Container transportation state monitoring method based on multi-parameter perception

Abstract

The embodiment of the invention provides a container transportation state monitoring method based on multi-parameter perception, which is applied to the technical field of logistics monitoring and comprises the following steps: by taking a Hall switch trigger signal as a time reference and a logic starting point, acquiring acceleration, attitude angle, position and temperature and humidity signals in a time window before and after triggering; performing time sequence correlation analysis on the acceleration, the attitude angle and the position signal to generate initial event judgment; when the initial event is determined to be physical door opening or hoisting start, extracting an impact decay time constant and an attitude angle deviation retention amount, and constructing a two-dimensional feature space to perform mechanical impact classification; asynchronously calling a temperature and humidity signal according to a judgment result for root verification; and generating a joint event code and executing differentiated resource control. According to the invention, triggering events with different properties can be accurately distinguished, risk grading of mechanical impact and source tracing of environmental abnormity are realized, terminal endurance time is obviously prolonged, and monitoring reliability is ensured.

Description

Technical Field

[0001] This application relates to the field of logistics monitoring technology, and in particular to a method for monitoring the status of container transportation based on multi-parameter sensing. Background Technology

[0002] With the development of global trade, the transportation scope of freight containers covers the entire world. A single logistics entity can operate millions of containers, distributed across all continents and major trade routes. During container transportation, it is necessary to monitor their location, status, and environmental parameters in real time to ensure cargo safety and transportation efficiency.

[0003] Currently, container monitoring equipment typically operates globally with the containers, moving with them after installation. The average return time to the maintenance point is very long, and on some routes, the equipment may not return to the domestic market for extended periods, making on-site maintenance virtually impossible. Therefore, monitoring equipment must possess high reliability and autonomous judgment capabilities, accurately identifying various abnormal events during transportation without human intervention.

[0004] Mobile communication networks vary across different regions, and international roaming protocols are complex. Equipment needs to maintain stable connections under various network environments and frequency switching conditions. At the same time, transportation routes cover a variety of extreme environments, including equatorial high temperatures, Arctic cold, humid islands, and inland temperature variations, which places higher demands on the adaptability and monitoring accuracy of the equipment.

[0005] In existing technologies, container monitoring methods mostly rely on single sensor triggers or manual inspections. Under globalized, multi-environmental, and low-maintenance conditions, it is difficult to accurately distinguish between different events such as physical door opening, hoisting, collision, drop, and water leakage. This can easily lead to false alarms or missed alarms, and cannot meet the intelligent management needs of large-scale global container transportation. Summary of the Invention

[0006] This application provides a container transportation status monitoring method based on multi-parameter perception, which enables accurate classification, risk grading, and root cause tracing of container transportation events. It also dynamically allocates resources based on event levels, reducing false alarm rates and improving endurance. To achieve the above objectives, this application adopts the following technical solution: A method for monitoring the status of container transportation based on multi-parameter sensing, the method comprising: Using the Hall switch trigger signal as the time reference and logic starting point, the acceleration signal, attitude angle signal, position signal and temperature and humidity signal are collected within the time window before and after the trigger moment; A time-series correlation analysis is performed on the acceleration signal, the attitude angle signal, and the position signal within the time window to generate an initial event determination result. When the initial event determination result is physical door opening or hoisting start, feature extraction is performed on the acceleration signal and attitude angle signal within the time window to obtain the impact decay time constant and tilt retention degree. Based on the impact decay time constant and the tilt retention degree, the mechanical impact state classification result is determined. Based on the mechanical impact state classification results, the temperature and humidity signals within the corresponding time period after the triggering time are retrieved, and the temporal variation characteristics of the temperature and humidity signals are analyzed; based on the temporal correlation between the temporal variation characteristics and the mechanical impact state classification results, the root cause verification of the box anomaly is performed to generate the root cause verification results. Based on the initial event determination result, the mechanical impact state classification result, and the root cause verification result, a joint event code is generated; based on the joint event code, a classification response mechanism is executed.

[0007] In some possible implementations, the step of performing time-series correlation analysis on the acceleration signal, the attitude angle signal, and the position signal within the time window to generate an initial event determination result includes: Based on the triggering time of the Hall switch, the acceleration signal, attitude angle signal and position signal are extracted within a preset time before and after the trigger; Temporal response features, attitude evolution features, and spatial migration features related to the triggering time are extracted from the acceleration signal, the attitude angle signal, and the position signal, respectively. The extracted temporal response features, attitude evolution features, and spatial migration features are matched temporally to construct the local response features and overall migration features after triggering. The initial event determination result is determined based on the coupling relationship between the local response features and the overall migration features.

[0008] In some possible implementations, determining the initial event determination result includes: After the Hall switch is triggered, the transient acceleration response in the local direction, the attitude deflection response in the unilateral direction, and the stabilization response after the initial attitude is restored are extracted within the corresponding time period. By performing time-series correlation on the acceleration transient response, the attitude deflection response, and the stabilization response, the short-range mechanical release process caused by the release of local constraints during the door opening process can be identified. The short-range mechanical release process constitutes the local response feature, and the absence of a continuous position migration process within the corresponding time period constitutes a lack of overall migration feature. When the short-range mechanical release process is detected and no continuous position migration process is detected, the initial event determination result is determined to be physical door opening based on the coupling relationship of the existence of local response features and the absence of overall migration features.

[0009] In some possible implementations, determining the initial event determination result includes: After the Hall switch is triggered, the continuous position migration process, attitude reciprocating swing process and acceleration ground transition process are extracted within the corresponding time period. By temporally coupling the continuous position migration process, the attitude reciprocating swing process, and the acceleration ground transition process, the overall lifting and transmission process formed when the box body changes from a supported state to a suspended state can be identified. When the overall lifting and transmission process is detected, it is determined that the Hall switch trigger is caused by the hoisting operation, and the initial event determination result is determined to be the start of hoisting.

[0010] In some possible implementations, determining the initial event determination result includes: After the Hall switch is triggered, the acceleration signal, attitude angle signal and position signal within a preset time before and after the trigger are jointly verified to generate a verification result; When the verification result indicates that the time-domain response feature, the attitude evolution feature, and the spatial migration feature are all missing, it is determined that the Hall switch triggering did not cause a change in the mechanical state and spatial position of the enclosure, and the initial event determination result is determined to be magnetic field interference.

[0011] In some possible implementations, determining the mechanical impact state classification result based on the impact decay time constant and the degree of tilt retention includes: When the initial event determination result is physical door opening or hoisting start, the mechanical response segment is determined from the acceleration signal and attitude angle signal within the time window based on the impact decay time constant and the tilt holding degree. Acceleration response feature sequences are extracted from the acceleration signals within the mechanical response segment, and attitude response feature sequences are extracted from the attitude angle signals within the mechanical response segment. The acceleration response feature sequences, the attitude response feature sequences, the impact decay time constant, and the tilt retention degree are correlated and analyzed to construct a mechanical response combination feature characterizing motion stability after impact. Based on the combined characteristics of the mechanical response, the classification result of the mechanical impact state is determined.

[0012] In some possible implementations, the mechanical impact condition classification result includes: If the attitude deflects to a limited extent after the impact and enters a recovery process, and the vibration duration corresponding to the recovery process is shorter than the attitude holding time, then the mechanical impact state classification result is determined to be a transportation collision state. If the posture deviates from the initial state after the impact and remains deviated in the subsequent period, and there is still continuous vibration after the impact decays, then the mechanical impact state classification result is determined to be a drop impact state. If the attitude deflects cumulatively in the same direction after the impact, or if the attitude recovery process does not begin after the impact decay ends, the mechanical impact state classification result is determined to be a rollover instability state.

