A water conservancy gate opening monitoring method and system based on the Internet of Things

Abstract

The application discloses a water conservancy project gate opening monitoring method and system based on an Internet of Things, relates to the technical field of water conservancy gate opening monitoring, and comprises the following steps: acquiring continuous opening angle data of a target gate in a preset monitoring period and a structure image sequence of a preset part of the target gate, constructing a dynamic opening feature sequence of the target gate based on the opening angle data, and synchronizing the structure image sequence and the dynamic opening feature sequence in time; processing the structure image sequence based on a preset few-sample deep learning model; and judging whether the visual damage confidence is lower than a first preset threshold value or not; the method has the beneficial effect that the best maintenance window is effectively avoided from being missed due to prediction lag, and the safe operation of the water conservancy project is ensured.

Description

Technical Field

[0001] This invention relates to the field of water conservancy gate opening monitoring technology, and in particular to a water conservancy project gate opening monitoring method and system based on the Internet of Things. Background Technology

[0002] As critical operational equipment in water conservancy projects, the safety and reliability of gates directly affect flood control and drought relief, water resource allocation, and the safety of people's lives and property. With the rapid development of Internet of Things (IoT) technology, sensor-based automated monitoring systems have been widely applied in water conservancy projects to collect real-time dynamic operational data of gates, such as opening and closing status, opening time series, and operational loads. Simultaneously, to assess the structural health of gates and prevent sudden accidents, visual inspection methods are often combined to periodically or in real-time acquire images and perform damage analysis on preset parts of the gates. These monitoring methods, combined with modern data analysis techniques, provide an important data foundation for gate fault diagnosis and remaining life prediction.

[0003] However, existing monitoring systems for hydraulic engineering gates typically treat time-series monitoring indicators reflecting mechanical operating status and visual inspection information reflecting structural health status as two heterogeneous modal data points. Specifically, life prediction or mechanical wear assessment models are performed independently, while visual damage information obtained from structural image analysis is usually only used as an auxiliary structural health indicator, lacking an effective, cross-modal deep feedback and self-calibration mechanism. This directly leads to the inability of existing models to accurately capture the synergistic effects of faults when the mechanical performance of the gate is accelerated by initial structural defects. As a result, the remaining life prediction value is seriously lagging or significantly deviates from the actual degradation state, greatly increasing the difficulty of early warning of complex faults and potentially missing the optimal maintenance window, thus triggering significant safety risks. Summary of the Invention

[0004] In view of the above-mentioned prior art, this application is hereby filed. Embodiments of this application provide a method for monitoring the opening degree of gates in water conservancy projects based on the Internet of Things, including: Acquire continuous opening angle data of the target gate and structural image sequence of a preset part of the target gate within a preset monitoring period; construct a dynamic opening feature sequence of the target gate based on the opening angle data; the structural image sequence is time-synchronized with the dynamic opening feature sequence. Based on a preset few-shot deep learning model, the sequence of structural images is processed to obtain a visual damage confidence sequence and quantified damage features of a preset part of the gate. Determine whether the confidence level of the visual impairment is lower than a first preset threshold; If the judgment result is yes, then according to the quantitative damage characteristics, the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence are assigned weight adjustment factors, and the dynamic opening feature sequence is re-acquired; based on the dynamic opening feature sequence after weight adjustment, the remaining life prediction value of the target gate is corrected, and it is determined whether the corrected remaining life prediction value is lower than the second preset value.

[0005] According to one aspect of this application, an Internet of Things-based water conservancy project gate opening monitoring system is provided, comprising: a data acquisition module: used to acquire a continuous dynamic opening feature sequence of the target gate within a preset monitoring period, and a structural image sequence of a preset part of the target gate synchronized with the dynamic opening feature sequence in time; Data processing module: used to process the image sequence of the structure based on a preset few-shot deep learning model to obtain the visual damage confidence sequence and quantified damage features of preset parts of the gate; Judgment module: used to determine whether the confidence level of the visual impairment is lower than a first preset threshold; Self-calibration module: If the judgment result is yes, then according to the quantitative damage characteristics, the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence are assigned weight adjustment factors, and the dynamic opening feature sequence is re-acquired; based on the dynamic opening feature sequence after weight adjustment, the remaining life prediction value of the target gate is corrected, and it is determined whether the corrected remaining life prediction value is lower than the second preset value.

[0006] According to another aspect of this application, an electronic device is provided, including a memory and a processor, the memory being used to store computer-executable instructions, and the processor being used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method described above.

[0007] According to another aspect of this application, a computer storage medium is provided that stores computer-executable instructions thereon, which, when executed by a processor, implement the steps of the method described above.

[0008] Compared with existing technologies, the method and system for monitoring the gate opening of water conservancy projects based on the Internet of Things, according to an embodiment of this application, introduces a weight adjustment mechanism based on visual impairment into the feature sequence of mechanical operation status. This achieves cross-modal collaborative monitoring capabilities that are unmatched by traditional monitoring systems, solves the problem of heterogeneous data being independent in existing technologies, and enables the entire monitoring system to detect the accelerated degradation of mechanical performance caused by structural defects in advance. This greatly improves the early detection sensitivity and warning capability of the monitoring system for collaborative faults, significantly improves the accuracy and reliability of the remaining life prediction value, provides a more reliable and forward-looking decision-making basis for the inspection and maintenance of water conservancy projects, effectively avoids missing the best maintenance window due to prediction lag, and ensures the safe operation of water conservancy projects. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a schematic diagram of the overall framework and process structure of a water conservancy project gate opening monitoring method based on the Internet of Things according to the present invention.

