Container interior cargo state monitoring method, device and electronic equipment

Abstract

Embodiments of the present disclosure disclose a method and device for monitoring the state of goods in a container and an electronic device. A specific embodiment of the method comprises: determining wind state information according to real-time vehicle state information, track state information, and real-time weather information of the area where the target tramcar is located; generating a predicted vibration signal according to the real-time vehicle state information, the track state information, the wind state information, vehicle basic information corresponding to the target tramcar, and a pre-trained vibration noise signal prediction model; performing noise signal stripping on a real-time vibration signal according to the predicted vibration signal to obtain a stripped vibration signal; determining a vibration state of the goods in the container according to the stripped vibration signal; and determining whether to initiate a vibration anomaly warning according to the vibration state and a dynamic warning threshold corresponding to the area where the target tramcar is located. The embodiment provides a vibration anomaly warning with high accuracy.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the fields of cargo transportation status monitoring and computer technology, and specifically to methods, apparatus and electronic devices for status monitoring of cargo inside containers. Background Technology

[0002] As a crucial mode of land transportation, railway transport plays a vital role in connecting coastal ports with inland areas and bridging inland regions. During medium- to long-distance transport of goods, especially high-value precision instruments, vibrations generated during transit can easily damage the cargo. Currently, the common approach is to monitor vibrations during transport using vibration sensors, thereby issuing early warnings when abnormal vibrations occur.

[0003] However, the container, as an intermediary, transmits vibrations generated by vehicle movement to the cargo. This vibration is both the cause of cargo vibration and the noise that interferes with the monitoring of abnormal cargo vibration. Due to the failure to effectively separate the noise vibration signal, the false alarm rate of abnormal vibration is high and the accuracy is poor. Summary of the Invention

[0004] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0005] Some embodiments of this disclosure propose methods, apparatuses, and electronic devices for monitoring the condition of cargo within containers, in order to address the technical problems mentioned in the background section above.

[0006] In a first aspect, some embodiments of this disclosure provide a method for monitoring the state of goods inside a container. The method includes: acquiring real-time vibration signals, real-time vehicle state information corresponding to a target railcar, and track state information, wherein the real-time vibration signals are collected by a triaxial sensor mounted on a cargo pallet used for carrying the goods, and the cargo pallet is placed inside the container carried by the target railcar; determining wind state information based on the real-time vehicle state information, the track state information, and real-time weather information of the area where the target railcar is located, wherein the wind state information represents a wind component perpendicular to the direction of travel of the target railcar; generating a predicted vibration signal based on the real-time vehicle state information, the track state information, the wind state information, the vehicle basic information corresponding to the target railcar, and a pre-trained vibration noise signal prediction model; performing noise stripping on the real-time vibration signal based on the predicted vibration signal to obtain a noise-stripped vibration signal; determining the vibration state of the goods inside the container based on the noise-stripped vibration signal; and determining whether to initiate a vibration anomaly warning based on the vibration state and a dynamic warning threshold corresponding to the area where the target railcar is located.

[0007] Secondly, some embodiments of this disclosure provide a status monitoring device for goods inside a container. The device includes: an acquisition unit configured to acquire real-time vibration signals, real-time vehicle status information corresponding to a target railcar, and track status information, wherein the real-time vibration signals are collected by a triaxial sensor installed on a cargo pallet used for carrying the goods, and the cargo pallet is installed inside a container carried by the target railcar; and a first determination unit configured to determine wind status information based on the aforementioned real-time vehicle status information, the aforementioned track status information, and the aforementioned real-time weather information of the area where the target railcar is located, wherein the wind status information represents wind direction perpendicular to the direction of travel of the target railcar. The system comprises: a generation unit configured to generate a predicted vibration signal based on the real-time vehicle status information, track status information, wind status information, basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model; a noise signal stripping unit configured to strip noise from the real-time vibration signal based on the predicted vibration signal to obtain a stripped vibration signal; a second determination unit configured to determine the vibration state of the cargo inside the container based on the stripped vibration signal; and a third determination unit configured to determine whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located.

[0008] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0009] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0010] The above-described embodiments of this disclosure have the following beneficial effects: The method for monitoring the state of goods inside a container, as described in some embodiments of this disclosure, achieves highly accurate vibration anomaly early warning. Specifically, the reason for poor accuracy in vibration anomaly early warning is that the container, as an intermediary, transmits vibrations generated by vehicle movement to the goods. This vibration is both the cause of vibration in the goods and noise that interferes with the monitoring of abnormal vibrations in the goods, as the noise vibration signal is not effectively separated. Based on this, the method for monitoring the state of goods inside a container, as described in some embodiments of this disclosure, firstly acquires real-time vibration signals, real-time vehicle state information corresponding to the target railcar, and track state information. The real-time vibration signals are collected by a triaxial sensor installed on a cargo pallet used to carry the goods, and the cargo pallet is placed inside the container transported by the target railcar. Secondly, based on the aforementioned real-time vehicle state information, track state information, and real-time weather information of the area where the target railcar is located, wind state information is determined. The wind state information represents the wind component perpendicular to the direction of travel of the target railcar. In practice, crosswinds can cause trams to sway and induce vibrations during operation. Furthermore, the degree of crosswind influence varies at different speeds. Therefore, this disclosure combines real-time vehicle status information, track status information, and real-time weather information to decompose the crosswinds (wind status information) that may induce vehicle vibrations. Then, based on the aforementioned real-time vehicle status information, track status information, wind status information, the basic vehicle information corresponding to the target tram, and a pre-trained vibration noise signal prediction model, a predicted vibration signal is generated. In practice, the train body, containers, cargo pallets, and cargo are often rigidly connected. Factors such as vehicle operation, track status, and weather can cause vibrations to be transmitted to the cargo through the containers, thus inducing cargo vibrations. Since the container is the intermediate body for vibration transmission, and the vibrations generated by the above factors can all be understood as noise vibrations, to reduce computational and data processing complexity, this disclosure treats all the above factors as a whole for predicting the (noise) vibration signal. In practice, noise and vibration signals can also be directly monitored by setting up another set of triaxial sensors at the bottom of the container. However, compared to the method disclosed herein, the number of sensors required increases exponentially, resulting in a significant increase in hardware and calibration costs. This is especially true given the large number of containers in use, where costs surge. Therefore, this disclosure employs an algorithmic prediction method to predict noise and vibration signals, thereby reducing hardware costs. Furthermore, based on the predicted vibration signal, noise is stripped from the real-time vibration signal to obtain the stripped vibration signal. This achieves direct noise and vibration signal stripping. Next, based on the stripped vibration signal, the vibration state of the cargo inside the container is determined. This maps the stripped vibration signal to an index-based vibration state.Finally, based on the aforementioned vibration state and the dynamic warning threshold corresponding to the area where the target tram is located, it is determined whether to initiate a vibration anomaly warning. In particular, considering the differences in environmental factors in different areas, this disclosure recognizes that setting a fixed threshold results in a high false alarm rate. Therefore, a method of setting dynamic thresholds for different areas is adopted to improve the accuracy of the warning. In summary, this method achieves highly accurate vibration anomaly warnings. Attached Figure Description

