equation is implemented according to the standard Kalman formula. After processing, a continuous and smooth geographic trajectory sequence is obtained, which serves as the basis for subsequent clustering and path analysis.

[0120] Use DBSCAN density clustering algorithm to spatially partition the location point set. Set DBSCAN parameters , , use the Haversine formula to calculate the spherical distance between any two points ,in The radius of the earth is 6371 km, and all distance calculations are unified in kilometers. The clustering results are output as location cluster sets: , each cluster Contains a set of spatially adjacent coordinate points, which are considered as the container's residence behavior at a certain location. The system calculates the centroid coordinates for each cluster Clustering time range , forming a location cluster feature table, the fields include: cluster number, location centroid, first entry time, last stay time, number of points, cluster duration, etc.

[0121] Traverse the cluster set in chronological order and count the continuous residence time after entering a cluster center point (i.e., entering a residence location). The residence time is defined as: ,in To enter the cluster The first time point, The last time point when the container leaves the cluster area. If the container enters the same cluster range again, it is considered a new residence event. The system records the number of residence events for each cluster point. , average dwell time , Standard deviation of dwell time , and build a dwell time series: , providing a time reference for subsequent cycle analysis.

[0122] Constructing a movement path map based on location time series .in Represents a set of cluster center points, each node corresponds to a residence location; Represents the actual movement path of the container between two locations, which is defined as the path segment formed from exiting from one cluster point to entering the next cluster point. Included fields: Start point ,end , path occurrence time , path time , path distance Path time , the path distance is obtained by accumulating the Haversine distance between trajectory points. The system constructs the directed path transfer probability matrix ,in: , Indicates that from clustering Move to The number of historical times, General flow paths and spatial transition probabilities for modeling containers.

[0123] The residence time series of each cluster point Perform periodicity analysis. Use the autocorrelation function (ACF) method to detect the periodic structure of the time series: ,in is the time lag, and is the mean and standard deviation of the residence time of the cluster point. If a significant peak appears at (ACF value ≥ 0.6), it is determined that there is a periodic dwell behavior, and the period length is (Unit is day). The system records the periodic residence pattern of each cluster point: {cluster number , cycle length , periodic intensity ACF( If the cycle strength is ≥ 0.8, it is marked as a “strong cycle point” and used for future position prediction.

[0124] Based on the most recent entry into the cluster Time , cycle length , predict the next departure time for: , and predict the next entry time for: , the system sets the prediction time window to ± , forming a prediction interval. If the prediction point is a strong period point, the confidence is marked as "high", otherwise it is "medium" or "low". The prediction result structure is as follows: {bucket ID, cluster point , current residency start time , predicted departure time , predict the next entry time , prediction confidence level}, integrating the periodicity of all cluster points, path transfer probability and future residence prediction time to form a complete container location behavior model. The output structure is: {bucket ID, permanent point list , periodic parameters per point , path transfer matrix , the next 48 hours position prediction sequence {( , , status)}, confidence level} The location pattern data is provided to the scheduling module as the container behavior insight result for recycling time planning, delivery path optimization and container idle analysis.

[0125] Preferably, step S4 includes:

[0126] Step S41: Perform demand warning screening on the demand forecast and status assessment data to obtain a warning bucket list;

[0127] Step S42: query the inventory resources according to the warning bucket list to obtain the resource availability status;

[0128] Step S43: Generate replenishment order data based on resource availability and the warning bucket list;

[0129] Step S44: generating bucket recycling plan data according to the resource availability status and the warning bucket list;

[0130] Step S45: Optimizing the logistics route based on the barrel recycling plan data and the replenishment order data to obtain an optimized route plan;

[0131] Step S46: Generate ISO-TANK subpackaging instructions based on the replenishment order data and resource availability to obtain subpackaging operation instructions;

[0132] Step S47: Generate an exception handling notification for the warning bucket list to obtain an exception handling notification;

[0133] Step S48: Integrate and distribute the scheduling instructions for the exception handling notification, the subpackaging operation instructions, and the optimized path plan to obtain the optimized scheduling instructions.