[0011] Figure 2 This is a schematic diagram of the structural damage evidence collaborative fault risk assessment process for a water conservancy project gate opening monitoring method based on the Internet of Things, according to the present invention.

[0012] Figure 3 This is a schematic diagram illustrating the process of generating structural damage evidence for a method for monitoring the opening of gates in water conservancy projects based on the Internet of Things, according to the present invention. Detailed Implementation

[0013] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0014] Application Overview In traditional hydraulic engineering gate monitoring systems, two heterogeneous modal data—time-series monitoring indicators reflecting mechanical operating status (such as dynamic opening characteristic sequences) and visual inspection information reflecting structural health (such as structural image sequences and their quantitative damage characteristics)—are typically treated as independent monitoring components. Specifically, lifespan prediction or mechanical wear assessment models are run independently, while visual damage information obtained from structural image analysis is usually only used as an auxiliary structural health indicator, lacking an effective, cross-modal, in-depth feedback and self-calibration mechanism. For example, when a gate experiences accelerated mechanical performance deterioration due to initial structural defects (such as minor damage in predetermined locations), traditional methods cannot accurately capture the synergistic effects of the fault because their prediction models do not provide real-time weighted intervention for mechanical characteristics based on the degree of structural damage. Failure to address these issues will result in severely lagging or significantly deviating remaining lifespan predictions from the actual degradation state, greatly increasing the difficulty of early warning of complex faults and potentially missing the optimal maintenance window, leading to significant safety risks.

[0015] Example 1: Reference Figures 1-3 As one embodiment of the present invention, a method for monitoring the opening degree of gates in water conservancy projects based on the Internet of Things is provided: Figure 1 The illustration shows a method for monitoring the opening degree of a water conservancy project gate based on the Internet of Things according to an embodiment of this application, including: S1~S7.

[0016] S1: Acquire continuous opening angle data of the target gate and structural image sequence of a preset part of the target gate within a preset monitoring period. Construct a dynamic opening feature sequence of the target gate based on the opening angle data. The structural image sequence and the dynamic opening feature sequence are synchronized in time.

[0017] In step S1, the opening angle data is acquired by a high-precision angle sensor (such as an absolute encoder or a magnetostrictive displacement sensor) deployed on the target gate opening adjustment mechanism. Within a preset monitoring period, the opening angle data of the target gate is continuously acquired at a preset sampling frequency. The opening angle data is the angle value obtained by measuring the target gate in its fully open or fully closed state as the zero point reference. Structural image sequence acquisition: The structural image sequence of the preset part is acquired at a preset frame rate by an industrial-grade image acquisition device deployed at a preset part of the target gate (such as bearing seat, connecting ear plate, bolt connection, etc.).

[0018] It should be noted that in the monitoring method for the opening of gates in water conservancy projects, the preset locations refer to those key areas where mechanical stress is concentrated and fatigue damage or structural defects are prone to occur during gate operation. Damage to these areas has a direct or accelerating effect on the mechanical operation of the gate. The preset locations include, but are not limited to, the connection points of the transmission mechanism, the support and guide mechanism, and key stress concentration areas. By acquiring and analyzing images of these locations, it is possible to ensure that initial structural defects that have a direct impact on the deterioration of the mechanical operation can be captured, thereby providing accurate visual input for subsequent weight adjustment and life prediction correction.

[0019] Furthermore, a dynamic opening feature sequence of the target gate is constructed based on the opening angle data, including: performing time-domain, frequency-domain, and / or wavelet-domain analysis on continuous angle measurement data to extract a feature set characterizing the mechanical operating state of the target gate, including but not limited to: angle change rate (i.e., operating angular velocity), second-order angle change rate (i.e., operating angular acceleration), operating stability indicators (such as the standard deviation of the angle sequence and fluctuation coefficient), and frequency characteristics reflecting operating friction or wear; the extracted multidimensional feature vectors are arranged in chronological order to construct a dynamic opening feature sequence, which contains feature components reflecting the degree of mechanical wear and the degree of deterioration of the operating state of the target gate; Synchronizing the structural image sequence with the dynamic opening feature sequence in time refers to uniformly calibrating the acquisition timestamps of angle sensors and image acquisition devices through an IoT gateway or time server to ensure that at any monitoring moment, the acquired structural image and the constructed dynamic opening feature vector have a unified and accurate time correspondence.

[0020] S2: Based on a preset few-sample deep learning model, the structural image sequence is processed to obtain the visual damage confidence sequence and quantified damage features of preset parts of the gate.