[0011] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0012] Figure 1 This is a flowchart of some embodiments of a method for monitoring the condition of goods inside a container according to the present disclosure; Figure 2 This is a schematic diagram illustrating the process of determining wind component; Figure 3 This is a schematic diagram of the process of generating predicted vibration signals; Figure 4 This is a schematic diagram of the process of generating predicted vibration signals; Figure 5 This is a schematic diagram of some embodiments of a status monitoring device for cargo inside a container according to the present disclosure; Figure 6 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0013] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0014] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0015] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0016] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0017] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0018] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of a method for monitoring the condition of cargo within a container according to the present disclosure. This method for monitoring the condition of cargo within a container includes the following steps: Step 101: Obtain real-time vibration signals, real-time vehicle status information and track status information corresponding to the target railcar.

[0020] In some embodiments, the execution subject (e.g., a computing device) of the method for monitoring the status of goods inside a container can acquire real-time vibration signals, real-time vehicle status information corresponding to the target railcar, and track status information through wired or wireless connections.

[0021] The target railcar can be a train (e.g., a railway train) that transports goods along pre-laid tracks. The cargo pallets are placed inside containers carried by the target railcar. Specifically, the target railcar can be in formation, pulling multiple trailers, each trailer carrying a container containing goods loaded onto cargo pallets.

[0022] The real-time vibration signal is acquired by a triaxial sensor mounted on the cargo pallet used for carrying the goods. Specifically, the triaxial sensor requires calibration before use. The real-time vibration signal is a sequence of signal values ​​in time series form. Because the real-time vibration signal is acquired by the triaxial sensor, the signal value set can contain vibration component values ​​(signal values) along the XYZ axes at the same time.

[0023] Real-time vehicle status information can characterize the time-varying acceleration and velocity values ​​of the target railcar as it travels along the track. Specifically, the real-time vehicle status can be represented as a time-series sequence of six values. This six-value set includes: acceleration components along the XYZ axes (3 components) and velocity components along the XYZ axes (3 components). In practice, real-time vehicle status information can be acquired in real time by IMU (Inertial Measurement Unit) sensors installed within the target railcar.

[0024] Track status information characterizes the track status of the area where the target tram is located. Specifically, since the track is continuous, the track status information can be a sequence of ternary values, which can include: track sampling point coordinates, track deflection angle, and track elevation. In practice, the acquisition of track status information first uses GPS (Global Positioning System) to obtain the current position of the target tram. Then, the coordinates of the track sampling points, track deflection angle, and track elevation corresponding to the current position are read from a pre-built BIM (Building Information Modeling) model of the track distribution as track status information. Alternatively, the current position can also be determined using the BeiDou or Galileo satellite navigation systems.

[0025] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra wideband) connections, and other currently known or future wireless connection methods.

[0026] It should be noted that the aforementioned computing devices can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed on the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.

[0027] Step 102: Determine the wind status information based on the real-time vehicle status information, track status information, and real-time weather information of the area where the target tram is located.

[0028] In some embodiments, the aforementioned executing entity may determine wind status information based on real-time vehicle status information, track status information, and real-time weather information of the area where the target tram is located.

[0029] The wind state information represents the wind component perpendicular to the direction of travel of the target railcar. Specifically, due to changes in track elevation and vehicle direction of travel, the wind state information can be a time-series sequence of binary values, where each binary value includes wind speed and wind direction. For example, wind direction can be represented by the angle between the wind and the X-axis in a geodetic coordinate system.

[0030] The real-time weather information represents the wind speed and direction in the area where the target tram is located. In practice, after locating the current position, the weather API (Application Programming Interface) can be called to query the wind speed and direction of the current location as real-time weather information.

[0031] In practice, firstly, the direction of travel of the target railcar can be determined using the speed values ​​(three components along the XYZ axes) corresponding to the vehicle's status information, applying the parallelogram law. Then, since the track status information includes track elevation, the plane containing the track can be determined based on the geodetic coordinate system and the track elevation. Next, by combining wind speed and direction, the wind component perpendicular to the direction of travel in the plane containing the track can be determined as the wind status information.

[0032] In some optional implementations of certain embodiments, the executing entity determines wind state information based on the real-time vehicle state information, the track state information, and the real-time weather information of the area where the target tram is located, including: Step S1: Determine the basic direction based on the real-time vehicle status information mentioned above.

[0033] The aforementioned basic direction is in the same direction as the movement direction of the aforementioned target railcar.

[0034] In practice, the velocity direction of the target railcar can be used as the base direction. As the target railcar moves, its velocity direction changes, and therefore the base direction also changes accordingly.

[0035] Step S2: Determine the plane where the track is located based on the above track status information.

[0036] In practice, firstly, the XoY plane contained in the geodetic coordinate system can be used as the base plane. Then, by combining the track elevation included in the track status information, the plane on which the track lies can be obtained.

[0037] Step S3: Based on the real-time weather information mentioned above, determine the wind component perpendicular to the basic direction within the plane where the track is located, and use it as the wind state information.

[0038] In practice, since the base direction changes as the target railcar moves, the wind component perpendicular to the base direction also changes accordingly, thus allowing wind state information to be represented in time series form.