[0013] In some possible implementations, the step of generating root cause verification results for box anomalies based on the temporal correlation between the temporal variation characteristics and the mechanical impact state classification results includes: The Hall switch triggering time period is determined based on the initial event judgment result, and the impact response time period is determined based on the mechanical impact state classification result; Retrieve the temperature and humidity signals during the period corresponding to the Hall switch triggering and the period corresponding to the impact response; Extract the temperature change feature sequence of the temperature signal, and extract the humidity change feature sequence of the humidity signal; Based on the temporal correlation between the temperature change characteristic sequence and the humidity change characteristic sequence, the abnormality verification result of the enclosure is determined.

[0014] In some possible implementations, determining the anomaly verification result of the enclosure includes: If the initial event determination result is physical door opening, and the Hall switch triggers a sudden change in the temperature and humidity signals within the corresponding time period and then stabilizes, the verification result is that the cabinet door is abnormally opened. If the mechanical impact state classification result is a drop impact state or a rollover instability state, and the humidity signal rises after the impact response period and lags behind the end of the mechanical response, then the verification result is water leakage from the box. If the temperature and humidity signals change abnormally during the period corresponding to the initial event and mechanical impact, but the position signal does not represent a continuous position migration process, the verification result is that the terminal seal has failed.

[0015] In some possible implementations, the execution of the classification response mechanism based on the joint event encoding includes: When the mechanical impact state classification result is a drop impact state, the collection time will be extended to the first preset time. When the mechanical impact state classification result is a handling collision state, the collection time will be shortened to the second preset time. When the root cause verification result is water leakage in the enclosure or failure of the terminal seal, local persistent storage is used and a remote alarm is triggered; When the root cause verification result indicates that the cabinet door is abnormally opened, a loop storage method is used to trigger a local alarm.

[0016] As can be seen from the above technical solution, this application has the following beneficial effects: 1. This invention significantly improves the accuracy of container transport status determination by constructing a monitoring architecture based on multi-parameter sensing and hierarchical fusion analysis. Using the Hall switch trigger signal as the time reference and logical starting point, this invention extracts local response features from the transient acceleration response, attitude deflection response, and stabilization response, and combines this with position signals to extract overall migration features. This enables precise differentiation of physical door opening, hoisting initiation, and magnetic field interference, solving the false alarm problem caused by electromagnetic interference in traditional Hall switches.

[0017] 2. This invention introduces two quantitative parameters: impact decay time constant and attitude angle offset retention, to construct a two-dimensional feature space for refined risk classification of mechanical impact events. It can accurately distinguish between three types of events of different severity: handling collisions, drop impacts, and rollover instability. An asynchronous retrieval strategy is used to verify the root cause of temperature and humidity signals. By identifying features such as abrupt recovery patterns and hysteretic monotonous rise patterns, it can accurately trace three different root causes of anomalies: abnormal door opening, water leakage from the enclosure, and terminal sealing failure. Through a differentiated resource control strategy driven by joint event coding, this invention dynamically allocates acquisition frequency, storage method, and alarm priority according to event level, extending terminal battery life while ensuring no loss of critical information. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 A flowchart of the overall monitoring method provided in the embodiments of this application; Figure 2 A flowchart for initial event determination (physical door opening) provided in the embodiments of this application; Figure 3 A flowchart for initial event determination (lifting start) provided for embodiments of this application; Figure 4 A flowchart for classifying mechanical impacts provided in this application embodiment; Figure 5 A root cause verification flowchart provided for embodiments of this application; Figure 6 A flowchart illustrating the differentiated resource control provided in this application embodiment. Detailed Implementation

[0020] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are used to distinguish different objects, not to limit a specific order.

[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] Research has revealed that existing technologies rely on a single sensor for independent threshold judgments. Hall effect switches are susceptible to electromagnetic interference, leading to false alarms, and accelerometers cannot distinguish between minor collisions and severe drops, resulting in high false alarm rates and coarse risk classification. Furthermore, the lack of correlation analysis capabilities between mechanical states and environmental changes makes it impossible to verify the causal relationship between door opening and temperature / humidity changes, hindering the tracing of the delayed root causes of cabinet damage and water leakage. Additionally, the fixed resource strategy leads to wasted battery life and storage space.

[0023] To address the aforementioned issues, this application provides a method for monitoring the status of container transportation based on multi-parameter sensing: Example 1

[0024] To solve the above problems, such as Figures 1-6 As shown, this embodiment sets up a typical container logistics transportation scenario. A container carrying precision instruments has an intelligent monitoring terminal integrated with multiple sensors installed at its door. This terminal includes: a Hall switch sensor for sensing the door's magnetic state; a six-axis inertial measurement unit for collecting three-axis acceleration and three-axis angular velocity; a global navigation satellite system receiver for acquiring geographical location, speed, and direction of motion; and a temperature and humidity sensor for collecting the temperature and relative humidity inside the container. The transportation process includes loading, road transport, port hoisting, sea navigation, and unloading.

[0025] I. Overall Framework of Monitoring Methods

[0026] This method uses the Hall switch trigger signal as the time reference and logic starting point. The Hall switch monitors the state of the door magnet: when the door is closed, the magnet is close to the Hall element, and the sensor outputs a low-level signal; when the door is opened, subjected to external strong magnetic field interference, or when the relative position of the magnet and the Hall element changes due to structural deformation during hoisting, the sensor outputs a high-level signal. The rising edge of this signal is defined as the Hall switch trigger signal.

[0027] This trigger signal is not an isolated alarm, but rather the time base and logical starting point of the entire monitoring process. It instructs the system to begin executing a series of complex data acquisition and analysis tasks. The trigger signal initiates the following five stages: Phase 1: Data Acquisition. Centered on the trigger moment, acquire acceleration, attitude angle, position, and temperature and humidity signals within the time windows before and after the trigger moment.

[0028] Phase Two: Initial Event Determination. Timing correlation analysis is performed on acceleration, attitude angle, and position signals to determine whether the event is physical door opening, hoisting initiation, or magnetic field interference.

[0029] The third stage: mechanical impact classification. If the initial judgment result is physical door opening or hoisting start, it means that the box has undergone real mechanical action. Then, the impact decay time constant and attitude angle offset hold amount are further extracted to construct a two-dimensional feature space and classify the mechanical impact into handling collision state, drop impact state or overturning instability state.

[0030] Phase 4: Root Cause Verification. Based on the initial event assessment and mechanical impact classification results, temperature and humidity signals corresponding to the following time period are asynchronously retrieved from the stored historical data. The characteristics of temperature and humidity changes are analyzed to verify the root causes of enclosure anomalies, including abnormal door opening, water leakage, or terminal seal failure.

[0031] Phase 5: Differentiated Resource Control. All the above judgment results are integrated to generate a joint event code, and differentiated resource control is implemented, dynamically adjusting the data collection duration, storage method, and alarm strategy.

[0032] II. System initialization and multi-source data time synchronization mechanism.

[0033] Before monitoring begins, the system needs to be initialized, the most critical step of which is establishing a unified time reference to ensure the accuracy of all subsequent time-series correlation analyses. Since different sensors have different sampling frequencies, different local clock sources, and clock drift, a precise time synchronization mechanism is necessary to align multi-source data onto a unified time axis. The system uses Coordinated Universal Time (UTC) output from the Global Navigation Satellite System (GNSS) receiver as its master clock. The Hall effect switch sensor, inertial measurement unit, and temperature and humidity sensor each maintain their own independent local hardware clock.

[0034] The system uses the following steps to achieve time synchronization of multi-source data: Step 1: Clock Base Establishment. The Global Navigation Satellite System (GNSS) receiver outputs a pulse signal once per second, with the rising edge of this pulse precisely aligned with the exact second of Coordinated Universal Time (UTC). The system records the UTC timestamp corresponding to the rising edge of each pulse. Simultaneously, record the current value of the local clock counter of each sensor. This pairing record forms the basis of the data points for clock calibration.