[0021] In step S2, it should be noted that the acquisition of the visual impairment confidence sequence and quantified impairment features includes: A few-shot deep learning model is trained in advance using a small number of images containing damage to the target gate; This training method is not a traditional end-to-end deep learning training, but rather a meta-learning approach. It aims to enable few-shot deep learning models to learn how to quickly adapt to new damage categories. During the training phase, the model uses a small image dataset containing typical gate damage (such as cracks, corrosion, and loose connectors) for meta-training. Through this training, the few-shot deep learning model can extract a general feature encoder sensitive to structural damage characteristics. Therefore, in actual monitoring, only a very small number of new damage samples collected on-site are needed to quickly adapt to and accurately identify specific damage in the current environment. This solves the technical problem of obtaining large-scale damage samples for water conservancy project gates. Based on a pre-set few-shot deep learning model, reasoning is performed on each frame of the acquired structural image sequence, and the output is a probability value reflecting the damage to a pre-set part of the target gate as the visual damage confidence. After acquiring a time-synchronized sequence of structural images, this sub-step is responsible for determining in real time whether there are potential structural damages in the images. For each input frame of structural image, the few-shot deep learning model generates a feature vector through its feature encoder. Subsequently, the few-shot deep learning model calculates the distance or similarity between this feature vector and the pre-learned damage prototype vector. Finally, the few-shot deep learning model transforms this similarity using functions such as Softmax or Sigmoid to generate a probability value between 0 and 1. This probability value is the current visual damage confidence level. Few-shot deep learning models extract and output features from structural image sequences that reflect the type and severity of damage as quantified damage features; Quantitative damage features aim to provide more refined damage severity information that can be used for quantification calculations than a single confidence level. Specifically, it can be represented in two forms: First, a high-dimensional feature vector output by the feature encoder without the final activation layer, which contains semantic information (type) and intensity information (severity) of the damage; Second, a normalized damage severity index calculated based on the high-dimensional feature vector, such as the proportion of the damaged area to the entire preset area calculated by an image segmentation algorithm, or the average pixel length of the crack.

[0022] The reason this invention adopts this technical solution is to mandate that the system output two types of data after visually detecting structural damage: first, Visual Damage Confidence (VDC) to determine whether to initiate correction; and second, Quantified Damage Features (QDF) to quantify the severity of the damage. The necessity of this operation lies in the fact that only by transforming structural damage information from qualitative alarms into quantifiable features can it serve as a feedback bridge, influencing the weight adjustment of the dynamic opening feature sequence in real time, thereby enabling the remaining lifetime prediction to respond to the accelerated degradation trend caused by structural defects.

[0023] A few-shot deep learning model was adopted to solve the problem of scarce damage samples in water conservancy projects, ensuring the adaptability of the method in actual deployment. The output is divided into visual damage confidence and quantitative damage features to realize a hierarchical feedback mechanism. The visual damage confidence is responsible for qualitative triggering (whether calibration is needed), while the quantitative damage features are responsible for quantitative correction (how much calibration). This hierarchical approach ensures that the system does not perform unnecessary calculations at any time, and the correction magnitude is precisely matched with the severity of structural damage. Cross-modal collaborative fault early warning was realized, which greatly improved the accuracy and real-time performance of remaining lifetime prediction. The quantitative damage features successfully mapped the severity of structural damage to the importance of the dynamic opening feature sequence. This not only solved the problem of lag in traditional remaining lifetime prediction, but also enabled the prediction results to accurately reflect the actual accelerated decay state of the gate under the influence of structural damage, thus gaining valuable early warning and maintenance time windows for operation and maintenance personnel.

[0024] S3: Determine whether the confidence level of visual impairment is lower than the first preset threshold.

[0025] S4: If the judgment result is negative, the remaining life prediction value of the target gate is obtained directly through the dynamic opening feature sequence.

[0026] At this point, the monitoring method directly uses the dynamic opening feature sequence as input and substitutes it into the pre-trained remaining life prediction model to obtain the remaining life prediction value of the target gate at the current moment. This remaining life prediction value is calculated based on the gate's pure mechanical operating state data (such as angular velocity, acceleration, characteristic fluctuations, etc.), without being affected by any weight adjustment factors or structural damage information. This process reflects that when the structural health status is good or uncertain, the system reverts to the remaining life prediction based on traditional mechanical wear theory, preventing unnecessary error corrections that could lead to an incorrectly shortened remaining life prediction value. This step ensures that the monitoring method can maintain a stable and efficient benchmark monitoring capability under most normal operating conditions. The remaining life prediction value obtained in this step will then enter the subsequent collaborative risk assessment process.

[0027] It should be noted that the remaining useful life prediction model is a data-driven degradation prediction model. Its function is to map the dynamic opening feature sequence (a feature vector sequence reflecting the mechanical operating state of the gate) to the remaining usable life of the target gate. Since the dynamic opening feature sequence is time series data, the remaining useful life prediction model can specifically adopt any of the following deep learning architectures or regression algorithms suitable for processing time series data: Models based on recurrent neural networks: For example, deep learning architectures such as Long Short-Term Memory Networks (LSTM) or Gated Recurrent Units (GRU) can be used. These models are good at capturing long-term dependencies and non-linear degradation trends in time series data. They can effectively learn the degradation patterns of dynamic opening feature sequences accumulated over time and perform extrapolation predictions. The similarity-matching model: This model matches the current dynamic opening feature sequence with the dataset of historical gate degradation paths that have failed. The remaining lifetime prediction is calculated based on the remaining time of the historical degradation path that is closest to the current dynamic opening feature sequence. Regardless of the specific architecture used, this remaining lifetime prediction model has been trained on a large amount of historical operating data and has the ability to automatically learn the laws of mechanical wear and performance degradation from dynamic opening feature sequences. It is the core of the calculation to obtain the remaining lifetime prediction value.

[0028] S5: If the judgment result is yes, then according to the quantitative damage characteristics, assign weight adjustment factors to the features in the dynamic opening feature sequence that reflect the degree of deterioration of the mechanical operating state of the target gate, and re-acquire the dynamic opening feature sequence; based on the dynamic opening feature sequence after weight adjustment, correct the predicted value of the remaining life of the target gate.