[0039] As an example, see Figure 2 The diagram illustrates the process of determining wind components. The target train moves along the track, thus its position changes. Sampling points are provided at times T, T+1, T+2, T+3, and T+4. Since the track is terrain-dependent, the base direction differs at different times. Specifically, the base direction at time T is D1, at time T+1 it is D2, at time T+2 it is D3, at time T+3 it is D4, and at time T+4 it is D5. Furthermore, because real-time weather information, especially wind direction and speed, often cannot achieve extremely high positional accuracy, this disclosure uses a set of wind direction and speed to uniformly represent real-time weather information within a region. Specifically, wind speed and direction can be represented by the wind vector W. Therefore, for times T, T+1, T+2, T+3, and T+4, the wind vector can be decomposed to obtain wind component V1 corresponding to time T, wind component V2 corresponding to time T+1, wind component V3 corresponding to time T+2, wind component V4 corresponding to time T+3, and wind component V5 corresponding to time T+4. Based on wind components V1, V2, V3, V4, and V5, wind state information in time series form, including binary value groups composed of wind speed component values ​​and wind direction, can be obtained.

[0040] Step 103: Generate predicted vibration signals based on real-time vehicle status information, track status information, wind status information, basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model.

[0041] In some embodiments, the aforementioned execution entity can generate a predicted vibration signal based on real-time vehicle status information, track status information, wind status information, basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model.

[0042] Among them, the basic vehicle information represents the basic parameters of the target railcar and the parameters of the cargo it carries. Specifically, the basic vehicle information may include: vehicle net weight, number of trailers, number of containers, container cargo loading capacity, and type of cargo loaded in the containers.

[0043] The vibration and noise signal prediction model can be a predictive model used to predict vibration signals induced by vehicle status, track status, and weather status. Specifically, since real-time vehicle status information, track status information, and wind status information are all in time series form, the vibration and noise signal prediction model can adopt an LSTM (Long Short-Term Memory) model.

[0044] In practice, the vibration noise signal prediction model can be based on the values ​​contained in real-time vehicle status information, track status information, and wind status information at the same time. The predicted vibration signal value at that time can be obtained. Through continuous prediction, the predicted vibration signal can be obtained.

[0045] As an example, the predicted vibration signal value at time T can be obtained by taking the six-element value group of real-time vehicle status information at time T, the three-element value group of track status information at time T, and the two-element value group of wind status information at time T as inputs.

[0046] Optionally, the vibration and noise signal prediction model includes: a vehicle state feature extraction module, a track state feature extraction module, a wind state feature extraction module, a vehicle basic information encoding module, and a vibration and noise signal predictor.

[0047] The vehicle state feature extraction module, track state feature extraction module, and wind state feature extraction module are set up in parallel to extract features from the time-series signals at the same moment.

[0048] In some optional implementations of certain embodiments, the execution entity generates a predicted vibration signal based on the real-time vehicle status information, the track status information, the wind status information, the vehicle basic information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model, including: Step S1: The vehicle basic information is encoded using the vehicle basic information encoding module to obtain vehicle basic information features.

[0049] As an example, see Figure 3The diagram illustrates the generation process of the predicted vibration signal. The vehicle basic information encoding module 304 is used to encode the vehicle basic information. Specifically, since the vehicle basic information may include: vehicle net weight, number of trailers, number of containers, container cargo loading capacity, and container cargo type, the vehicle basic information changes relatively little during the target railcar's cargo transportation process (e.g., changes in the number of trailers, containers, and container cargo loading capacity occur after loading and unloading at the station). Furthermore, since the vehicle basic information is text-based data, the vehicle basic information encoding module 304 adopts the Text-Embedding-v4 model. The Text-Embedding-v4 model belongs to the Qwen3-Embedding model series, which supports multiple long vector dimensions (e.g., 2048, 1536, 1024), thus encompassing richer information features. Furthermore, considering that the vehicle basic information changes relatively infrequently, the vehicle basic information features obtained after being encoded by the vehicle basic information encoding module 304 are stored in a cache. When the vehicle basic information changes, the vehicle basic information encoding module 304 will re-encode the vehicle basic information and replace the old vehicle basic information stored in the cache with the new vehicle basic information features, thereby reducing the ineffective use of computing resources.

[0050] Step S2: By using the vehicle state feature extraction module, the track state feature extraction module, and the wind state feature extraction module in parallel, extract the local features corresponding to the real-time vehicle state information, the track state information, and the wind state information within the variable sliding window to obtain the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence.

[0051] The variable sliding window's length dynamically changes based on the aforementioned real-time vehicle status information. In practice, when extracting features from time-series data, a sliding window with a fixed length is often used. However, considering that in freight transportation scenarios, changes in vehicle status information (speed, acceleration), such as when a target railcar is traveling at high speed, even subtle changes in track or wind direction and speed can cause severe vibrations, and vice versa. Therefore, this disclosure dynamically adjusts the variable sliding window's length based on vehicle status information. When the vehicle speed or acceleration represented by the vehicle status information is small, the variable sliding window's length increases (increasing the observation scale), thereby capturing subtle vibration changes over a long period through a large sliding window. When the vehicle speed or acceleration represented by the vehicle status information is small, the variable sliding window's length can be decreased (reducing the observation scale), thereby achieving timely and effective capture of severe vibrations over a short period. In particular, to avoid extreme cases where the window length is too small or too large, a window length threshold (minimum window length threshold W) is set for the variable sliding window. min and the maximum window length threshold W max ), at the minimum window length threshold W min and the maximum window length threshold W max Within the defined threshold range, a mapping relationship between vehicle state information (speed and acceleration) and window length is pre-built to quickly adjust the window length.