[0035] Step 2: Local clock frequency measurement. The frequency of the local clock for each sensor. The actual frequency is determined by a hardware crystal oscillator, but there is a slight deviation between the actual and nominal frequencies. The system accurately calculates the actual frequency by measuring the difference in local clock counts between two consecutive GNS pulses. The denominator of 1 second represents the global navigation satellite system pulse interval. The system performs clock calibration every ten minutes, updating the actual frequency value after each calibration.

[0036] Step 3: Absolute Time Conversion. For any given sensor sampling point, the system records its local clock counter value. The absolute time corresponding to this sampling point. Calculate using the following formula: in The time is the Global Navigation Satellite System time at the time of the most recent calibration. This is the local clock counter value at that calibration time. This is the current actual frequency.

[0037] Step 4: Multi-source data resampling. After the above conversion, all sensor data have obtained a unified timestamp based on Coordinated Universal Time (UTC). However, due to the different sampling frequencies of different sensors (200 Hz for the inertial measurement unit, 1 Hz for the global navigation satellite system, and 1 Hz for the temperature and humidity sensor), the data needs to be resampled onto a unified time axis before performing point-by-point time series analysis. The system uses the nearest neighbor interpolation method to extend the low-frequency signals (Global Navigation Satellite System position, temperature and humidity) onto the time axis to the same 200 Hz as the inertial measurement unit. Specifically, for the target time point... The method involves finding the sampling point in the original data whose timestamp is closest to the given point, and directly copying the value of that sampling point as the interpolation result. This method is simple, efficient, and does not introduce additional signal distortion, making it suitable for location and environmental data that are not continuously changing.

[0038] Through the above synchronization mechanism, all sensor data is converted into a unified timestamp based on Coordinated Universal Time, with synchronization accuracy controlled at the millisecond level.

[0039] III. Data Acquisition and Time Window Configuration.

[0040] When the Hall switch sensor output signal transitions from low to high, the system records this transition moment as a reference moment. The system then uses this reference time as the center and extracts a signal of a preset duration both before and after it, forming a "time window". The design of the time window needs to fully consider the dynamic characteristics of different physical signals: mechanical vibration signals change rapidly and have a short duration, requiring a shorter time window to capture transient features; while position migration and environmental change signals change slowly and have a long duration, requiring a longer time window to observe the complete evolution process.

[0041] (a) Time window for acceleration and attitude angle signals.

[0042] Acceleration and attitude angle signals are acquired by an inertial measurement unit at a sampling frequency of 200 Hz, capable of capturing transient changes at the millisecond level. The system sets the time windows for acceleration and attitude angle to 30 seconds before and after each measurement.

[0043] This timeframe was determined based on statistical analysis of a large amount of actual transportation data. The specific method is as follows: Data from over 1000 real physical door opening events was collected. For each event, starting from the moment the Hall switch was triggered, the envelope of the acceleration signal was continuously monitored, and the time required for the envelope to decay to below 5% of its peak value was recorded. Statistical analysis results show that in 95% of the events, the mechanical response (including the impact peak and subsequent vibration decay) completely ends within 30 seconds. The remaining 5% of events typically involve complex multiple impacts or structural damage, with longer response times, but these require separate handling as abnormal situations. Therefore, a 30-second time window can cover the complete mechanical response process of the vast majority of events, while avoiding the introduction of irrelevant noise due to an excessively long time window.

[0044] (ii) Time window for global navigation satellite system position signals.

[0045] Global Navigation Satellite System (GNSS) receivers typically update at 1 Hz, meaning they output one position point per second. The system sets the time window for the position signal to 5 minutes before and after, for a total of 600 seconds of position data collection (300 seconds before and after, plus the trigger time, for a total of 601 sampling points).

[0046] This timeframe is determined based on an analysis of positional change characteristics in typical container shipping scenarios. In port lifting scenarios, the entire process, from the spreader contacting the container to the container being completely lifted off the ground and beginning horizontal movement, typically takes between 30 seconds and 2 minutes. A 5-minute time window can fully capture the transition from a stationary to a moving state, including the stable state before lifting, positional changes during lifting, and the movement trajectory after lifting. For stopover scenarios in road transport, 5 minutes is also sufficient to distinguish between temporary stops and long-term stops.

[0047] (iii) Time window of temperature and humidity signals.

[0048] Temperature and humidity sensors typically have significant thermal and humidity inertia, with a response time constant of approximately 15 to 30 seconds. This means that sensor readings do not immediately reflect environmental changes but require a certain amount of time to stabilize to a new value. The system sets the time window for the temperature and humidity signals to 30 minutes before and after the initial window.

[0049] This duration was determined based on the analysis of the dynamic characteristics of the temperature and humidity sensor. The sensor's response process can be described by a first-order inertial element: ,in For a moment Sensor readings, The initial stable value before the change, This is the final stable value after the change. The time constant is used. For typical environmental changes, such as a sudden temperature change caused by opening the cabinet door, the sensor readings need approximately five times the time constant, or 150 seconds, to reach more than 95% of the stable value. A 30-minute time window, or 1800 seconds, is sufficient to capture the entire process of the change, including initial stabilization, the occurrence of the sudden change, the transition process, and subsequent recovery, providing sufficient data support for root cause verification.

[0050] (iv) Dynamic adjustment mechanism of time window.

[0051] The aforementioned time window configuration is stored in the terminal's non-volatile memory as a default value. In practical applications, the system can dynamically adjust it via remote commands based on the transportation environment. For example, in ocean-going shipping scenarios, since container attitude changes may last for a long time, such as ship rolling, the attitude angle time window can be extended to 60 seconds; in cold chain transportation scenarios, where temperature and humidity changes require higher precision, the temperature and humidity time window can be extended to 60 minutes to capture slower temperature rise trends. This dynamic adjustment mechanism enables the system to adapt to the needs of different transportation scenarios.

[0052] IV. Initial event determination.

[0053] By constructing two types of macroscopic response patterns—local response features and global migration features—and analyzing their coupling relationship, we can achieve accurate classification of the triggering causes of Hall switches.

[0054] (a) Definition and extraction of local response features.

[0055] Local response characteristics refer to the mechanical vibrations or attitude disturbances generated by the enclosure within a short period of time after an event occurs, without any change in spatial position. These correspond to local state changes in the enclosure's own structure, such as door opening or latch release. This characteristic is composed of a temporal combination of three sub-characteristics.

[0056] Sub-characteristic 1: Acceleration transient response.

[0057] Acceleration transient response refers to a very short-duration spike pulse that appears in the acceleration signal. From a signal processing perspective, the system identifies this feature through the following steps: First, calculate the resultant acceleration to eliminate the influence of direction: in This is the triaxial acceleration value (after subtracting the gravitational component). Set the impact detection threshold. ( The threshold is set based on statistical analysis of more than 1,000 hours of actual transportation data: under normal driving, bumpy, and jolting conditions, 95% of the peak values ​​of the combined acceleration are less than 0.3g. Therefore, setting the threshold to 0.5g can effectively filter out vibrations during normal transportation and only record impact events that may affect the goods.

[0058] The criteria for determining the transient response of acceleration are: the synthetic acceleration exceeds the threshold for three consecutive sampling points, and the impact duration is less than 0.5 seconds. The impact duration is calculated by starting from the first sampling point that exceeds the threshold and ending at the last sampling point that exceeds the threshold, and calculating the time difference between the two.

[0059] Sub-feature 2: Attitude deflection response.

[0060] Attitude yaw response refers to a unidirectional, transient angular change that appears in the attitude angle signal. The system identifies this feature through the following steps: Obtain the roll angle from the attitude calculation module Pitch angle and heading angle For physical door opening events, the main focus is on changes in the roll angle, because the door typically opens around a vertical or horizontal axis, which will cause a slight disturbance to the roll direction of the container.