[0029] In step S5, it should first be explained that the weight adjustment factor is assigned, including: The pre-defined quantitative damage features and the degree of deterioration of the mechanical operating state of the target gate are a monotonically increasing mapping relationship with accelerated change characteristics. This aims to simulate the nonlinear and accelerated induction effect of quantitative damage features on the mechanical performance of the target gate. When the quantitative damage features are small, their influence on the degree of deterioration of the mechanical operating state is slow; however, when the quantitative damage features accumulate to a certain extent, the rate at which they induce the degree of deterioration of the mechanical operating state will increase sharply. The multidimensional quantitative damage features are integrated into a dimensionless normalized damage score between [0, 1]. The specific formula is as follows: ; in, This represents the damage severity coefficient at the current moment, reflecting the relative contribution of quantified damage characteristics to the degree of deterioration in the mechanical operating condition. The normalized damage score at the current moment is obtained by normalizing the quantized damage features, where k is the acceleration exponent, a pure number greater than 1, ensuring that when... When increasing, The monotonically increasing curve pattern aligns with the logic that structural defects lead to an exponential increase in the deterioration of mechanical operating conditions. Based on the monotonically increasing mapping relationship of accelerated change characteristics, the quantified damage characteristics at the current moment are transformed into a single damage severity coefficient. This step is achieved through the above formula. accomplish; The injury severity coefficient is converted into a weighting adjustment factor. The weighting adjustment factor is greater than 1 and has a monotonically increasing relationship with the injury severity coefficient. The specific formula is as follows: ; in, , which is the weight adjustment factor at the current moment, and 1 is a dimensionless technique, representing the inherent weight of the feature itself in terms of the degree of deterioration of the mechanical operating state of the target gate; By weighting the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence with weight adjustment factors, we obtain the weighted dynamic opening feature sequence. Let the dynamic opening feature vector be... The i-th feature These are features reflecting the degree of deterioration in the mechanical operating condition of the target gate, adjusted for weights. The specific calculation formula is as follows: ; in, This is a feature that reflects the degree of deterioration in the mechanical operating status of the target gate after weight adjustment.

[0030] For example, a specific instance is used to demonstrate the workflow and final effect of the entire weight adjustment mechanism; Assumptions and dehumidification data: Normalized damage score Structural image analysis revealed severe damage; after normalization... ; Acceleration index k: k=3; Characteristics of the deterioration of the mechanical operating condition of the original target gate : Select the variance of the target gate's operating angular velocity, with dimensions of The current observation value is ; Step-by-step demonstration: First, convert it into a damage severity coefficient. ; Secondly, it is converted into a weight adjustment factor. =1 + 0.125 = 1.125; The weighted average yields the characteristics reflecting the deterioration of the target gate's mechanical operating state, adjusted for weights: the dimensionless weighting adjustment factor... =1.125 weighted to the original feature reflecting the degree of deterioration in the mechanical operating condition of the target gate. superior: ; At this point of severe structural damage, the original characteristic value reflecting the degree of deterioration in the mechanical operating condition of the target gate... It was magnified to This amplified feature value will then be used to correct the predicted remaining life of the target gate. Through this amplification, the monitoring method can identify the synergistic effect of mechanical deterioration and structural damage earlier, correcting the prediction results to a shorter remaining life, thereby achieving the purpose of early warning.

[0031] Figure 2 This is a schematic diagram of the structural damage evidence collaborative fault risk assessment process for a water conservancy project gate opening monitoring method based on the Internet of Things according to the present invention. S6: Based on the predicted remaining life of the target gate, obtain the dynamic trend anomaly generation results; Obtain the results of dynamic trend anomaly generation, including: Determine whether the corrected prediction of the remaining life of the gate is lower than the second preset threshold. The second preset threshold is a pre-set time length, the selection of which has clear engineering significance. It usually represents the minimum safety margin time required from when the system issues an early warning to when the operation and maintenance personnel can carry out emergency repairs or shutdowns. If the judgment result is yes, a dynamic trend anomaly is generated; indicating that the target gate is in a critical decline stage based on its mechanical operating state (and this state has been accelerated by the risk of structural damage), and is expected to reach the end point of safe operation in the short term. At this time, a dynamic trend anomaly is immediately generated. This dynamic trend anomaly is a clear alarm signal, which indicates that the performance of the target gate based on mechanical performance indicators has deteriorated rapidly and is trending towards failure or serious malfunction. The generation of this alarm signal constitutes the first level of early warning issued by this monitoring method, and provides key dynamic evidence for subsequent collaborative failure risk assessment. If the judgment result is negative, the dynamic trend anomaly generation result is "no dynamic trend anomaly". This indicates that based on the current mechanical operation status data, the predicted service life of the gate is still within the safety margin and has not yet reached the critical point that requires urgent attention from maintenance personnel. Therefore, the current prediction conclusion of the monitoring method is that the gate does not have the risk of accelerated degradation, or its degradation rate is still within an acceptable normal range.

[0032] At this point, the monitoring method will clearly generate a dynamic trend anomaly result as "no dynamic trend anomaly." This "no dynamic trend anomaly" result is a clear signal indicating that the current decline trend based on mechanical performance is healthy or risk-free. This "no" result will be used as input for subsequent collaborative failure risk assessment. In collaborative failure assessment, a collaborative failure risk will only be determined when both the dynamic trend anomaly and structural damage evidence are "yes." Therefore, the "no dynamic trend anomaly" result will ensure that even if structural damage evidence (i.e., structural defects) is subsequently generated, the final collaborative failure risk assessment result will still be "no," thus avoiding issuing the highest level alarm before the mechanical performance has truly reached the critical point, and ensuring the rigor and accuracy of risk assessment.