[0052] As an example, see further. Figure 3The variable sliding window can have a length of 5 (note that this window length is for illustrative purposes only and is not a limitation on the window length). In this case, the execution entity will, under the premise of time alignment, input the local features corresponding to the real-time vehicle state information, the track state information, and the wind state information within the variable sliding window into the vehicle state feature extraction module 301, the track state feature extraction module 302, and the wind state feature extraction module 303 in parallel. Specifically, the vehicle state feature extraction module 301 includes 5 serially connected convolutional blocks and 1 fully connected layer. The track state feature extraction module 302 includes 3 serially connected convolutional blocks and 1 fully connected layer. The wind state feature extraction module 303 includes 1 serially connected convolutional block and 1 fully connected layer. The convolutional blocks include a convolutional layer, a ReLU activation function, and a Max-Pool layer. In particular, the convolutional layers in both the vehicle state feature extraction module 301 and the track state feature extraction module 302 use 3×3 convolutional kernels. The wind state feature extraction module 303 uses 1×1 convolutional kernels. The vehicle state feature extraction module 301 and the aforementioned track state feature extraction module 302 both use 3×3 convolutional kernels because 3×3 convolutional kernels have a higher computational density. The wind state feature extraction module 303 uses 1×1 convolutional kernels because it is limited by the feature dimension constraints corresponding to wind state information; therefore, it is adjusted to a 1×1 convolutional kernel. Furthermore, the vehicle state feature extraction module 301, the aforementioned track state feature extraction module 302, and the aforementioned wind state feature extraction module 303 all adopt linear network structures, thus possessing high architectural flexibility and facilitating adaptive adjustments to the number of network layers according to requirements.

[0053] Step S3: Overlay the above-mentioned vehicle basic information features, the above-mentioned local vehicle state feature sequence, the above-mentioned local track state feature sequence, and the above-mentioned local wind state feature sequence to obtain the overlaid feature sequence.

[0054] In practice, the basic information features of the vehicle, as well as the local vehicle state features, local track state features, and local wind state features corresponding to the same variable sliding window, can be spliced ​​together according to a preset splicing order to obtain the corresponding superimposed feature sequence.

[0055] As an example, see Figure 4 The diagram shows the generation process of the predicted vibration signal. Taking the local vehicle state feature 402, local track state feature 403, and local wind state feature 404 corresponding to the sliding time window at time T as an example, the vehicle basic information feature 401, local vehicle state feature 402, local track state feature 403, and local wind state feature 404 can be spliced ​​together according to a preset splicing order to obtain the corresponding superimposed feature 405.

[0056] In some optional implementations of certain embodiments, the above-mentioned feature superposition of the vehicle basic information features, the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence yields a superimposed feature sequence including: Step S31: For each local vehicle state feature in the above local vehicle state feature sequence, perform the following processing steps: Step S311: Determine the feature forgetting probability corresponding to the above local vehicle state features.

[0057] The aforementioned feature forgetting probability represents the probability of whether vehicle basic information features participate in feature superposition. When vehicle basic information features participate in feature superposition, the feature forgetting probability is reset to the default probability (10%). When vehicle basic information features do not participate in feature superposition, the feature forgetting probability increases. Specifically, the value range of the feature forgetting probability is [40%, 100%], increasing in increments of 10%. In particular, a higher feature forgetting probability indicates a higher probability of vehicle basic information features participating in feature superposition. Especially, the reason why the lower limit of the value range is not set to 0% is that: the feature forgetting probability is a probability value, and there are situations where the feature forgetting probability is high, but the vehicle basic information features still do not participate in feature superposition. If the lower limit is set to 0%, an extreme case is more likely to occur, causing the vehicle basic information features to continuously not participate in feature superposition, thus leading to feature forgetting.

[0058] In practice, continue to refer to Figure 4Based on the aforementioned analysis of vehicle basic information, its change frequency is small, and therefore the change frequency of the corresponding vehicle basic information features is also small. The conventional approach is to use vehicle basic information as the initial input (the initial input corresponding to neuron t1), but there are the following problems: (1) Since the vibration noise signal predictor adopts a recurrent neural network architecture, it will have a feature forgetting problem, which will lead to feature forgetting of vehicle basic information features; (2) Vehicle basic information is not unchanging, but its change frequency is small. If it is only used as the initial input, the change of vehicle basic information will be ignored, thus affecting the subsequent prediction accuracy. Therefore, on this basis, vehicle basic information features can be used as part of the input of each neuron and superimposed with local vehicle state features, local vehicle state features and local wind state features. At this time, new problems will arise: (3) Since the change frequency of vehicle basic information features is small, combined with the characteristics of the recurrent neural network architecture, the next neuron will receive the features processed by the previous neuron (which may already contain the description of vehicle basic information features). If vehicle basic information features are used as the input of each neuron, there will be a large number of duplicate feature extractions, which will lead to a large number of invalid feature calculations. Therefore, this disclosure sets a feature forgetting probability to control whether the vehicle basic information features are superimposed with local vehicle state features and local wind state features. When the feature forgetting probability representation needs to participate in feature superposition, the vehicle basic information features, local vehicle state features, local track state features, and local wind state features (stored in the cache) are superimposed to obtain the superimposed features. When the feature forgetting probability representation does not need to participate in feature superposition, in order to ensure the consistency of the dimensionality of the neuron input features, a 0 matrix with the same feature dimension as the vehicle basic information features can be superimposed with the local vehicle state features, local track state features, and local wind state features to obtain the superimposed features. Simultaneously, when the vehicle basic information features participate in feature superposition, the feature forgetting probability is initialized to the default probability; when the vehicle basic information features do not participate in feature superposition, the feature forgetting probability increases by 10% increments with each neuron.

[0059] As an example, taking neuron t1 as an example, the corresponding feature forgetting probability is the default probability (40%). When the feature forgetting probability of neuron t1 indicates that the vehicle basic information feature needs to participate in feature superposition, the vehicle basic information feature, local vehicle state feature, local track state feature, and local wind state feature are superimposed to obtain the superimposed feature. At this time, the feature forgetting probability of neuron t2 is the default probability. When the feature forgetting probability of neuron t1 indicates that the vehicle basic information feature does not need to participate in feature superposition, a 0 matrix with the same feature dimension as the vehicle basic information feature is superimposed with the local vehicle state feature, local track state feature, and local wind state feature to obtain the superimposed feature. At the same time, the feature forgetting probability of neuron t2 is updated to the default probability (40%) + 10% (50%). In particular, when the feature forgetting probability of a neuron is 100%, the feature forgetting probability of the next neuron is also initialized to the default probability. For example, if the feature forgetting probability of the previous neuron of neuron tn is 100%, it indicates that the vehicle basic information feature must participate in the construction of the superimposed feature of the previous neuron of neuron tn. Therefore, the feature forgetting probability of neuron tn is initialized to the default probability.