[0061] Calculate the difference in attitude angles before and after triggering: in The average value within the 5 seconds prior to triggering. This is the average value within 2 seconds after the trigger.

[0062] The criterion for determining attitude deflection response is: the deflection amplitude is within... to The deflection lasts for less than 2 seconds and is consistent with the preset door opening direction. The preset direction is set by configuration parameters during terminal installation; for example, left-opening is the positive direction and right-opening is the negative direction.

[0063] Sub-characteristic 3: Stable response.

[0064] The stabilization response refers to the process by which the attitude angle gradually returns to its original value after an attitude deflection. The system identifies this feature through the following steps: After the attitude deflection response occurs, the attitude angle is continuously monitored. The changes, among which for The angle of rotation of the housing around the longitudinal axis at any given moment. Calculate the attitude angle and the initial attitude reference. Deviation: The criterion for determining the stabilization response is: within 10 seconds after the deflection occurs, the deviation value... Descending to Within this range, and subsequently remaining within this range. The duration of the stabilization process is also recorded; this parameter is used for subsequent mechanical impact classification.

[0065] When the above three features appear sequentially on the time axis (in the order of acceleration transient response → attitude deflection response → stabilization response), and the time interval between adjacent features is within a reasonable range (acceleration response to attitude deflection is less than 1 second, attitude deflection to the start of stabilization is less than 5 seconds), then it is determined that there is a complete short-range mechanical release process, that is, the local response feature is detected.

[0066] (II) Definition and extraction of overall migration features.

[0067] Overall migration characteristics refer to the continuous displacement changes of the container's geographical location that exceed the positioning error range of the Global Navigation Satellite System after an event occurs. This corresponds to the movement of the container as a whole in space, such as hoisting and transportation.

[0068] The system identifies overall migration characteristics through the following steps: Obtain the position coordinates output by the Global Navigation Satellite System And convert it to plane coordinates in meters. This is to calculate the displacement distance.

[0069] Set location migration detection threshold The threshold is set at 10 meters. This threshold is based on the fact that the positioning accuracy of a civilian global navigation satellite system receiver is approximately 3-5 meters in open environments, and may reach 10 meters in complex environments such as ports. Setting the threshold to 10 meters can effectively filter out false migrations caused by positioning drift.

[0070] Calculate the displacement between adjacent sampling points: When the cumulative displacement of three or more consecutive sampling points exceeds And the displacement directions of each segment are consistent (the angle between adjacent displacement vectors is less than 100°). If the position is continuously shifted, then a continuous position shift is determined to have occurred.

[0071] The criteria for determining the overall migration characteristics are: there is a continuous location migration process, and there is a significant distance difference between the starting point and the ending point of the migration (the straight-line distance between the ending point and the starting point exceeds 10 meters).

[0072] (III) Coupling analysis of local response characteristics and global migration characteristics.

[0073] The system performs time-series correlation analysis on the extracted local response features and global migration features, and determines the initial event judgment result based on their existence and coupling relationship.

[0074] Logic 1: Physical door opening.

[0075] After the Hall switch is triggered, the system performs the following analysis steps: Step 1: Acceleration Signal Analysis. The system filters the acceleration signal to remove the gravitational component and low-frequency noise. A sliding window method is used to detect impact: window length 0.5 seconds, step size 0.1 seconds. If the maximum resultant acceleration within a window exceeds 0.5g, and the ratio of the root mean square value of this window to the values ​​of the preceding and following windows is greater than 3 (i.e., the impact is significantly higher than the background noise), the transient response of that acceleration is recorded. The peak impact value is recorded. Impact duration and the moment of impact .

[0076] Step 2: Attitude Angle Signal Analysis. The system monitors the roll angle change within 1 second after triggering. The sequence of differences between the roll angle and the initial reference is calculated. .like The maximum value is at to Between, and the direction of change is unidirectional (i.e. If the deflection is monotonically increasing or monotonically decreasing (without reverse oscillation), then record the attitude deflection response. Record the deflection direction. (Positive or negative), deflection amplitude and the time of deflection .

[0077] Step 3: Stabilization Detection. The system continuously monitors the change in roll angle after attitude deflection occurs. The deviation value is calculated. ,when The first time less than Record the moment when stabilization begins. If within 10 seconds Continue to maintain The following records the stabilization response. The time of stabilization completion is also recorded. Duration of the stabilization process .

[0078] Step 4: Timing Verification. The system verifies the order in which the three features appear: It should be near the trigger moment (±0.2 seconds). Should be in Then (less than 1 second), Should be in Then (less than 5 seconds). If the timing relationship is satisfied, then the local response feature is determined to exist.

[0079] Step 5: Position Signal Analysis. The system analyzes the Global Navigation Satellite System position data within 5 minutes before and after the trigger. The average coordinates of the position points within 30 seconds before and after the trigger are calculated. If the distance between the two coordinates is less than 10 meters, and there is no effective displacement trajectory consisting of more than three consecutive sampling points throughout the entire process, then the overall migration characteristics are considered missing.

[0080] Step 6: Comprehensive Judgment. When local response characteristics exist but overall migration characteristics are missing, the system determines the initial event judgment result as physical door opening. This judgment result means that the Hall switch is triggered by a real door opening action, and not by other reasons.

[0081] Judgment Logic 2: Lifting Start.

[0082] After the Hall switch is triggered, the system performs the following analysis steps: Step 1: Position Signal Analysis. The system analyzes the position data of the Global Navigation Satellite System. A trajectory detection algorithm is used: the displacement vector between consecutive sampling points is calculated. When the cumulative displacement exceeds 10 meters and the directional consistency index (average cosine of the angle between adjacent vectors) is greater than 0.8, it is marked as the start of position migration. If the migration continues for more than three sampling points (i.e., more than 3 seconds), the continuous position migration process is recorded. The start time of migration is recorded. Migration direction and migration speed (Bits removed by time).

[0083] Step 2: Attitude Angle Signal Analysis. The system monitors changes in the heading and roll angles after triggering. The attitude oscillation process exhibits periodic characteristics. The system identifies these characteristics using the following methods: The attitude angle sequence is bandpass filtered with a pass frequency of 0.2 Hz to 1 Hz, corresponding to a period of 1 second to 5 seconds.

[0084] Calculate the zero-crossing rate of the filtered signal. If the number of zero-crossings exceeds 3 times within 10 seconds, i.e. at least 1.5 cycles, and the waveform exhibits sinusoidal oscillation characteristics with an amplitude attenuation rate of less than 50% for adjacent peaks, it is determined to be a reciprocating oscillation.

[0085] Record the swing amplitude (Difference between peak values), oscillation period and oscillation duration .

[0086] Step 3: Acceleration Signal Analysis. The system analyzes the changes in the synthesized acceleration. The acceleration transition process from ground to surface exhibits typical characteristic patterns: First, check if there is an enhancement phase: the synthetic acceleration first rises to more than 1.2g, that is, the upward acceleration exceeds 0.2 times the force of gravity, and lasts for at least 0.1 seconds.

[0087] Then, check if there is a suspension phase: the synthetic acceleration drops below 0.2g (close to a free suspension state) for at least 0.5 seconds.

[0088] If the above two stages occur consecutively in time (the suspension stage begins within 1 second after the lifting stage ends), then record the acceleration transition process from ground to surface. Record the peak lifting value. Suspension start time .

[0089] Step 4: Temporal Coupling Analysis. The system verifies the temporal coupling relationship between the three features: Location migration start time With Hall switch trigger time The difference should be within ±3 seconds.

[0090] The starting time of the oscillating posture should be at Within 5 seconds thereafter.