[0134] In the embodiment of the present invention, the exhaustion time of all buckets is extracted from the "demand forecast and status assessment" database , Current liquid level percentage , container status label Status, prediction confidence level. Set the low liquid level warning threshold to =20%, the exhaustion time warning window is = 72 hours. The system performs screening logic on all buckets: If or , it is marked as "Nearly Exhausted"; if the Status contains tags such as "Abnormal Use," "Collision," or "Illegal Replacement," it is marked as "Needs Inspection"; if the barrel is empty and has been stationary for more than 7 days, it is marked as "Pending Recycling." Each barrel is categorized into a corresponding alert type, generating a structured alert barrel list: {Bucket ID, Customer ID, Location (lat, lon), Liquid Type, Alert Type, Predicted Exhaustion Time, Current Liquid Level, Status Tag}.

[0135] Get the current available inventory of each liquid type in each warehouse from the warehouse database , and query the inventory quantity of empty barrels at the same time , Available ISO-TANK storage capacity (tank ID), filling time for each type of liquid , available delivery vehicle resources The system uses liquid type as the query dimension to construct a resource status table: {liquid type, warehouse ID, inventory quantity, number of empty tanks, ISO-TANK tank ID, available tank quantity, vehicle ID, vehicle service area, available time window}, forming a resource availability status dataset as input for order generation and scheduling optimization.

[0136] Clustered by liquid type and customer location, the warning barrels with similar geographical locations and the same liquid are grouped into replenishment task orders. For each task, the system selects the warehouse closest to the customer and with inventory that meets the demand as the supplier and calculates the required liquid volume. ,in = Barrel capacity × (1 - current liquid level%). Generate a replenishment order structure: {order ID, liquid type, supply warehouse ID, target barrel ID list, replenishment quantity per barrel, customer location information, target replenishment time window, priority}. The system sorts orders by predicted exhaustion time, prioritizing urgent orders. Each order is associated with a unique order number for subsequent scheduling and tracking.

[0137] Filter out all buckets in the "empty bucket", "abnormal state", "static overdue" and other states. Calculate the distance matrix based on the current location information of the bucket and the warehouse distribution. , select the shortest distance warehouse as the recycling destination, and generate a recycling task: {recycling task ID, bucket ID, current location, recycling warehouse ID, recycling reason (empty bucket / inspection / abnormal), recycling priority, recycling time window}. The system matches the recycling task with the replenishment task in the geographical dimension, and subsequently merges and optimizes the path.

[0138] Construct a vehicle routing problem model with time windows (VRPTW). Define the node set N = {customer bucket point, warehouse point}, with edge weights representing the travel time or distance between two points. Constraints include: each route starts and ends at a warehouse, the total route load ≤ vehicle capacity, and the service time at each node meets the predicted exhaustion or reservation window. The route optimization objective is to minimize the total route length or the number of vehicles used. A genetic algorithm or ant colony algorithm is used for route search, outputting a route plan structure: {route ID, vehicle ID, starting warehouse, sequence of visited nodes (bucket ID or warehouse ID), estimated arrival time sequence, total route length, and task type (replenishment / recovery / mixed)}.

[0139] The required liquid volume for each replenishment order is summarized and matched against available ISO-TANK resources. Tanks with sufficient remaining capacity and that have passed quality inspection are selected. A repackaging instruction is generated: {repackaging task ID, ISO-TANK ID, liquid type, total repackaging volume, target customer ID or bucket ID list, repackaging start time, repackaging end time, repackaging person in charge, quality inspection requirements}. The system considers repackaging parallelization capacity and tank switching time to optimize the allocation sequence.

[0140] Extract all buckets with status tags such as "collision," "abnormal consumption," "dumping," and "illegal replacement." Analyze the time, frequency, and current status of these events. Combined with customer level and historical fault records, prioritize handling. Generate an exception handling notification: {notification ID, bucket ID, event type, event time, recommended handling method (on-site inspection / remote diagnosis / recycling and replacement), recommended handling time window, responsible personnel, and handling priority level}. This notification data is pushed to the operations and maintenance management platform, generating a handling task.