[0060] Step S4: Based on the superimposed feature sequence and the vibration noise signal predictor, the predicted vibration signal is obtained.

[0061] The vibration noise signal predictor adopts a recurrent neural network architecture, specifically a "many to many" mode, which can generate the predicted vibration signal value corresponding to each sliding time window. By predicting the vibration noise signal through the superimposed features corresponding to each sliding time window, a continuous predicted vibration signal can be obtained.

[0062] As an example, see further. Figure 4 The vibration noise signal predictor includes neurons t1, t2, t3, t4, ..., tn. Taking neuron t1 as an example, the superimposed feature 405 corresponding to the sliding time window at time T is input into neuron t1 to obtain the predicted vibration signal value corresponding to the sliding window at time T. By concatenating the predicted vibration signal values ​​output by multiple neurons (neurons t1, t2, t3, t4, ..., tn) in time sequence, the predicted vibration signal 406 is obtained.

[0063] In some alternative implementations of certain embodiments, the vibration noise signal prediction model is trained through the following steps: Step S1: Construct training samples and their corresponding sample labels.

[0064] The training samples include: historical vehicle status information, historical track status information, historical wind status information, and historical vehicle basic information. The sample labels corresponding to the training samples correspond to historical vibration signals collected by a triaxial sensor located on the bottom of the container.

[0065] In practice, factors such as vehicle operation, track conditions, and weather can cause vibrations to be transmitted from the container to the cargo, thus inducing cargo vibration. Therefore, capturing the noise generated by these factors individually is equivalent to capturing them collectively. Simultaneously, the container, acting as an intermediary, allows for the direct capture of vibration noise (noise vibration signals) caused by these factors by installing a triaxial sensor on it. Therefore, historical vehicle status information, historical track status information, historical wind status information, and historical vehicle basic information can be used as training samples. The vibration signal values ​​from historical vibration signals collected by the triaxial sensor located on the container floor that correspond to the training samples can be used as sample labels, thus constructing the training samples and sample labels.

[0066] Specifically, triaxial sensors can be installed on the floor of a small number of containers and on the cargo pallets to continuously generate new training samples for iterative model updates. This method still has a significant cost advantage compared to installing two sets of triaxial sensors on all containers.

[0067] Step S2: Initialize the model parameters for the vibration noise signal prediction model that has not been trained.

[0068] In practice, the model parameters of each neuron in the vibration noise signal prediction model can be randomly initialized.

[0069] Step S3: In response to the completion of initialization, supervised model training is performed on the untrained vibration noise signal prediction model based on the training samples and corresponding sample labels to obtain the vibration noise signal prediction model.

[0070] In practice, a supervised model for predicting vibration and noise signals can be trained using forward and backward propagation based on training samples and their corresponding labels, thus obtaining the predicted model. Specifically, convergence or the number of training iterations can be used as the condition for training to end.

[0071] Step 104: Based on the predicted vibration signal, perform noise removal on the real-time vibration signal to obtain the removed vibration signal.

[0072] In some embodiments, the aforementioned execution entity can perform noise stripping on the real-time vibration signal based on the predicted vibration signal to obtain the stripped vibration signal.

[0073] In practice, the following formula can be used to remove noise from real-time vibration signals to obtain the noise-removed vibration signal: , in, Indicates the serial number. This indicates the vibration signal after peeling. Represents real-time vibration signals. Indicates the predicted vibration signal. Indicates the vibration signal after peeling. One signal value, Indicates the first [unit] in the real-time vibration signal One signal value, Indicating the predicted vibration signal, the first... Each signal value.

[0074] In practice, weighting coefficients can be set for the predicted vibration signal to control the influence of noise signal stripping. The value range of the weighting coefficient corresponding to the predicted vibration signal can be (0, 1).

[0075] Step 105: Determine the vibration state of the cargo inside the container based on the vibration signal after stripping.

[0076] In some embodiments, the aforementioned executing entity can determine the vibration state of the cargo inside the container based on the vibration signal after stripping. The vibration state can characterize the vibration changes of the cargo, specifically, it can be characterized by the vibration amplitude.

[0077] In practice, the vibration amplitude can be solved using the Archid formula. In particular, when solving for the vibration state, the stripped vibration signal can be further processed by filtering and smoothing to reduce noise interference.

[0078] Step 106: Determine whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located.

[0079] In some embodiments, the aforementioned execution entity may determine whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located.

[0080] The dynamic warning threshold is dynamically set according to changes in the region. This is because different regions are induced by different abnormal vibrations. Therefore, this disclosure abandons the fixed warning threshold method and sets regional thresholds (dynamic warning thresholds) for areas containing tracks to improve the accuracy of warnings and avoid false alarms and missed alarms.

[0081] In practice, dynamic early warning values ​​are set for different regions. For example, they can be set based on expert experience. Alternatively, they can be automatically updated based on the historical vibration state sequence corresponding to that region. When the vibration state... In some optional implementations of certain embodiments, the aforementioned dynamic warning threshold is generated through the following steps: Step S1: Obtain the historical vibration state sequence corresponding to the area where the target railcar is located.

[0082] In practice, the historical vibration states of a target tram passing through an area at different times can be obtained as a historical vibration state sequence. Specifically, to ensure the effectiveness of the determined dynamic warning threshold, for the same area, the historical vibration states of multiple different trams passing through the area at different times can also be collected as a historical vibration state sequence. The historical vibration state can be characterized by the historical vibration amplitude.

[0083] Step S2: Construct a vibration feature matrix based on the above historical vibration state sequence.

[0084] As an example, suppose a transportation route Depend on Composed of (geographical) regions, therefore, transportation routes This can be expressed as: ,in, Indicates the sequence number. For regions... (For example, for the area where the target tram is located), settings can be set. One sampling point. For the region. It is possible to collect data on multiple train services passing through the area. Therefore, for the area... Constructed historical vibration state sequence It can be represented as: , in, This indicates that the data for the first shift is in Historical vibration data collected at each sampling point This indicates that the data for the second shift is in Historical vibration data collected at each sampling point Indicates the first Data for each shift Historical vibration data collected from each sampling point.