[0033] S7: Obtain the visual damage confidence and the duration of quantitative damage characteristics of any preset location within a preset monitoring period, and obtain the structural damage evidence generation result based on the duration. Obtain the structural damage evidence generation results, including: Determine whether, within a preset monitoring period, the confidence level of visual damage at any preset location is consistently lower than a first preset threshold, and the corresponding quantitative damage characteristics are consistently higher than a third preset threshold, in order to confirm the persistence and severity of structural defects over time. As can be seen from the above, this judgment contains two conditions that must be true simultaneously, and these conditions must remain constant over time. Visual impairment confidence condition: Within the preset monitoring period, the visual impairment confidence (reflecting the probability of impairment) of any preset part must be continuously lower than the first preset threshold. This threshold (first preset threshold) is used to exclude instantaneous false alarms caused by environmental interference (such as changes in light). Only when it is continuously lower than this threshold can it be preliminarily confirmed as a real visual impairment. Quantitative damage characteristic condition: During the same duration, the corresponding quantitative damage characteristic (a value reflecting the type and severity of damage) must be continuously higher than the third preset threshold. This threshold (the third preset threshold) is used to screen out damage with sufficient severity and exclude minor defects that do not affect the safe operation of the gate. Persistence determination: "Persistence" means that the above two conditions must be met continuously or discontinuously within a preset monitoring period for a preset duration determination condition. Only when both conditions meet the persistence requirement in time can it be identified as a stable and serious structural defect. If the judgment result is yes, structural damage evidence is generated. This structural damage evidence is a clear signal that the target gate has a stable and serious structural defect, providing reliable structural health evidence for subsequent collaborative fault determination. If the judgment result is negative, the structural damage evidence generation result will be no structural damage evidence. This indicates that although a damage signal may be captured at a certain point in time, the damage signal does not have the persistence or stability required for engineering to identify a risk source. For example, a momentary ambient light interference may cause a brief decrease in the confidence of visual damage (not meeting the first duration exceeding the fourth preset duration), or although it persists, its severity index never reaches the critical threshold that needs attention (not meeting the second duration exceeding the fifth preset duration). The monitoring method will explicitly generate a result of "no structural damage evidence" for the structural damage evidence generated. This "no structural damage evidence" result is a clear signal indicating that the current structural health status based on the visual-quantitative modality is normal or risk-free. This "no" result will serve as input for subsequent collaborative failure risk assessment. In collaborative failure assessment, both dynamic trend anomaly and structural damage evidence must be "yes" simultaneously for a collaborative failure risk to be determined. Therefore, the generation of "no structural damage evidence" will ensure that the final collaborative failure risk assessment result is negative, effectively preventing erroneous collaborative assessments caused solely by abnormal mechanical performance indicators (dynamic trend anomalies), and improving the accuracy of the highest-level early warning.

[0034] S8: Based on the dynamic trend anomaly generation results and structural damage evidence generation results, the determination result of the target gate having a collaborative failure risk is obtained.

[0035] The determination result of the risk of collaborative failure of the target gate is obtained, including: When both abnormal dynamic trends and evidence of structural damage are present, the target gate is deemed to have a risk of coordinated failure.

[0036] Anomalies in dynamic trends: derived from the mechanical-life prediction mode, indicating that the gate's performance degradation rate has accelerated to a critical state; Evidence of structural damage: derived from the visual-structural mode, indicating the presence of stable and severe structural root defects. The assessment of collaborative failure risk signifies that the monitoring method has confirmed that the current rapid deterioration in mechanical performance is not an isolated, accidental event, but rather an accelerated consequence of persistent and severe structural defects. This risk assessment offers higher accuracy and reliability than single-modal alerts, effectively avoiding false alarms and guiding maintenance personnel to take immediate emergency measures to address the root causes, thereby significantly reducing the probability of major safety incidents.

[0037] If abnormal dynamic trends and evidence of structural damage do not coexist, the target gate is deemed to have no risk of coordinated failure. "Not existing simultaneously" means that the "yes" condition is not met simultaneously among the two key independent pieces of evidence: abnormal dynamic trends and evidence of structural damage. Specifically, this includes the following three situations: Anomalies in dynamic trends but no evidence of structural damage: This indicates that the remaining life prediction is below the safety threshold (generating anomalies in dynamic trends), but the monitoring method fails to identify a persistent and serious structural defect as the root cause (generating no evidence of structural damage). In this case, the abnormality in mechanical performance may be caused by non-structural factors. The monitoring method will determine it as an abnormality in a single mechanical unit and recommend routine troubleshooting, but it is not classified as the most dangerous risk of synergistic failure. No dynamic trend anomaly but evidence of structural damage: This indicates that the monitoring method visually confirms a persistent and serious structural defect (generating evidence of structural damage), but based on the judgment of the remaining life prediction value, the mechanical performance of the gate has not yet reached the critical deterioration state (generating the result as no dynamic trend anomaly). In this case, the monitoring method regards the structural defect as a potential risk and recommends that the operation and maintenance personnel perform predictive maintenance, but not emergency shutdown. No abnormal dynamic trends and no evidence of structural damage: This indicates that the remaining life prediction has a sufficient safety margin (generated result: no abnormal dynamic trends), and the monitoring method has not detected any persistent and severe structural defects (generated result: no evidence of structural damage). This represents that the target gate is in a safe and healthy operating state, and the monitoring result is normal. Regardless of which of the above situations applies, since the logic of "dual evidence" is not met, it is ultimately determined that there is no risk of coordinated failure in the target gate. This logic ensures the rigor and accuracy of the highest level of early warning, avoids overreaction or misjudgment, and ensures that operation and maintenance resources can be accurately invested in the most dangerous coordinated failures that are actually accelerated by structural defects.