[0091] The lifting phase of the acceleration-off-ground transition process should be consistent with It may occur at the same time or slightly earlier.

[0092] When the above timing relationship is satisfied, the system determines that these features together constitute the overall lifting and transfer process of the box body from the ground-supported state to the suspended state.

[0093] Step 5: Comprehensive Judgment. When the overall lifting and transfer process is detected, the system determines the initial event judgment result as the start of hoisting. This means that the Hall switch triggering is caused by the hoisting operation, possibly due to deformation or vibration of the housing caused by contact with the lifting equipment, which triggers the Hall switch.

[0094] Judgment Logic 3: Magnetic field interference.

[0095] After the Hall switch is triggered, the system performs joint verification of the acceleration, attitude angle, and position signals for the corresponding time period: Step 1: Acceleration signal verification. The system calculates the root mean square value of acceleration within 10 seconds before and after the trigger moment. and The formula for calculating the root mean square value is: in The acceleration value at the sampling point. This represents the number of sampling points. If... If the change is less than 5%, it is determined that no acceleration disturbance has occurred.

[0096] Step 2: Attitude Angle Signal Verification. The system calculates the standard deviation of the attitude angles within 10 seconds before and after the trigger. and .like and If no attitude evolution has occurred, then it is determined that no attitude evolution has taken place.

[0097] Step 3: Position signal verification. The system calculates the average displacement of the position point before and after the trigger. If the distance between the average coordinates of the position point in the 30 seconds before the trigger and the average coordinates in the 30 seconds after the trigger is less than 10 meters, and the maximum displacement within the entire time window is less than 10 meters, then it is determined that no position migration has occurred.

[0098] Step 4: Comprehensive Judgment. When the system confirms that none of the above three types of signals have undergone mechanical state changes or spatial migration changes exceeding the noise level, it determines that the Hall switch triggering has not caused any physical changes to the enclosure. Therefore, the system infers that the Hall switch triggering can only be caused by external strong magnetic field interference, and determines the initial event judgment result as magnetic field interference.

[0099] V. Classification of Mechanical Impacts.

[0100] By introducing two quantitative parameters with clear physical meaning, namely the impact decay time constant and the attitude angle offset retention amount, a two-dimensional feature space is constructed to achieve refined risk classification of mechanical impact events.

[0101] (a) Delineation of the mechanical response zone.

[0102] The mechanical response zone refers to the data segment from the occurrence of an impact until the housing attitude and vibration tend to stabilize. The system defines the mechanical response zone through the following steps: Starting point determination: from the Hall switch trigger moment Begin by searching forward for acceleration signals. Find the first signal exceeding the impact threshold. The sampling points were used as the impact initiation points. If the forward search does not find it, then... This is the starting point.

[0103] End point determination: from the impact initiation point Begin by searching backwards for acceleration signals. Find the last one that exceeds the impact threshold. The sampling points are then searched, and the search continues until the combined acceleration of all sampling points is lower than a certain value within 10 consecutive seconds. The starting point of the 10-second window is taken as the end point of the impact. .

[0104] Mechanical response section: This is the mechanical response section. This section includes the peak impact, the decay process, and subsequent vibrations.

[0105] (ii) Impact decay time constant The calculation.

[0106] Impact decay time constant It is a key parameter describing the rate of impact energy dissipation. Its calculation process is as follows: Step 1: Composite acceleration calculation. This involves processing the triaxial acceleration signals. , , Calculate the resultant acceleration: in The value is time, in seconds. The composite acceleration eliminates the influence of direction and reflects the total vibration intensity.

[0107] Step 2: High-pass filtering. The synthesized acceleration signal is high-pass filtered with a cutoff frequency of 1 Hz. A second-order Butterworth filter is used, with the following transfer function: in Cutoff angular frequency ( ), The quality factor is used to remove the gravity component (DC component) and low-frequency motion drift, while retaining the impact-related vibration components.

[0108] Step 3: Envelope extraction. This involves processing the filtered signal... Take the absolute value, then pass it through a low-pass filter with a cutoff frequency of 10 Hz to obtain the envelope. The low-pass filter also uses a second-order Butterworth filter, the purpose of which is to smooth the signal and extract the contour of the impulse. Envelope This reflects the decay trend of impact energy over time.

[0109] Step 4: Peak Identification. Locate the envelope within the mechanical response range. maximum point Record the time of its occurrence. Sum of values.

[0110] Step 5: Extracting the attenuation curve. Using the peak point... Starting from the point, extract a 3-second segment of the decay curve. ,in The 3-second cutoff time is based on the analysis of typical impact decay processes, where most impacts have completed their main decay within 3 seconds of peak value.

[0111] Step 6: Exponential Function Fitting. The least squares method is used to fit an exponential function to the truncated decay curve. The fitted function is in the form of: Taking the natural logarithm, the function becomes linear: set up , Then the fitted line is Slope ,therefore: The specific calculation of the least squares method: For Data points The estimated value of the slope is: The summation traversal arrive Calculate back, .

[0112] Step 7: Result Validation. Calculate the coefficient of determination for the fit. : in These are the fitted values. for The mean. If This indicates that the exponential decay model is not fitting well, and the impact may not be a single exponential decay pattern (e.g., there are multiple impacts). In this case, the system should use multi-segment fitting or use the time required for decay to half of the peak value. As an alternative parameter .

[0113] (iii) Attitude angle offset holding amount The calculation.

[0114] Attitude angle offset hold It is a key parameter describing the permanent deformation or displacement caused by impact on the box. Its calculation process is as follows: Step 1: Initial attitude baseline calculation. Collect attitude angle data (roll angle) within 5 seconds before the start of the mechanical response segment. Pitch angle ), calculate their arithmetic mean as the initial attitude reference: in Sampling time, The number of sampling points within 5 seconds (for a sampling rate of 200 Hz). Taking a 5-second average is to smooth out minor fluctuations and obtain a stable initial attitude.

[0115] Step 2: Maximum Deflection Recording. Within the mechanical response range, calculate the deviation of the attitude angle relative to the initial reference at each sampling point: Record the maximum absolute value of the deviation and The change in roll angle and the corresponding time of occurrence should also be recorded. For most impact events, the change in roll angle is more significant, but the change in pitch angle should also be recorded for subsequent analysis.

[0116] Step 3: Final state attitude calculation. Determine the end point of the mechanical response segment. .Pick The attitude angle data within the last 10 seconds are used to calculate the arithmetic mean as the final attitude: in , This represents the number of sampling points over 10 seconds. Taking the average over 10 seconds ensures that the vibration has completely decayed and the attitude has reached a stable state.

[0117] Step 4: Calculation of Attitude Angle Offset. The attitude angle offset is defined as the absolute value of the difference between the final attitude and the initial attitude reference: Choose the larger one as the main offset preservation amount: This value reflects the permanent changes caused by the impact to the housing; the larger the value, the more severe the tilting, deformation, or displacement caused by the impact.

[0118] (iv) Construction of two-dimensional feature space and determination of classification boundary.

[0119] The system uses the impact decay time constant For horizontal coordinate and attitude angle offset holding amount A two-dimensional feature space is constructed for the vertical axis. The classification boundary is determined based on cluster analysis of historical transportation accident data.

[0120] Cluster analysis methods: Data Acquisition: Collect over 1000 known types of mechanical impact events, including artificially simulated handling collisions (300), drop impacts (300), and rollover instability (400). Record the feature vector for each event. .

[0121] Cluster analysis: The K-means clustering algorithm (K=3) was used to divide the samples into three classes. The goal of the K-means algorithm is to minimize the sum of squares within each class. in For the first Class sample set, For the first The center point of the class.

