Intelligent data monitoring system and method based on ultrasonic water meter

Abstract

The invention discloses an intelligent data monitoring system and method based on an ultrasonic water meter, and the method comprises the steps: determining a data grading label through the comparison between the vibration features of the vibration data of an ultrasonic sensor in a preset time period and a recent historical feature baseline; therefore, the system can intelligently identify the abnormal degree of the current running state of the water meter. Furthermore, based on the data grading label, a data transmission strategy and a data compression strategy for the original vibration signal are specified, so that the data transmission and storage efficiency is optimized to the greatest extent on the premise of ensuring that key information is not lost. The self-adaptive strategy not only solves the problem of low data transmission and storage efficiency in a traditional scheme, but also ensures that enough detailed data can be obtained at a critical moment, and resources can be efficiently utilized during conventional operation, so that the optimal balance between data integrity and accuracy and transmission and storage cost is realized.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring, and more specifically, to a data intelligent monitoring system and method based on ultrasonic water meters. Background Art

[0002] With the in-depth development of Internet of Things technology and smart city construction, higher requirements are put forward for the management and monitoring of water resources. Due to problems such as wear and reduced accuracy of traditional mechanical water meters, they can no longer meet the needs of modern refined management. Ultrasonic water meters are becoming an important development direction in the field of intelligent water utilities due to their advantages of no mechanical moving parts, high accuracy, wide range ratio, good long-term stability, and low power consumption. By effectively monitoring the operation data of ultrasonic water meters, not only can the water usage situation be grasped in real time, and abnormal situations such as leakage and water theft be detected in a timely manner, but also key data support can be provided for the operation and maintenance of water utilities companies, pipe network optimization, and user services, thereby improving the efficiency of water resource utilization and management level. Therefore, it is particularly necessary to construct an efficient and intelligent ultrasonic water meter data monitoring solution.

[0003] However, the existing ultrasonic water meter data monitoring solutions still face many challenges in practical applications. Traditional monitoring methods often adopt fixed-frequency or full-volume data transmission modes, that is, regardless of the operation state of the water meter, the collected original vibration signals or the data after simple processing are uploaded to the cloud platform. This "one-size-fits-all" data transmission strategy results in a huge amount of data, which not only occupies a large amount of network bandwidth resources, significantly increases the data transmission cost, but also poses extremely high requirements on the storage and processing capabilities of the cloud platform. Especially in the scenario of large-scale deployment, the transmission and storage costs of massive data will become a bottleneck restricting the popularization and operation efficiency of the intelligent water utility system. In addition, the fixed-mode data transmission may also lead to the failure to transmit key original high-fidelity data in a timely and complete manner when abnormal situations occur, affecting the efficiency of fault diagnosis and problem solving.

[0004] Therefore, an optimized data intelligent monitoring solution for ultrasonic water meters is expected. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide an intelligent data monitoring system and method based on an ultrasonic water meter, which determine data classification tags by comparing the vibration characteristics of the vibration data of the ultrasonic sensor within a predetermined time period with the recent historical feature baseline, enabling the system to intelligently identify the abnormal degree of the current water meter operation state. Further, based on the data classification tags, data transmission strategies and data compression strategies for the original vibration signal are specified, so as to optimize the data transmission and storage efficiency to the greatest extent on the premise of ensuring that key information is not lost. This adaptive strategy not only solves the problem of low data transmission and storage efficiency in traditional solutions, but more importantly, it ensures that sufficient detailed data can be obtained at critical moments, while efficiently utilizing resources during normal operation, achieving the best balance among data integrity, accuracy, and transmission and storage costs.

[0006] According to one aspect of the present application, an intelligent data monitoring method based on an ultrasonic water meter is provided, which includes: The edge computing center receives the original vibration signal of a preset time period collected by the vibration sensor of the target ultrasonic water meter; The edge computing center extracts the vibration characteristics of the original vibration signal to obtain the vibration signal waveform coding characteristics; Based on the comparison result between the vibration signal waveform coding characteristics and the recent historical feature baseline, the edge computing center specifies data transmission strategies and data compression strategies for the original vibration signal; The edge computing center processes the original vibration signal based on the data transmission strategies and data compression strategies to obtain an adaptive compression data packet and transmits it to the cloud platform.