[0038] In this application embodiment, "dynamic trend anomaly", "structural damage evidence" and "cooperative failure risk determination" are regarded as a cross-modal collaborative early warning and risk confirmation technical solution. The core purpose of this invention is to overcome the two major problems of high false alarm rate and inaccurate fault characterization in traditional gate monitoring systems. In traditional independent monitoring methods, if an alarm is triggered only based on the modified gate remaining life prediction value being lower than the safety threshold, the anomaly may be caused by instantaneous, non-structural interference, which is prone to false alarm. Conversely, if an alarm is triggered only based on the visual damage confidence level being continuously lower than the threshold, it cannot provide a time basis for when the mechanical performance will deteriorate sharply. This technical solution ensures the effectiveness of early warning and the accuracy of decision-making. The advantages of this operation are: it achieves high-reliability early warning; by using logical AND for judgment, this method avoids false alarms caused by environmental interference or momentary failure of a single sensor, and can accurately characterize the root cause of the fault, clearly indicating that the current mechanical performance deterioration is accelerated by structural defects, providing clear guidance for subsequent maintenance strategies. This mechanism ensures the timeliness and decision-making value of the early warning, because it only triggers the highest level alarm when the life prediction value is close to the threshold (indicating that maintenance time is urgent) and the structural defect has been confirmed (indicating that the problem is serious). In terms of alternative technologies, existing technologies mainly include: independent threshold alarm, that is, an alarm is triggered when either mechanical or structural indicators exceed the limit, which is limited by the inability to distinguish the fault type and the high false alarm rate; and simple threshold persistence judgment, which only judges whether the structural damage continues to exist, but does not combine it with the remaining life prediction for cross-validation, and cannot establish the degree of impact of structural defects on mechanical performance. The collaborative fault risk judgment mechanism of this invention effectively avoids the limitations of all the above-mentioned existing technologies, and realizes the deep integration and verification of cross-modal data at the risk decision-making level.

[0039] Figure 3 This is a schematic diagram of the structural damage evidence generation process of a method for monitoring the opening of water conservancy engineering gates based on the Internet of Things according to the present invention. Further, determining whether structural damage exists includes: The first duration during which the confidence level of visual damage at any preset part of the target gate is lower than a first preset threshold is obtained within a preset monitoring period; The second duration of the quantitative damage characteristics that are higher than the third preset threshold within the preset monitoring period is obtained. This step focuses on the numerical indicator of the quantitative damage characteristics (such as crack length and corrosion area percentage), and counts the actual duration of the characteristics that are continuously higher than the third preset threshold, which is the second duration. This ensures that the focus is not only on the question of "whether there is damage", but also on the question of "whether the damage is serious enough". When the first duration exceeds the fourth preset duration and the second duration exceeds the fifth preset duration, structural damage evidence is generated.

[0040] It should be noted that the first preset threshold is the lower limit of the confidence level of visual impairment. Below this value, it indicates that the monitoring system believes that the probability of impairment is high. The threshold is set based on the performance curve and fault tolerance requirements of the few-sample deep learning model. It is usually determined by testing the false alarm rate under various lighting and environmental interferences in experiments. Its goal is to set it at a level that can exclude normal images with high confidence, so as to ensure that the impairment initially confirmed by the monitoring system has a high degree of authenticity. The third preset threshold is the lower limit of the severity of the quantitative damage characteristics. If it is higher than this value, it indicates that the structural damage has reached the engineering level that must be paid attention to. The setting of this threshold is strictly based on the design specifications and maintenance manuals of hydraulic engineering gates. For example, if the manual stipulates that steel structure cracks with a length of more than 5 mm require intervention, this threshold is to convert 5 mm into the corresponding normalized quantitative damage characteristic value. It excludes minor defects that have no engineering significance and ensures that the monitoring system focuses its resources on serious damage with potential risks. The fourth preset duration is the minimum effective duration of the first duration. This duration is used to filter out brief anomalies of a few seconds or minutes caused by short network interruptions, image acquisition flicker, or instantaneous changes in ambient light. Its setting is based on the monitoring system's requirement to eliminate instantaneous interference. For example, it can be set to 30 minutes or 1 hour to ensure that damaged images must be continuously and stably captured for more than this period of time to confirm the reliability of the confidence level. The fifth preset duration is the minimum effective duration of the second duration. This duration is used to confirm that the change in severity is not an instantaneous fluctuation, but a stable deterioration trend. It is set based on the minimum time window of the gate operation and the structural response time. It is usually set to a longer period, such as 6 hours or 24 hours, to ensure that the damage severity index remains stable above the danger threshold in multiple gate opening and closing cycles, thereby establishing the persistence of severity.