[0085] Step S3: Determine the reference vibration parameters based on the above vibration characteristic matrix.

[0086] The reference vibration parameters include the mean vibration amplitude and the standard deviation of vibration amplitude.

[0087] As an example, continue with the region For example, the average vibration amplitude can be expressed by the following formula: , in, and All of these represent serial numbers.

[0088] area The corresponding standard deviation of the vibration amplitude can be expressed by the following formula: .

[0089] Step S4: Determine the above dynamic early warning threshold based on the reference vibration parameters and the pre-constructed threshold determination function.

[0090] As an example, continue with the region For example, its corresponding dynamic early warning threshold can be expressed as: Specifically, the dynamic early warning threshold consists of an upper threshold and a lower threshold.

[0091] The threshold determination function is characterized by the following formula: , in, Indicates the dynamic early warning threshold. and All are weighting coefficients. This is the lower limit of the safety threshold to avoid false negatives. In particular, the weighting coefficients The "1.5" and "0.2" in the formula are empirical values. Weighting coefficients. and weighting coefficients It is a fixed value.

[0092] Optionally, as vibration states continue to occur (when a train passes through the area), a sliding window approach can be used to continue collecting historical vibration state sequences, and the dynamic warning threshold can be updated according to steps S1 to S5 above.

[0093] In practice, the deviation of vibration quantity corresponding to the vibration state can first be calculated using the following formula: , in, Indicates the vibration offset. The amplitude of vibration represents the vibration state.

[0094] Secondly, a vibration anomaly warning will be initiated when the vibration state and vibration offset meet the following conditions: , in, This represents the sensitivity coefficient, with a default value of 1.8. It can also be adjusted adaptively depending on the type of goods.

[0095] In some optional implementations of some embodiments, the above method further includes: Step S1: In response to initiating a vibration anomaly warning, determine the target decision path based on the above track status information, the above wind status information, the above vehicle foundation information, and the above track status information.

[0096] The target decision path is the area where the target tram is located (e.g., by region). (For example, the decision path within the corresponding vehicle control decision tree.)

[0097] In practice, for the area where the target tram is located (e.g., by area) Taking this as an example, a control decision tree can be constructed by combining factors such as track conditions (track deflection angle, track elevation), weather (wind speed, wind direction), vehicle basic information (vehicle net weight, number of trailers, number of containers, container cargo loading capacity, and container cargo type), as well as their corresponding safe operating speed and safe operating acceleration. Specifically, given the known values ​​of the aforementioned track condition information, wind condition information, vehicle basic information, and track condition information, a satisfactory decision path can be determined through tree traversal, serving as the target decision path.

[0098] Step S2: Based on the real-time vehicle status information mentioned above, update the decision boundary corresponding to the target decision path to obtain the updated vehicle control decision tree.

[0099] In practice, when the real-time vehicle status information (acceleration and speed) is less than the safe operating speed and safe operating acceleration corresponding to the target decision path, the real-time vehicle status information (acceleration and speed) is updated to the upper limit of the safe operating speed and safe operating acceleration corresponding to the target decision path. This method continuously optimizes the safe operating speed and safe operating acceleration for the area, guiding the speed of rail vehicles when passing through that area, thereby reducing the probability of cargo vibration.

[0100] Step S3: In response to the re-entry of the target railcar, generate vehicle control suggestion information based on the updated vehicle control decision tree.

[0101] In practice, when the target railcar re-enters the area (e.g., within the area...), For example, when retrieving track status information, wind status information, vehicle basic information, and track status information, the updated vehicle control decision tree can be traversed to determine the corresponding safe operating speed and safe operating acceleration as vehicle control suggestion information.

[0102] Step S4: Send the above vehicle control suggestion information to the control terminal corresponding to the target tram.

[0103] The control terminal can be the visual control terminal corresponding to the railway locomotive of the target rail train.

[0104] The above-described embodiments of this disclosure have the following beneficial effects: The method for monitoring the state of goods inside a container, as described in some embodiments of this disclosure, achieves highly accurate vibration anomaly early warning. Specifically, the reason for poor accuracy in vibration anomaly early warning is that the container, as an intermediary, transmits vibrations generated by vehicle movement to the goods. This vibration is both the cause of vibration in the goods and noise that interferes with the monitoring of abnormal vibrations in the goods, as the noise vibration signal is not effectively separated. Based on this, the method for monitoring the state of goods inside a container, as described in some embodiments of this disclosure, firstly acquires real-time vibration signals, real-time vehicle state information corresponding to the target railcar, and track state information. The real-time vibration signals are collected by a triaxial sensor installed on a cargo pallet used to carry the goods, and the cargo pallet is placed inside the container transported by the target railcar. Secondly, based on the aforementioned real-time vehicle state information, track state information, and real-time weather information of the area where the target railcar is located, wind state information is determined. The wind state information represents the wind component perpendicular to the direction of travel of the target railcar. In practice, crosswinds can cause trams to sway and induce vibrations during operation. Furthermore, the degree of crosswind influence varies at different speeds. Therefore, this disclosure combines real-time vehicle status information, track status information, and real-time weather information to decompose the crosswinds (wind status information) that may induce vehicle vibrations. Then, based on the aforementioned real-time vehicle status information, track status information, wind status information, the basic vehicle information corresponding to the target tram, and a pre-trained vibration noise signal prediction model, a predicted vibration signal is generated. In practice, the train body, containers, cargo pallets, and cargo are often rigidly connected. Factors such as vehicle operation, track status, and weather can cause vibrations to be transmitted to the cargo through the containers, thus inducing cargo vibrations. Since the container is the intermediate body for vibration transmission, and the vibrations generated by the above factors can all be understood as noise vibrations, to reduce computational and data processing complexity, this disclosure treats all the above factors as a whole for predicting the (noise) vibration signal. In practice, noise and vibration signals can also be directly monitored by setting up another set of triaxial sensors at the bottom of the container. However, compared to the method disclosed herein, the number of sensors required increases exponentially, resulting in a significant increase in hardware and calibration costs. This is especially true given the large number of containers in use, where costs surge. Therefore, this disclosure employs an algorithmic prediction method to predict noise and vibration signals, thereby reducing hardware costs. Furthermore, based on the predicted vibration signal, noise is stripped from the real-time vibration signal to obtain the stripped vibration signal. This achieves direct noise and vibration signal stripping. Next, based on the stripped vibration signal, the vibration state of the cargo inside the container is determined. This maps the stripped vibration signal to an index-based vibration state.Finally, based on the aforementioned vibration state and the dynamic warning threshold corresponding to the area where the target tram is located, it is determined whether to initiate a vibration anomaly warning. In particular, considering the differences in environmental factors in different areas, this disclosure recognizes that setting a fixed threshold results in a high false alarm rate. Therefore, a method of setting dynamic thresholds for different areas is adopted to improve the accuracy of the warning. In summary, this method achieves highly accurate vibration anomaly warnings.