[0122] Clustering results: The three centroids obtained from the cluster analysis are as follows: Transportation and Collision Center: Fall impact category center: Center of flipping instability class: Determining the classification boundaries: The boundary lines between each category are taken as the perpendicular bisector of the line connecting the centers of the two categories. The final determined boundaries are as follows: Collision area during transport ( and ).

[0123] Fall impact zone ( and ).

[0124] Overturning instability region [ or( and )).

[0125] In practical applications, the above boundary values ​​can be dynamically adjusted according to the type of transport vehicle, such as road trucks, rail freight, and ocean-going vessels.

[0126] (v) Determination of classification results.

[0127] The system will calculate Mapping is performed in a two-dimensional feature space, and the final classification result is determined based on the location of the feature points and the actual signal characteristics: Conditions for determining a transport collision: Feature points fall into the transport collision area ( and ).

[0128] The actual signal characteristics are consistent with: the attitude deflects finitely after the impact. And, within 10 seconds, it enters the stabilization process ( Furthermore, the vibration duration during the recovery process (the time from the end of the impact to the vibration decaying to below the threshold) is shorter than the attitude maintenance time (the duration of attitude deflection).

[0129] When the above conditions are met, it is determined to be a handling collision state. This state corresponds to a minor scratch or collision, in which the container can quickly regain stability and has minimal impact on cargo safety.

[0130] Conditions for determining a drop impact condition: Feature points fall into the drop impact zone ( and ).

[0131] The actual signal characteristics are consistent with: the attitude deviates from the initial state after the impact. And it remains deviated in the subsequent period (it does not return to stability after the final state attitude is stable), and there is still continuous vibration after the impact decays (there is still more than 30% of the vibration after decay for more than 1 second).

[0132] When the above conditions are met, the condition is determined to be a drop impact condition. This condition corresponds to the container falling from a certain height, resulting in slight structural deformation and continuous vibration, posing a significant threat to the safety of the cargo.

[0133] Conditions for determining a state of overturning instability: Feature points fall into the flipping instability region ( ,or and ).

[0134] The actual signal characteristics are consistent with: after the impact, the attitude deflects cumulatively in the same direction (after multiple impacts, the attitude angle increases successively without stabilizing), or the attitude recovery process has not yet begun after the impact decay ends. (Not present or longer than 30 seconds).

[0135] When the above conditions are met, the condition is determined to be a rollover instability state. This state corresponds to the container overturning or severely tipping over, posing the highest threat to cargo safety.

[0136] VI. Root cause verification.

[0137] The asynchronous retrieval strategy and pattern recognition of temperature and humidity change characteristics enable the tracing of the root cause of abnormalities in the enclosure.

[0138] (a) The meaning of asynchronous call strategy.

[0139] Asynchronous retrieval refers to the system retrieving temperature and humidity data from the historical storage database before and after a corresponding time period based on the event type and occurrence time indicated by the initial event determination and mechanical impact classification. This read operation is time-independent, does not block other analysis processes, and does not require the temperature and humidity sensors to respond in real time when the event occurs. This design fully considers the lag in temperature and humidity changes; that is, environmental changes such as water leakage in the enclosure may occur several hours after a mechanical impact, thus requiring data to be retrieved from storage for analysis afterward.

[0140] (ii) Rules for dynamically determining the corresponding time period.

[0141] The system dynamically determines the time period for retrieving temperature and humidity data based on the judgment results. Different judgment results correspond to different retrieval reference points. This is because the environmental change caused by the physical opening of the door occurs instantaneously, and can be captured by using the trigger moment as the reference. However, water seepage caused by a drop has a lag, requiring monitoring to be extended backward from the end of the mechanical response. The system dynamically determines the time period for retrieving temperature and humidity data based on the judgment results: When a physical door opening is detected, data from 30 minutes before and after the trigger time is retrieved to capture environmental changes and the recovery process.

[0142] When the condition is determined to be a drop impact state or an overturning instability state, data from 24 hours prior to the end of the mechanical response section is retrieved to monitor the lag environmental change of water seepage.

[0143] When a collision is detected during transport, data is not retrieved immediately. Instead, continuous 24-hour monitoring is initiated, with temperature and humidity trends analyzed every hour to prevent the risk from escalating.

[0144] When the problem is determined to be magnetic field interference, no root cause verification is performed.

[0145] When multiple judgment results coexist, the maximum union of all corresponding time periods is taken as the retrieval range.

[0146] (III) Extraction of temperature and humidity change characteristics.

[0147] The system extracts the following variation characteristics from the retrieved temperature and humidity signals: Step 1: Baseline Calculation. Collect temperature and humidity data from 10 minutes before the event or during a stable period before the event, and calculate the stable mean and standard deviation: Step 2: Detection of the Starting Point of Change. Starting from the event time, scan the temperature and humidity data forward and backward. The criterion for determining the starting point of change is: the values ​​of three consecutive sampling points exceed the baseline plus or minus three times the standard deviation, i.e.: Record the time of the first out-of-range sampling point. A range of 3 times the standard deviation corresponds to a 99.7% confidence interval in statistics; anything outside this range can be considered a significant anomaly.

[0148] Step 3: Calculate the magnitude of change. Starting from the point of change... Next, find the peak value of the temperature or humidity. or The range of change is: Step 4: Calculate the duration of the change. Starting from the point of change... The time from the start to the point when the data begins to regress to the baseline (or continues to deviate and no longer regresses) Up to this point, the duration of the change is: If the data no longer regresses, then This refers to the time until the data ends.

[0149] Step 5: Recovery Process Analysis. If the data shows a trend of regression to the baseline after the changes, calculate the recovery process parameters: Recovery starting point: The point in time when the data begins to regress to the baseline.

[0150] Recovery completion time: The point at which the data stabilizes within the baseline plus or minus one standard deviation.

[0151] Recovery time constant: The characteristic time of the recovery process is obtained by exponential fitting.

[0152] (iv) Determination of verification results.

[0153] Verification result 1: The cabinet door was opened abnormally.

[0154] When the initial event determination result is physical door opening, the system retrieves temperature and humidity data for the 30 minutes before and after the door opening action. If the analysis reveals the following characteristic patterns, the verification result is that the cabinet door was abnormally opened: Abrupt change characteristics: At the moment the door opens (within ±5 seconds), both the temperature and humidity signals exhibit a steep abrupt change. Temperature change amplitude. Humidity variation range This threshold is based on statistical analysis of more than 500 actual door opening events: the temperature difference between the inside and outside of the box usually exceeds 5°C, and the humidity difference usually exceeds 20%.

[0155] Recovery characteristics: Within 5–30 minutes after the door is closed, both signals begin to revert to their pre-event stable values, forming a complete recovery process. The recovery time constant is consistent with the insulation and moisture retention characteristics of the enclosure, typically 5–15 minutes.

[0156] Timing consistency: The start time of temperature and humidity changes is consistent and aligned with the door opening time.

[0157] This characteristic pattern indicates that after a brief exchange between the internal and external environments, the environment recovered thanks to the chamber's own insulation and moisture-retaining capabilities, confirming that the chamber door was indeed opened. If no temperature and humidity change characteristics are detected during the corresponding time period, the verification result is that there is no environmental anomaly.

[0158] Verification result 2: Water leakage from the enclosure.

[0159] When the mechanical impact classification result is a drop impact or a rollover instability state, the system retrieves the humidity signal within 24 hours after the impact. If the analysis reveals the following characteristic pattern, the verification result is water leakage from the enclosure: Hysteresis characteristics: In the first 1–6 hours after the impact, the humidity signal remains stable (deviation from baseline is less than 1%). Subsequently, the humidity signal began to show a sustained upward trend, with the start time of the rise significantly lagging behind the end time of the mechanical response segment (more than 1 hour later).