[0007] According to another aspect of the present application, an intelligent data monitoring system based on an ultrasonic water meter is provided, which includes: An original vibration signal acquisition module for the edge computing center to receive the original vibration signal of a preset time period collected by the vibration sensor of the target ultrasonic water meter; A vibration characteristic extraction module for the edge computing center to extract the vibration characteristics of the original vibration signal to obtain the vibration signal waveform coding characteristics; A data processing strategy generation module for the edge computing center to specify data transmission strategies and data compression strategies for the original vibration signal based on the comparison result between the vibration signal waveform coding characteristics and the recent historical feature baseline; A data processing module for the edge computing center to process the original vibration signal based on the data transmission strategies and data compression strategies to obtain an adaptive compression data packet and transmit it to the cloud platform.

[0008] Compared with the prior art, a data intelligent monitoring system and method based on an ultrasonic water meter provided by the present application determine data classification labels by comparing the vibration characteristics of the vibration data of an ultrasonic sensor within a predetermined time period with a recent historical feature baseline, which enables the system to intelligently identify the abnormal degree of the current water meter operation state. Further, based on the data classification labels, data transmission strategies and data compression strategies for the original vibration signal are specified, so as to optimize the data transmission and storage efficiency to the greatest extent on the premise of ensuring that key information is not lost. This adaptive strategy not only solves the problem of low data transmission and storage efficiency in traditional solutions, but more importantly, it ensures that sufficient detailed data can be obtained at critical moments, while resources can be efficiently utilized during normal operation, achieving the best balance among data integrity, accuracy, and transmission and storage costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 is a flowchart of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of the present application; Figure 2 is a schematic diagram of data flow of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of the present application; Figure 3 is a flowchart of sub-step S3 of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of the present application; Figure 4 is a flowchart of sub-step S31 of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of the present application; Figure 5 is a block diagram of a data intelligent monitoring system based on an ultrasonic water meter according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] Hereinafter, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0012] As shown in this application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on the user terminal and / or server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the previous or following operations are not necessarily executed precisely in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more steps can be removed from these processes.

[0015] Next, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described here.

[0016] In the technical solution of this application, a data intelligent monitoring method based on an ultrasonic water meter is proposed. Figure 1 FIG. is a flowchart of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of this application. Figure 2 FIG. is a schematic diagram of data flow of a data intelligent monitoring method based on an ultrasonic water meter according to an embodiment of this application. As Figure 1 and Figure 2 shown, the data intelligent monitoring method based on an ultrasonic water meter according to the embodiment of this application includes the steps: S1, the edge computing center receives the original vibration signals of a preset time period collected by the vibration sensor of the target ultrasonic water meter; S2, the edge computing center extracts the vibration characteristics of the original vibration signals to obtain the vibration signal waveform coding characteristics; S3, the edge computing center designates a data transmission strategy and a data compression strategy for the original vibration signals based on the comparison result between the vibration signal waveform coding characteristics and the recent historical feature baseline; S4, the edge computing center processes the original vibration signals based on the data transmission strategy and the data compression strategy to obtain an adaptive compression data packet and transmits it to the cloud platform.

[0017] Specifically, in S1, the edge computing center receives the original vibration signals of a preset time period collected by the vibration sensor of the target ultrasonic water meter. It should be understood that in the actual application scenario of the intelligent water service system, ultrasonic water meters are often distributed at various nodes of the urban water supply network. These water meters continuously collect the original vibration signals that reflect the water usage status and the operation status of the equipment through the built-in vibration sensors. With the continuous growth of the data volume, how to efficiently and intelligently process and manage these massive original signals in real time has become the key to improving the water supply safety and operation and maintenance efficiency. Therefore, in the technical solution of this application, by introducing an edge computing center, that is, deploying a data processing unit with a certain computing power on a local computing node close to the data source end (such as a community, a building, or a regional level), the system can directly receive the original vibration signals within a preset time period collected by the vibration sensor of the target ultrasonic water meter, without uploading all the original data to the cloud, thereby effectively alleviating the network bandwidth pressure and improving the response speed.

[0018] In the specific implementation process, the vibration sensor built in or externally connected to the ultrasonic water meter will continuously sample the minute vibrations generated by the states such as the fluid flow in the pipeline and the operation of mechanical components according to the system-set cycle. These sampled data are first cached in the local storage module in the form of digital signals and are uniformly sent to the edge computing center within the jurisdiction area via wired or wireless communication after the end of the set time window. Here, by moving the preliminary data aggregation and analysis tasks to the edge side, not only is the dependence on the core network and cloud platform resources significantly reduced, but also a more agile and efficient response to abnormal events is achieved.