[0105] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a status monitoring device for goods inside a container, these device embodiments being similar to... Figure 1 Corresponding to the method embodiments shown, the status monitoring device for cargo inside a container can be specifically applied to various electronic devices.

[0106] like Figure 5 As shown, a state monitoring device 500 for cargo inside a container, according to some embodiments, includes: an acquisition unit 501, a first determination unit 502, a generation unit 503, a noise signal stripping unit 504, a second determination unit 505, and a third determination unit 506. The acquisition unit 501 is configured to acquire real-time vibration signals, real-time vehicle state information corresponding to the target railcar, and track state information. The real-time vibration signals are collected by a triaxial sensor mounted on a cargo pallet used for cargo transport, and the cargo pallet is placed inside the container carried by the target railcar. The first determination unit 502 is configured to determine wind state information based on the real-time vehicle state information, the track state information, and real-time weather information of the area where the target railcar is located. The wind state information represents the wind component perpendicular to the direction of travel of the target railcar. The generation unit 503 is configured to determine wind state information based on the real-time vehicle state information, the track state information, and real-time weather information of the area where the target railcar is located. The system generates a predicted vibration signal based on the status information, track status information, wind status information, vehicle basic information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model. A noise signal stripping unit 504 is configured to strip noise from the real-time vibration signal based on the predicted vibration signal to obtain a stripped vibration signal. A second determining unit 505 is configured to determine the vibration state of the cargo inside the container based on the stripped vibration signal. A third determining unit 506 is configured to determine whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located.

[0107] It is understandable that the units described in the status monitoring device 500 for cargo inside the container are related to the reference... Figure 1The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the condition monitoring device 500 for cargo inside a container and the units contained therein, and will not be repeated here.

[0108] The following is for reference. Figure 6 It illustrates a schematic diagram of the structure of an electronic device (e.g., a computing device) suitable for implementing some embodiments of the present disclosure. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of this disclosure. Figure 6 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system and a computer program. The computer program includes program instructions, which, when executed, cause the processor to perform any of the methods described above. The processor provides computational and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the execution of the computer program in the non-volatile storage medium; when executed by the processor, the computer program causes the processor to perform any of the methods described above. The network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present disclosure and does not constitute a limitation on the computer device to which the present disclosure is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0109] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0110] In one embodiment, the processor is configured to run a computer program stored in a memory to perform the following steps: acquiring real-time vibration signals, real-time vehicle status information and track status information corresponding to the target railcar, wherein the real-time vibration signals are collected by a triaxial sensor mounted on a cargo pallet used for carrying goods, and the cargo pallet is placed inside a container carried by the target railcar; determining wind status information based on the real-time vehicle status information, the track status information and the real-time weather information of the area where the target railcar is located, wherein the wind status information represents the wind component perpendicular to the direction of travel of the target railcar; generating a predicted vibration signal based on the real-time vehicle status information, the track status information, the wind status information, the basic vehicle information corresponding to the target railcar and a pre-trained vibration noise signal prediction model; performing noise signal stripping on the real-time vibration signal based on the predicted vibration signal to obtain a stripped vibration signal; determining the vibration state of the goods inside the container based on the stripped vibration signal; and determining whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located.

[0111] This disclosure also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, and the method implemented when the program instructions are executed can be referred to the various embodiments of the methods described above.

[0112] The aforementioned computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. Alternatively, the aforementioned computer-readable storage medium may be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0113] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0114] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for monitoring the status of goods inside a container, characterized in that, include: The system acquires real-time vibration signals, real-time vehicle status information and track status information corresponding to the target railcar. The real-time vibration signals are collected by a three-axis sensor installed on a cargo pallet used to carry the cargo, which is placed inside a container carried by the target railcar. Based on the real-time vehicle status information, the track status information, and the real-time weather information of the area where the target railcar is located, wind status information is determined, wherein the wind status information represents the wind component perpendicular to the direction of travel of the target railcar. Based on the real-time vehicle status information, the track status information, the wind status information, the basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model, a predicted vibration signal is generated. The vibration and noise signal prediction model includes: a vehicle status feature extraction module, a track status feature extraction module, a wind status feature extraction module, a vehicle basic information encoding module, and a vibration and noise signal predictor. The vehicle status feature extraction module, the track status feature extraction module, and the wind status feature extraction module are set in parallel to perform parallel feature extraction on the time-series signal at the same moment. Based on the predicted vibration signal, noise signal stripping is performed on the real-time vibration signal to obtain the stripped vibration signal; Based on the vibration signal after stripping, the vibration state of the goods inside the container is determined; Based on the vibration state and the dynamic early warning threshold corresponding to the area where the target railcar is located, it is determined whether to initiate a vibration anomaly warning, wherein... The step of generating a predicted vibration signal based on the real-time vehicle status information, the track status information, the wind status information, the basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model includes: The vehicle basic information is encoded using the vehicle basic information encoding module to obtain vehicle basic information features. The vehicle state feature extraction module, the track state feature extraction module, and the wind state feature extraction module are used in parallel to extract the local features corresponding to the real-time vehicle state information, the track state information, and the wind state information within a variable sliding window, thereby obtaining a local vehicle state feature sequence, a local track state feature sequence, and a local wind state feature sequence. The window length of the variable sliding window changes dynamically according to the real-time vehicle state information. The vehicle basic information features, the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence are superimposed to obtain the superimposed feature sequence. The predicted vibration signal is obtained based on the superimposed feature sequence and the vibration noise signal predictor, wherein, The method of superimposing the vehicle basic information features, the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence to obtain the superimposed feature sequence includes: For each local vehicle state feature in the local vehicle state feature sequence, the following processing steps are performed: Determine the feature forgetting probability corresponding to the local vehicle state feature, wherein the feature forgetting probability represents the probability of whether the vehicle basic information feature participates in feature superposition. When the vehicle basic information feature participates in feature superposition, the feature forgetting probability is reset to the default probability. When the vehicle basic information feature does not participate in feature superposition, the feature forgetting probability increases. Based on the feature forgetting probability, the vehicle basic information features are superimposed with the local vehicle state features, the local track state features corresponding to the local vehicle state features, and the local wind state features corresponding to the local vehicle state features to obtain the superimposed features.