[0160] Monotonic increase characteristics: The upward trend in humidity is continuous, with humidity values ​​increasing monotonically for more than 6 consecutive hours (the average value for each hour is greater than that for the previous hour), and the cumulative increase exceeds 10%. There are no signs of stabilization during the upward process (no return to baseline).

[0161] Temperature decreases synchronously: As humidity increases, the temperature signal shows a synchronous downward trend. For every 10% increase in humidity, the temperature decreases by 0.5~2°C, which is consistent with the physical law of heat absorption by water evaporation.

[0162] No recovery characteristic: The humidity did not show a trend of reverting to the baseline throughout the entire 24-hour monitoring window.

[0163] This characteristic pattern indicates that after the enclosure is damaged by impact, external moisture (such as rainwater or seawater) slowly seeps into the enclosure and accumulates and evaporates inside. The lag in humidity increase is due to the time required for moisture penetration; once the penetration process begins, it is irreversible.

[0164] Verification result 3: Terminal seal failure.

[0165] If the system detects abnormal changes in temperature and humidity signals (e.g., temperature fluctuates drastically by more than 10°C in a short period of time, or humidity rises abnormally by more than 30%), but synchronous analysis of the position signal reveals the following conditions, then the verification result is that the terminal seal has failed: Unchanged location: During the period when the temperature and humidity anomaly occurred, the geographical location of the container did not change from one environmental area to another, that is, the change in the global navigation satellite system coordinates was less than 10 meters.

[0166] No mechanical incidents: No incidents such as physical door opening or falling impacts that could damage the enclosure structure have been identified previously.

[0167] Sensor malfunction: Temperature and humidity changes do not conform to physical laws, such as temperature and humidity rising synchronously instead of changing in opposite directions, or the rate of change exceeding the sensor's response capability.

[0168] This characteristic pattern indicates that the environmental anomaly was not caused by the opening of the cabinet door or damage to the cabinet, but rather by the breach in the sealing of the monitoring terminal's own casing, which caused external air or moisture to directly affect the sensor readings, or by a malfunction of the sensor itself.

[0169] (v) Continuous monitoring mechanism.

[0170] For events classified as transport collisions, although the initial risk level is low, there is still a possibility that they may evolve into water leakage in the enclosure, such as micro-cracks leading to slow seepage. Therefore, after saving the summary data, the system will initiate a continuous monitoring task: Monitoring cycle: Temperature and humidity data are retrieved once per hour within 24 hours after the event occurs.

[0171] Trend analysis: For each retrieved data point, calculate the average humidity over the past hour and compare it with the average of the previous hour. Construct a humidity time series. ,in For the first Average humidity level per hour.

[0172] Anomaly detection: If a monotonically increasing humidity value is detected for more than 3 consecutive hours ( If the cumulative increase exceeds 10%, it is judged as a "continuous upward trend in humidity".

[0173] Status escalation: When a continuous upward trend in humidity is detected, the system automatically updates the verification result to water leakage in the enclosure and triggers a high-priority alarm.

[0174] This mechanism ensures that low-risk events are not overlooked due to initial classification, thus enabling full lifecycle tracking of events.

[0175] VII. Differentiated Resource Control By integrating multi-stage decision results through joint event coding, differentiated resource management strategies are driven to achieve intelligent allocation of limited terminal resources, such as power, storage space, and communication bandwidth.

[0176] (a) Rules for generating joint event codes The joint event coding adopts a structured format, consisting of three fields: initial event type, mechanical impact classification, and root cause verification result, concatenated with underscores. The first field's value is either physical door opening, hoisting initiation, or magnetic field interference; the second field's value is either handling collision state, drop impact state, rollover instability state, or none; the third field's value is either abnormal door opening, water leakage from the enclosure, or terminal seal failure, or none. For example, normal door opening without abnormalities is coded as "physical door opening_none_none," door opening causing a sudden change in temperature and humidity is coded as "physical door opening_none_abnormal door opening," and a drop causing water leakage is coded as "hoisting initiation_drop impact state_enclosure water leakage." The total code length does not exceed 64 bytes and is stored in the terminal memory as a lookup index for differentiated resource control strategies.

[0177] (ii) Differentiated resource control strategy The system executes differentiated resource control strategies based on the joint event code, as follows: When the first field is "magnetic interference", the system determines that the event is invalid interference, restores the default sampling frequency, clears all sensor data caches within the trigger time window, and suppresses the alarm to free up storage space.

[0178] When the second field is "transportation collision status", the system determines it as a low-risk event, maintains the current acquisition frequency, and extracts key feature values ​​from the mechanical response segment to form a summary data packet, including event timestamp, peak impact acceleration, attitude angle offset hold-up, impact decay time constant and joint event code, not exceeding 200 bytes. It does not trigger a real-time alarm, but starts 24-hour continuous monitoring to track temperature and humidity changes and prevent risk evolution.

[0179] When the second field is "drop impact state" or "rollover instability state", the system determines it as a high-risk event, increases the acceleration and attitude angle acquisition frequency to 400 Hz and continues for 10 minutes, performs lossless compression and permanent storage of the raw data within the entire event time window, including complete acceleration waveform, attitude angle curve, positioning trajectory and temperature and humidity curve, and reports a high-priority alarm through the wireless network, repeating three times to ensure delivery.

[0180] When the third field is "box leakage" or "terminal seal failure", the system determines it as a root cause problem that needs to be continuously monitored, increases the temperature and humidity collection frequency to once every 10 seconds for 72 hours, continuously saves the temperature and humidity data for 24 hours before and after the abnormal period, and immediately reports a high-priority alarm.

[0181] (III) Strategy Overlay Rules When multiple judgment results exist simultaneously, such as "physical door opening" followed by "fall impact", the system executes resource control strategies according to the following rules: Collection duration: Take the maximum collection duration among all triggering strategies.

[0182] Storage method: Select the most stringent storage method (continuous data storage > extended data storage > summary data storage > clear cache).

[0183] Alarm strategy: Take the highest priority (high priority alarm > alarm suppression).

[0184] For example, for the joint event code "physical door opening_drop impact state_box water leakage", the system will execute the superposition of strategy three and strategy four: the acceleration acquisition frequency is increased to 400 Hz, the temperature and humidity acquisition frequency is increased to once every 10 seconds, the complete mechanical response data is saved, the extended temperature and humidity data is saved, and a high-priority alarm is reported.

[0185] VIII. Attitude Calculation Methods.

[0186] To ensure the accuracy of the attitude angle signal, the system uses the quaternion method for attitude calculation to avoid the gimbal lock problem that occurs when the pitch angle is close to 90°.

[0187] (a) Definition of quaternion.

[0188] A quaternion consists of one real part and three imaginary parts, represented as: ,in The imaginary unit. The modulus of a quaternion is defined as: For attitude representation, unit quaternions are typically used. ).

[0189] (ii) Determining the initial quaternion.

[0190] When the system starts up and the enclosure is stationary, the vector output of the accelerometer... This is the direction of gravity. The initial roll and pitch angles can be calculated using this vector: Initial heading angle It is usually set to zero, or obtained through calibration with a magnetometer. From Euler angles The formula for converting to quaternions is: (III) Update of Quaternion Differential Equations.

[0191] Attitude updates are performed using quaternion differential equations: in Angular velocity vector Quaternions , This represents quaternion multiplication.

[0192] Definition of quaternion multiplication: For and their product for: After discretization, numerical integration is performed using the first-order Runge-Kutta method (Euler method): in For a sampling rate of 200 Hz, the sampling interval is... Second.

[0193] (iv) Normalization process.

[0194] After each update, the modulus of the quaternion may deviate from 1 due to accumulated error. The system performs normalization processing: (v) Conversion from quaternions to Euler angles.

[0195] Convert the updated quaternion to Euler angles (roll angles). Pitch angle Heading angle The conversion formula is: in This is the arctangent function in the four quadrants, returning an angle range of... ; The return angle range is .