[0041] Furthermore, the first duration during which the confidence level of visual impairment at any preset part of the target gate is lower than a first preset threshold within a preset monitoring period includes: The system counts the number of consecutive frames with a visual impairment confidence level below a first preset threshold within a preset monitoring period. During the monitoring period, the visual impairment confidence level sequence is compared frame by frame with the first preset threshold. Only when the confidence level of a certain frame is lower than the threshold is it counted in a consecutive counter. The system records the number of all consecutive image frames with a confidence level below the threshold. This number is the consecutive frame count. This consecutive frame count excludes frames with a confidence level higher than the threshold, ensuring that the statistics represent the smallest unit of time in which the structural impairment state persists. The first duration is determined based on the number of consecutive frames and the acquisition frame rate of the structural image sequence, using the following formula: ; in, The first duration (in seconds) represents the actual total duration of continuous structural damage. This represents the number of consecutive frames, specifically the total number of image frames whose visual impairment confidence level is continuously below a first preset threshold. The frame rate for acquiring the structural image sequence (unit: frames / second); When the first duration meets the continuous duration determination condition, it is determined that the first duration exceeds the fourth preset duration. The first duration (i.e. the calculated actual continuous duration) is used to compare with the continuous duration determination condition, which adopts a more robust determination logic based on the number of monitoring sub-windows. Only when this continuous duration recurs in multiple time windows and reaches a preset number (sixth preset number) is the first duration finally established to have engineering significance of continuity, thus determining that it has exceeded the fourth preset duration, providing a solid evidence in the time dimension for generating reliable structural damage evidence.

[0042] Furthermore, when the first duration meets the continuous duration judgment condition, this judgment condition is used to eliminate instantaneous interference and occasional errors in structural damage monitoring, ensuring that the judged structural defects are stable and persistent. This condition is achieved by introducing time window division and quantitative statistics to quantify the persistence, including: The preset monitoring period is divided into time windows to obtain multiple continuous monitoring sub-windows. The preset monitoring period is a macroscopic time range (e.g., 30 days), while the monitoring sub-window is a fixed, shorter time segment within the period (e.g., 24 hours or 1 week). Time window division discretizes continuous monitoring data into a series of ordered segments. For example, 30 days can be divided into 30 24-hour monitoring sub-windows. The system counts the number of monitoring sub-windows where the confidence level of visual impairment is consistently below a first preset threshold. Within each sub-window, the system checks whether there is a continuous or cumulative monitoring period in which the confidence level of visual impairment is consistently below the first preset threshold. If a sub-window meets this threshold condition (for example, 80% of the image frames show a confidence level below the threshold within 24 hours), the sub-window is marked as an "abnormal sub-window" and included in the total count. When the number of monitoring sub-windows reaches the sixth preset number, the duration is determined to meet the continuous duration judgment condition. The sixth preset number is a key threshold set in advance. Its physical meaning is: only when the abnormal state of structural damage spans multiple discontinuous monitoring sub-windows (for example, within 30 days, five different 24-hour windows confirm the continued existence of damage) is it finally determined to meet the continuous duration judgment condition. This judgment method based on the number of sub-windows is more robust than the simple "how many seconds" judgment method. It allows for normal or interfering data between two damage confirmations, which is more in line with the fluctuating and continuous characteristics of structural damage under environmental influence in actual engineering. This judgment ultimately confirms the stability and reliability of the structural damage evidence.

[0043] Example 2: As one embodiment of the present invention, which differs from the previous embodiment, a water conservancy project gate opening monitoring system based on the Internet of Things includes: Data acquisition module: used to acquire the continuous dynamic opening feature sequence of the target gate within a preset monitoring period, as well as the structural image sequence of a preset part of the target gate that is time-synchronized with the dynamic opening feature sequence; Data processing module: Used to process structural image sequences based on a preset few-shot deep learning model to obtain visual damage confidence sequences and quantified damage features of preset parts of the gate; Judgment module: used to determine whether the confidence level of visual impairment is lower than the first preset threshold; Self-calibration module: If the judgment result is yes, then according to the quantified damage characteristics, the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence are assigned weight adjustment factors, and the dynamic opening feature sequence is re-acquired; based on the dynamic opening feature sequence after weight adjustment, the remaining life prediction value of the target gate is corrected. If the judgment result is negative, the remaining life prediction value of the target gate is obtained directly through the dynamic opening feature sequence. Collaborative Judgment Module: Used to obtain dynamic trend anomaly generation results based on the predicted remaining life of the target gate; The system obtains the visual damage confidence level and the duration of quantitative damage characteristics for any preset location within a preset monitoring period, and generates structural damage evidence based on the duration. Based on the results of dynamic trend anomaly generation and structural damage evidence generation, the determination result of the target gate having a risk of collaborative failure is obtained.

[0044] Example 3: In one embodiment of the present invention, which differs from the previous embodiment, the electronic device includes one or more processors and a memory.

[0045] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0046] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0047] In one example, the electronic device may also include input devices and output devices, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown). In addition, depending on the specific application, the electronic device may include any other suitable components.

[0048] Example 4: Embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps described in the "Exemplary Methods" section above according to the various embodiments of this application.

[0049] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0050] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not restrict the application from being implemented using the specific details described above.

[0051] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0052] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0053] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0054] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for monitoring the opening degree of gates in water conservancy projects based on the Internet of Things, characterized in that, include: Acquire continuous opening angle data of the target gate and structural image sequence of a preset part of the target gate within a preset monitoring period; construct a dynamic opening feature sequence of the target gate based on the opening angle data; the structural image sequence is time-synchronized with the dynamic opening feature sequence. Based on a preset few-shot deep learning model, the sequence of structural images is processed to obtain a visual damage confidence sequence and quantified damage features of a preset part of the gate. Determine whether the confidence level of the visual impairment is lower than a first preset threshold; If the judgment result is yes, then the weight adjustment factor is assigned to the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence according to the quantitative damage characteristics, and the dynamic opening feature sequence is re-acquired. Based on the weighted dynamic opening feature sequence, the remaining life prediction of the target gate is corrected. If the judgment result is negative, the remaining life prediction value of the target gate is directly obtained through the dynamic opening feature sequence. Based on the predicted remaining life of the target gate, the dynamic trend anomaly generation results are obtained; The system obtains the visual damage confidence level and the duration of quantitative damage characteristics of any preset location within a preset monitoring period, and obtains the structural damage evidence generation result based on the duration. Based on the dynamic trend anomaly generation results and structural damage evidence generation results, the determination result of the target gate having a risk of collaborative failure is obtained.

2. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things according to claim 1, characterized in that: The process of obtaining the dynamic trend anomaly generation result includes: Determine whether the corrected predicted remaining lifespan of the gate is lower than a second preset threshold. If the judgment result is yes, then the dynamic trend anomaly generation result is dynamic trend anomaly; If the judgment result is negative, the dynamic trend anomaly generation result will be "no dynamic trend anomaly". The obtained structural damage evidence generation results include: Determine whether, within the preset monitoring period, the confidence level of visual damage at any preset location is consistently lower than the first preset threshold, and the corresponding quantitative damage characteristics are consistently higher than the third preset threshold. If the judgment result is yes, then the structural damage evidence generation result is structural damage evidence; If the judgment result is negative, the structural damage evidence generation result will be no structural damage evidence. The determination result that the target gate has a risk of collaborative failure includes: When the abnormal dynamic trend and the evidence of structural damage coexist, it is determined that the target gate has a risk of coordinated failure. If the abnormal dynamic trend and the evidence of structural damage do not coexist, then the target gate is determined to have no risk of coordinated failure.

3. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things according to claim 1, characterized in that: The acquisition of the visual impairment confidence sequence and quantified impairment features includes: The few-shot deep learning model is trained in advance using images containing damage to the target gate; Based on a pre-set few-shot deep learning model, reasoning is performed on each frame of the acquired structural image sequence, and the probability value reflecting the damage to a pre-set part of the target gate is output as the visual damage confidence. The few-shot deep learning model extracts and outputs features reflecting the type and severity of damage in the structural image sequence as the quantified damage features.

4. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things according to claim 1, characterized in that: The weighting adjustment factor includes: A monotonically increasing mapping relationship with accelerating change characteristics is established between the quantified damage characteristics and the degree of deterioration of the mechanical operating state of the target gate; Based on the monotonically increasing mapping relationship of the accelerated change characteristics, the quantified damage characteristics at the current moment are transformed into a single damage severity coefficient. The severity coefficient of injury is converted into a weighting adjustment factor, wherein the weighting adjustment factor is greater than 1 and has a monotonically increasing relationship with the severity coefficient of injury. The dynamic opening feature sequence is obtained by weighting the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence with the weight adjustment factor.

5. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things according to claim 2, characterized in that: The obtained structural damage evidence generation results also include: The first duration during which the confidence level of visual damage at any preset part of the target gate is lower than the first preset threshold is obtained within a preset monitoring period; The second duration during which the quantitative damage characteristics corresponding to the preset monitoring period are higher than the third preset threshold is obtained; When the first duration exceeds the fourth preset duration and the second duration exceeds the fifth preset duration, the structural damage evidence is generated.

6. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things according to claim 5, characterized in that: The first duration during which the confidence level of visual impairment at any preset part of the target gate is lower than the first preset threshold within a preset monitoring period includes: The number of consecutive frames in which the confidence level of visual impairment is lower than the first preset threshold within the preset monitoring period is counted. The first duration is determined based on the number of consecutive frames and the acquisition frame rate of the structured image sequence; When the first duration meets the continuous duration determination condition, it is determined that the first duration exceeds the fourth preset duration.

7. The method for monitoring the gate opening degree of a water conservancy project based on the Internet of Things as described in claim 6, characterized in that: The condition that the first duration satisfies the continuous duration determination condition includes: The preset monitoring period is divided into time windows to obtain multiple consecutive monitoring sub-windows; The number of monitoring sub-windows whose visual impairment confidence level is consistently below the first preset threshold is counted. When the number of monitoring sub-windows reaches the sixth preset number, it is determined that the duration meets the continuous duration determination condition.

8. A gate opening monitoring system for water conservancy projects based on the Internet of Things, characterized in that, include: Data acquisition module: used to acquire the continuous dynamic opening feature sequence of the target gate within a preset monitoring period, as well as the structural image sequence of a preset part of the target gate that is time-synchronized with the dynamic opening feature sequence; Data processing module: used to process the image sequence of the structure based on a preset few-shot deep learning model to obtain the visual damage confidence sequence and quantified damage features of preset parts of the gate; Judgment module: used to determine whether the confidence level of the visual impairment is lower than a first preset threshold; Self-calibration module: If the judgment result is yes, then according to the quantitative damage characteristics, the features reflecting the deterioration of the mechanical operating state of the target gate in the dynamic opening feature sequence are assigned weight adjustment factors, and the dynamic opening feature sequence is re-acquired; Based on the weighted dynamic opening feature sequence, the remaining life prediction of the target gate is corrected. If the judgment result is negative, the remaining life prediction value of the target gate is directly obtained through the dynamic opening feature sequence. Collaborative Judgment Module: Used to obtain dynamic trend anomaly generation results based on the predicted remaining life of the target gate; The system obtains the visual damage confidence level and the duration of quantitative damage characteristics of any preset location within a preset monitoring period, and obtains the structural damage evidence generation result based on the duration. Based on the dynamic trend anomaly generation results and structural damage evidence generation results, the determination result of the target gate having a risk of collaborative failure is obtained.

9. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method as described in any one of claims 1 to 7.

10. A computer storage medium storing computer-executable instructions thereon, characterized in that: When the computer-executable instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 7.

Images