2. The method according to claim 1, characterized in that, The method further includes: In response to initiating a vibration anomaly warning, a target decision path is determined based on the wind state information, the vehicle basic information, and the track state information, wherein the target decision path is the decision path within the vehicle control decision tree corresponding to the area where the target tram is located; Based on the real-time vehicle status information, the decision boundary corresponding to the target decision path is updated to obtain the updated vehicle control decision tree. In response to the re-entry of the target railcar, vehicle control suggestion information is generated based on the updated vehicle control decision tree; The vehicle control suggestion information is sent to the control terminal corresponding to the target tram.

3. The method according to claim 2, characterized in that, The step of determining wind status information based on the real-time vehicle status information, the track status information, and the real-time weather information of the area where the target tram is located includes: Based on the real-time vehicle status information, a basic direction is determined, wherein the basic direction is in the same direction as the movement direction of the target railcar; Based on the track status information, determine the plane where the track is located; Based on the real-time weather information, the wind component perpendicular to the base direction within the plane where the track is located is determined and used as the wind state information.

4. The method according to claim 3, characterized in that, The dynamic early warning threshold is generated through the following steps: Obtain the historical vibration state sequence corresponding to the area where the target tram is located; Based on the historical vibration state sequence, a vibration feature matrix is ​​constructed; Based on the vibration characteristic matrix, determine the reference vibration parameters; The dynamic early warning threshold is determined based on the reference vibration parameters and a pre-constructed threshold determination function.

5. The method according to claim 4, characterized in that, The vibration noise signal prediction model is trained through the following steps: Construct training samples and corresponding sample labels. The training samples include: historical vehicle status information, historical track status information, historical wind status information, and historical vehicle basic information. The sample labels corresponding to the training samples correspond to historical vibration signals collected by a triaxial sensor set on the bottom of the container. Initialize the model parameters for the vibration noise signal prediction model that has not been trained. Upon completion of initialization, supervised model training is performed on the untrained vibration noise signal prediction model based on the training samples and their corresponding labels to obtain the vibration noise signal prediction model.

6. A status monitoring device for goods inside a container, characterized in that, include: The acquisition unit is configured to acquire real-time vibration signals, real-time vehicle status information and track status information corresponding to the target railcar. The real-time vibration signals are collected by a triaxial sensor installed on a cargo pallet used for cargo carrying. The cargo pallet is installed inside a container carried by the target railcar. The first determining unit is configured to determine wind state information based on the real-time vehicle state information, the track state information, and the real-time weather information of the area where the target railcar is located, wherein the wind state information represents the wind component perpendicular to the direction of travel of the target railcar. The generation unit is configured to generate a predicted vibration signal based on the real-time vehicle status information, the track status information, the wind status information, the vehicle basic information corresponding to the target railcar, and a pre-trained vibration noise signal prediction model. The vibration noise signal prediction model includes: a vehicle status feature extraction module, a track status feature extraction module, a wind status feature extraction module, a vehicle basic information encoding module, and a vibration noise signal predictor. The vehicle status feature extraction module, the track status feature extraction module, and the wind status feature extraction module are set in parallel to perform parallel feature extraction on the time-series signal at the same moment. The noise signal stripping unit is configured to strip the noise signal from the real-time vibration signal based on the predicted vibration signal to obtain the stripped vibration signal. The second determining unit is configured to determine the vibration state of the cargo inside the container based on the vibration signal after stripping. The third determining unit is configured to determine whether to initiate a vibration anomaly warning based on the vibration state and the dynamic warning threshold corresponding to the area where the target railcar is located. The step of generating a predicted vibration signal based on the real-time vehicle status information, the track status information, the wind status information, the basic vehicle information corresponding to the target railcar, and a pre-trained vibration and noise signal prediction model includes: The vehicle basic information is encoded using the vehicle basic information encoding module to obtain vehicle basic information features. The vehicle state feature extraction module, the track state feature extraction module, and the wind state feature extraction module are used in parallel to extract the local features corresponding to the real-time vehicle state information, the track state information, and the wind state information within a variable sliding window, thereby obtaining a local vehicle state feature sequence, a local track state feature sequence, and a local wind state feature sequence. The window length of the variable sliding window changes dynamically according to the real-time vehicle state information. The vehicle basic information features, the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence are superimposed to obtain the superimposed feature sequence. The predicted vibration signal is obtained based on the superimposed feature sequence and the vibration noise signal predictor, wherein, The method of superimposing the vehicle basic information features, the local vehicle state feature sequence, the local track state feature sequence, and the local wind state feature sequence to obtain the superimposed feature sequence includes: For each local vehicle state feature in the local vehicle state feature sequence, the following processing steps are performed: Determine the feature forgetting probability corresponding to the local vehicle state feature, wherein the feature forgetting probability represents the probability of whether the vehicle basic information feature participates in feature superposition. When the vehicle basic information feature participates in feature superposition, the feature forgetting probability is reset to the default probability. When the vehicle basic information feature does not participate in feature superposition, the feature forgetting probability increases. Based on the feature forgetting probability, the vehicle basic information features are superimposed with the local vehicle state features, the local track state features corresponding to the local vehicle state features, and the local wind state features corresponding to the local vehicle state features to obtain the superimposed features.

7. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 5.

8. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images