[0019] Specifically, in S2, the edge computing center extracts the vibration features of the original vibration signals to obtain the vibration signal waveform coding features. It should be understood that in the face of these high-frequency, massive, and complex original data, it is difficult to accurately capture potential abnormal patterns only by simple statistics or fixed threshold methods, nor can it meet the requirements of flexible hierarchical management and efficient processing of multi-level anomalies. Therefore, in the technical solution of this application, vibration feature extraction based on one-dimensional convolutional coding is performed on the original vibration signals to obtain a vibration signal waveform feature coding vector as the vibration signal waveform coding features. In the specific implementation process, the edge computing center first receives a segment of original time-series vibration signals collected by the vibration sensor of the ultrasonic water meter, and then inputs this time-series data into a pre-trained one-dimensional convolutional network model. This model can automatically identify and abstract the features closely related to the equipment operation status at different time scales, such as the minute waveform changes caused by fluid disturbance, mechanical wear, or external interference. In this way, the system can transform the complex and redundant original vibration data into highly condensed and discriminative coding vectors, laying a solid foundation for subsequent comparison with the historical baseline, abnormal level determination, and formulation of adaptive compression strategies.

[0020] Specifically, in S3, the edge computing center designates a data transmission strategy and a data compression strategy for the original vibration signal based on the comparison result between the vibration signal waveform coding feature and the recent historical feature baseline. In a specific example of this application, as Figure 3 shown, S3 includes: S31, the edge computing center compares the vibration signal waveform coding feature with the recent historical feature baseline to obtain the deviation degree; S32, based on the deviation degree, determines a data classification label, where the data classification label includes highly suspected anomaly, mildly suspected anomaly, and normal; S33, based on the data classification label, designates a data transmission strategy and a data compression strategy for the original vibration signal.

[0021] Specifically, in S31, the edge computing center compares the vibration signal waveform coding feature with the recent historical feature baseline to obtain the deviation degree. In a specific example of this application, as Figure 4 shown, S31 includes: S311, obtaining the baseline feature vector of the recent historical vibration signal as the recent historical feature baseline; S312, inputting the vibration signal waveform feature coding vector and the baseline feature vector into a fine-grained waveform feature difference modeling network to obtain a vibration waveform deviation baseline response coding vector as the vibration waveform deviation baseline response coding feature; S313, based on the vibration waveform deviation baseline response coding feature, obtaining the deviation degree.

[0022] More specifically, in S311, the baseline feature vector of the recent historical vibration signal is obtained as the recent historical feature baseline. Since the vibration signal itself is affected by various factors such as the environment, equipment aging, and water flow changes, it is often difficult to accurately judge whether there is an anomaly based on data at a single moment or an isolated period. The recent historical feature baseline represents the original vibration signals collected by the target ultrasonic water meter within a set time window (such as several hours, several days, or several weeks) for feature extraction and statistical analysis, forming a reference vector that reflects the vibration characteristics of the device under normal operation in the recent period. This baseline can not only reflect the health level and typical working conditions of the device itself but also effectively avoid misjudgments caused by occasional noise or short-term disturbances.

[0023] In the specific implementation process, the edge computing center will regularly select a preset time period (such as the most recent 24 hours, the most recent week, etc.) from the stored historical vibration data, and use the same one-dimensional convolutional network model as the real-time data to perform batch feature extraction on these historical original vibration signals. Each time segment will generate a corresponding vibration signal waveform coding vector. Subsequently, the system uses methods such as weighted average, median, and principal component analysis to fuse this series of coding vectors into a baseline feature vector with strong representativeness and high anti-interference ability. It is worth mentioning that this baseline, as an important reference standard for all newly collected data during the current monitoring period, will be stored in the edge computing center and dynamically updated with the changes in the equipment operating conditions and the environment. By obtaining and maintaining such a dynamically updated recent historical feature baseline, the system can accurately compare the newly collected vibration signals, effectively distinguishing which changes fall within the normal fluctuation range. This not only greatly improves the accuracy of anomaly detection and hierarchical management but also provides a scientific basis for formulating subsequent adaptive compression strategies, enabling key events to be discovered and processed in a timely manner, while conventional data can be efficiently compressed and transmitted, overall optimizing the data transfer efficiency and operation and maintenance response ability of the intelligent water system.