[0196] (vi) Motion interference suppression.

[0197] When the enclosure experiences significant linear acceleration, such as during acceleration, deceleration, or collisions, the accelerometer output includes not only the gravitational component but also the motion acceleration. In this case, attitude calculations based on the accelerometer will be interfered with. The system employs the following suppression strategy: Motion detection: Calculating composite acceleration .like If so, it is determined that there is a large acceleration.

[0198] Update strategy: When a large acceleration is detected, the system pauses attitude updates and maintains the valid attitude from the previous moment. Specifically: (No accelerometer correction is used; integration is done solely using the gyroscope).

[0199] Recovery mechanism: When If the motion continues for more than 1 second, the system determines that the motion has ended and re-initializes the attitude (using the current acceleration vector as the direction of gravity to correct the accumulated error).

[0200] Using the above methods, the system can maintain the accuracy of attitude angles under motion disturbance environments, providing a reliable data foundation for subsequent impact classification and event determination.

[0201] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the status of container transportation based on multi-parameter sensing, characterized in that, The method includes: Using the Hall switch trigger signal as the time reference and logic starting point, the acceleration signal, attitude angle signal, position signal and temperature and humidity signal are collected within the time window before and after the trigger moment; A time-series correlation analysis is performed on the acceleration signal, the attitude angle signal, and the position signal within the time window to generate an initial event determination result. When the initial event determination result is physical door opening or hoisting start, feature extraction is performed on the acceleration signal and attitude angle signal within the time window to obtain the impact decay time constant and tilt retention degree. Based on the impact decay time constant and the tilt retention degree, the mechanical impact state classification result is determined. Based on the mechanical impact state classification results, the temperature and humidity signals within the corresponding time period after the triggering time are retrieved, and the temporal variation characteristics of the temperature and humidity signals are analyzed; based on the temporal correlation between the temporal variation characteristics and the mechanical impact state classification results, the root cause verification of the box anomaly is performed to generate the root cause verification results. Based on the initial event determination result, the mechanical impact state classification result, and the root cause verification result, a joint event code is generated; based on the joint event code, a classification response mechanism is executed.

2. The method according to claim 1, characterized in that, The step of performing time-series correlation analysis on the acceleration signal, attitude angle signal, and position signal within the time window to generate initial event determination results includes: Based on the triggering time of the Hall switch, the acceleration signal, attitude angle signal and position signal are extracted within a preset time before and after the trigger; Temporal response features, attitude evolution features, and spatial migration features related to the triggering time are extracted from the acceleration signal, the attitude angle signal, and the position signal, respectively. The extracted temporal response features, attitude evolution features, and spatial migration features are matched temporally to construct the local response features and overall migration features after triggering. The initial event determination result is determined based on the coupling relationship between the local response features and the overall migration features.

3. The method according to claim 2, characterized in that, The determination of the initial event judgment result includes: After the Hall switch is triggered, the transient acceleration response in the local direction, the attitude deflection response in the unilateral direction, and the stabilization response after the initial attitude is restored are extracted within the corresponding time period. By performing time-series correlation on the acceleration transient response, the attitude deflection response, and the stabilization response, the short-range mechanical release process caused by the release of local constraints during the door opening process can be identified. The short-range mechanical release process constitutes the local response feature, and the absence of a continuous position migration process within the corresponding time period constitutes a lack of overall migration feature. When the short-range mechanical release process is detected and no continuous position migration process is detected, the initial event determination result is determined to be physical door opening based on the coupling relationship of the existence of local response features and the absence of overall migration features.

4. The method according to claim 2, characterized in that, The determination of the initial event judgment result includes: After the Hall switch is triggered, the continuous position migration process, attitude reciprocating swing process and acceleration ground transition process within the corresponding time period are extracted; By temporally coupling the continuous position migration process, the attitude reciprocating swing process, and the acceleration ground transition process, the overall lifting and transmission process formed when the box body changes from a supported state to a suspended state can be identified. When the overall lifting and transmission process is detected, it is determined that the Hall switch trigger is caused by the hoisting operation, and the initial event determination result is determined to be the start of hoisting.

5. The method according to claim 2, characterized in that, The determination of the initial event judgment result includes: After the Hall switch is triggered, the acceleration signal, attitude angle signal and position signal within a preset time before and after the trigger are jointly verified to generate a verification result; When the verification result indicates that the time-domain response feature, the attitude evolution feature, and the spatial migration feature are all missing, it is determined that the Hall switch triggering did not cause a change in the mechanical state and spatial position of the enclosure, and the initial event determination result is determined to be magnetic field interference.

6. The method according to claim 1, characterized in that, The step of determining the mechanical impact state classification result based on the impact attenuation time constant and the degree of tilt retention includes: When the initial event determination result is physical door opening or hoisting start, the mechanical response segment is determined from the acceleration signal and attitude angle signal within the time window based on the impact decay time constant and the tilt holding degree. Acceleration response feature sequences are extracted from the acceleration signals within the mechanical response segment, and attitude response feature sequences are extracted from the attitude angle signals within the mechanical response segment. The acceleration response feature sequences, the attitude response feature sequences, the impact decay time constant, and the tilt retention degree are correlated and analyzed to construct a mechanical response combination feature characterizing motion stability after impact. Based on the combined characteristics of the mechanical response, the classification result of the mechanical impact state is determined.

7. The method according to claim 6, characterized in that, The mechanical impact condition classification results include: If the attitude deflects to a limited extent after the impact and enters a recovery process, and the vibration duration corresponding to the recovery process is shorter than the attitude holding time, then the mechanical impact state classification result is determined to be a transportation collision state. If the posture deviates from the initial state after the impact and remains deviated in the subsequent period, and there is still continuous vibration after the impact decays, then the mechanical impact state classification result is determined to be a drop impact state. If the attitude deflects cumulatively in the same direction after the impact, or if the attitude recovery process does not begin after the impact decay ends, the mechanical impact state classification result is determined to be a rollover instability state.

8. The method according to claim 1, characterized in that, The step of performing root cause verification on the box anomaly based on the temporal correlation between the temporal change characteristics and the mechanical impact state classification results includes: The Hall switch triggering time period is determined based on the initial event judgment result, and the impact response time period is determined based on the mechanical impact state classification result; Retrieve the temperature and humidity signals during the period corresponding to the Hall switch triggering and the period corresponding to the impact response; Extract the temperature change feature sequence of the temperature signal, and extract the humidity change feature sequence of the humidity signal; Based on the temporal correlation between the temperature change characteristic sequence and the humidity change characteristic sequence, the abnormality verification result of the enclosure is determined.

9. The method according to claim 8, characterized in that, The determination of the abnormality verification result of the enclosure includes: If the initial event determination result is physical door opening, and the Hall switch triggers a sudden change in the temperature and humidity signals within the corresponding time period and then stabilizes, the verification result is that the cabinet door is abnormally opened. If the mechanical impact state classification result is a drop impact state or a rollover instability state, and the humidity signal rises after the impact response period and lags behind the end of the mechanical response, then the verification result is water leakage from the box. If the temperature and humidity signals change abnormally during the period corresponding to the initial event and mechanical impact, but the position signal does not represent a continuous position migration process, the verification result is that the terminal seal has failed.

10. The method according to claim 1, characterized in that, The step of executing a classification response mechanism based on the joint event encoding includes: When the mechanical impact state classification result is a drop impact state, the collection time will be extended to the first preset time. When the mechanical impact state classification result is a handling collision state, the collection time will be shortened to the second preset time. When the root cause verification result is water leakage in the enclosure or failure of the terminal seal, local persistent storage is used and a remote alarm is triggered; When the root cause verification result indicates that the cabinet door is abnormally opened, a loop storage method is used to trigger a local alarm.

Images