[0141] The three types of tasks are aggregated in chronological order and geographical location to generate a complete scheduling task package. Each task package includes: a list of tasks to be executed (replenishment, recycling, anomaly inspection); route planning and schedule; required resources (vehicle ID, tank ID, operator ID); operation nodes and confirmation mechanisms (code scanning confirmation, liquid level verification, GPS lock). The system encodes the task package into a unified format (such as XML or JSON structure) and distributes it to the logistics scheduling platform, warehousing system and operation and maintenance terminal through the API interface to form a closed-loop execution mechanism. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and scope of the equivalent elements of the application documents are included in the present invention.

[0142] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A sensor data processing method for liquid packaging barrels, applied to smart chip shared barrels, characterized in that: The smart chip sharing barrel has a pre-embedded sensor array inside. The sensor array includes a pressure sensor, a tilt sensor, a vibration sensor, a GNSS positioning module, and a real-time clock chip fixedly installed in the bottom and side wall structures of the barrel body. The sensor data processing method of the liquid packaging barrel includes the following steps: Step S1: Obtain the original sensor data of the smart chip shared bucket; perform bucket status judgment on the original sensor data to obtain status event data; locally cache and transmit the status event data to obtain a real-time perception data stream; Step S2: Perform perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and historical barrel loading information; calculate liquid usage rate data based on the historical barrel loading information and the optimized perception data; perform comprehensive container status determination based on the historical barrel loading information and the optimized perception data to obtain currently loaded liquid type data; and generate comprehensive barrel status data based on the liquid type data and the liquid usage rate data. Step S3: Perform customer usage and consumption analysis on the comprehensive bucket state data to obtain customer consumption characteristics; perform dynamic depletion time prediction based on the comprehensive bucket state data and customer consumption characteristics to obtain depletion prediction data; perform container location pattern analysis on the comprehensive bucket state data to obtain location pattern data; perform container life cycle assessment based on the location pattern data, depletion prediction data, and comprehensive bucket state data to obtain demand forecast and status assessment data; Step S4: Perform barrel status warning and logistics optimization based on demand forecast and status assessment data to obtain an optimized path plan; perform trade scheduling and execution based on the optimized path plan to obtain optimized scheduling instructions.

2. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: acquiring raw sensor data through the sensor array built into the smart chip shared bucket, wherein the raw sensor data includes pressure data, vibration data, tilt data, positioning data, and clock data; Step S12: performing data standardization processing on the original sensing data to obtain standardized sensing data; Step S13: Bind the standardized perception data with the bucket identification information to obtain bucket-related data; Step S14: Perform status event determination on the bucket associated data to obtain status event data.

3. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: The perception optimization and loading retrieval of the real-time perception data stream in step S2 includes: Receive and verify the real-time perception data stream and obtain the verified data packet; Performing perception data preprocessing on the verified data packets to obtain optimized perception data; Historical liquid information is retrieved based on the optimized perception data to obtain the historical loading information of the barrel.

4. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: Calculating the liquid usage rate data based on the historical barrel loading information and the optimized perception data in step S2 includes: Perform time window data screening on the optimized perception data to obtain the period liquid level data; Calculate the liquid level change rate data based on the period liquid level data; Performing container shape correction on the liquid level change rate data according to the container shape parameters in the barrel historical loading information to obtain the shape correction change rate; Performing liquid characteristic compensation on the shape correction change rate to obtain the characteristic compensation change rate; Calculate the basic usage speed based on the characteristic compensation change rate; Identify abnormal pattern markers based on base usage velocity; Generates fluid usage rate data based on base usage rates and abnormal pattern markers.

5. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: In step S2, the comprehensive determination of the container status based on the historical loading information of the bucket and the optimized perception data includes: Extract the bucket's tilt and vibration data from the optimized perception data; The bucket status is determined based on the inclination angle and vibration data to obtain the bucket status mark; Perform position change analysis based on the optimized perception data to obtain position change data; Determine the transport status based on the position change data and the barrel status mark to obtain the transport status mark; Performing a liquid level consistency check on the optimized sensing data to obtain liquid level consistency data; Perform liquid replacement detection based on the liquid level consistency data and the transportation status mark to obtain a liquid replacement mark; Generate delivery status mark based on location change data; Generate a container status determination result according to the barrel status mark, the transportation status mark, the liquid replacement mark, and the delivery status mark; The type data of the liquid currently loaded is determined according to the container status determination result.

6. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: In step S3, the customer's usage and consumption analysis of the comprehensive bucket data includes: Classify and associate the comprehensive bucket data to obtain classified associated data; Obtain historical usage data for each bucket; extract historical usage patterns from the historical usage data; Analyze customer consumption characteristics based on historical usage patterns and classified association data.

7. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: In step S3, dynamic exhaustion time prediction based on comprehensive bucket data and customer consumption characteristics includes: Extract current liquid level data and usage speed sequence from comprehensive barrel status data; Perform short-term trend analysis on the speed series to obtain short-term trend data; Perform customer feature matching on the current liquid level data and the customer consumption features to obtain customer matching features; Adjust the cycle factors based on short-term trend data and customer matching characteristics to obtain the cycle adjustment speed; Perform comprehensive rate prediction based on the periodic adjustment speed to obtain a predicted rate sequence; Calculate the exhaustion time based on the predicted rate sequence and the current liquid level data to obtain the time node prediction data; The current liquid level data, predicted rate series and time node prediction data are integrated to obtain the depletion prediction data.

8. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: The container position pattern analysis of the comprehensive bucket status data in step S3 includes: Extracting position history sequences from comprehensive bucket data; Perform location cluster analysis on the location history sequence to obtain a location cluster set; Calculate the length of stay based on the location cluster set; Extracting movement path patterns from position history sequences; Identify periodic patterns based on movement path patterns and dwell time to obtain position periodic patterns; Predict the dwell time according to the position cycle law to obtain the dwell prediction result; Generate location pattern data based on the dwell prediction results.

9. The sensor data processing method for liquid packaging barrels according to claim 1, characterized in that: Step S4 includes: Step S41: Perform demand warning screening on the demand forecast and status assessment data to obtain a warning bucket list; Step S42: query the inventory resources according to the warning bucket list to obtain the resource availability status; Step S43: Generate replenishment order data based on resource availability and the warning bucket list; Step S44: generating bucket recycling plan data according to the resource availability status and the warning bucket list; Step S45: Optimizing the logistics route based on the barrel recycling plan data and the replenishment order data to obtain an optimized route plan; Step S46: Generate ISO-TANK subpackaging instructions based on the replenishment order data and resource availability to obtain subpackaging operation instructions; Step S47: Generate an exception handling notification for the warning bucket list to obtain an exception handling notification; Step S48: Integrate and distribute the scheduling instructions for the exception handling notification, the subpackaging operation instructions, and the optimized path plan to obtain the optimized scheduling instructions.

10. A sensor data processing system for a liquid packaging barrel, characterized in that: A sensor data processing system for a liquid packaging barrel is configured to execute the sensor data processing method for a liquid packaging barrel according to claim 1, wherein the sensor data processing system for the liquid packaging barrel comprises: The bucket-side status perception module is used to obtain the original sensor data of the smart chip shared bucket; determine the bucket status of the original sensor data to obtain status event data; and locally cache and transmit the status event data to obtain a real-time perception data stream. The comprehensive barrel state fusion module is used to perform perception optimization and loading retrieval on the real-time perception data stream to obtain optimized perception data and historical barrel loading information; calculate liquid usage rate data based on the historical barrel loading information and optimized perception data; perform comprehensive container status judgment based on the historical barrel loading information and optimized perception data to obtain the current loaded liquid type data; and generate comprehensive barrel state data based on the liquid type data and liquid usage rate data. The cycle behavior prediction module is used to analyze customer usage and consumption based on the comprehensive bucket data to obtain customer consumption characteristics; dynamically predict depletion time based on the comprehensive bucket data and customer consumption characteristics to obtain depletion prediction data; analyze container location patterns based on the comprehensive bucket data to obtain location pattern data; and perform container life cycle assessment based on location pattern data, depletion prediction data, and comprehensive bucket data to obtain demand forecast and status assessment data. The intelligent trade scheduling module is used to perform barrel status warning and logistics optimization based on demand forecast and status assessment data to obtain an optimized path plan; trade scheduling and execution are carried out according to the optimized path plan to obtain optimized scheduling instructions.

Images