[0024] More specifically, in S312, the vibration signal waveform feature encoding vector and the baseline feature vector are input into the fine-grained waveform feature difference modeling network to obtain the vibration waveform offset baseline response encoding vector as the vibration waveform offset baseline response encoding feature. It should be understood that in the actual monitoring scenario of the ultrasonic water meter, due to the high-dimensional, strong temporal and complex fluctuation characteristics of the original vibration signal, it is difficult for traditional methods to accurately capture the deep differences between the device state and the baseline features from the massive data. Therefore, without breaking through the limitations of fixed thresholds or shallow statistical features to deeply analyze the vibration signal waveform feature encoding vector and the baseline feature vector, in the technical solution of this application, the vibration signal waveform feature encoding vector and the baseline feature vector are input into the fine-grained waveform feature difference modeling network to obtain the vibration waveform offset baseline response encoding vector as the vibration waveform offset baseline response encoding feature. That is, by constructing a local interaction response mechanism based on the ordered feature intensity, the feature patterns reflecting the small offsets of the device operating state are extracted from the vibration signal, thereby providing a decision-making basis for anomaly grading and resource optimization. In this process, first, the vibration signal waveform feature encoding vector and the baseline feature vector are sorted based on the eigenvalue size to eliminate the random interference of the original feature arrangement, so as to focus on the feature intensity distribution itself; subsequently, the ordered vector is decomposed into multiple local feature segments through equal-grained feature segmentation to form local feature sequences describing different intensity intervals, enabling subsequent analysis to conduct fine-grained comparisons on the vibration patterns within a specific feature intensity range; then, each corresponding pair of vibration signal local feature segments and baseline local feature segments are input into the transfer response inference unit to capture the complex dynamic associations between the vibration signal and the baseline features and improve the discrimination ability for different abnormal patterns; finally, the interaction response information of each local region is integrated through the sequence transfer module to form an encoding vector that can comprehensively reflect the overall offset trend of the vibration signal, laying a foundation for the quantitative evaluation of the deviation degree. This fine-grained feature difference modeling ability provides a reliable basis for the subsequent dynamic adjustment of the data compression strategy, avoiding both the bandwidth waste caused by high-frequency transmission of normal data and ensuring that the key features of abnormal events are completely retained, thus achieving a balance between monitoring accuracy and resource efficiency in the intelligent water service scenario.

[0025] Specifically, first, the vibration signal waveform feature coding vector and the baseline feature vector are sorted based on the eigenvalue magnitude to obtain the vibration signal waveform feature sorted coding vector and the baseline feature sorted coding vector. It should be understood that the original arrangement order of the vibration signal waveform feature coding vector and the baseline feature vector is often affected by factors such as the sensor acquisition timing and the signal preprocessing process, resulting in random distribution positions of the feature elements. This randomness will interfere with the accurate judgment of the equipment state differences. For example, features of the same intensity level may be misjudged as abnormal fluctuations due to different arrangement orders. Therefore, in the technical solution of this application, the vibration signal waveform feature coding vector and the baseline feature vector are sorted based on the eigenvalue magnitude to obtain the vibration signal waveform feature sorted coding vector and the baseline feature sorted coding vector.

[0026] In the specific implementation process, the system first sorts each feature element in the vibration signal waveform feature coding vector and the baseline feature vector, rearranges the element positions in ascending or descending order, and generates the vibration signal waveform feature sorted coding vector and the baseline feature sorted coding vector. For example, if there are high-value feature elements representing high-frequency vibration energy in the original vibration signal waveform feature coding vector that are scattered at different index positions, these high-value elements will be concentrated and arranged in a specific interval of the vector after sorting. This structured representation mechanism based on the intensity distribution lays a key foundation for achieving stable and reliable anomaly grading and resource dynamic allocation in the intelligent water service scenario.

[0027] In a specific example of this application, the following formula is used to sort the vibration signal waveform feature coding vector and the baseline feature vector based on the eigenvalue magnitude to obtain the vibration signal waveform feature sorted coding vector and the baseline feature sorted coding vector; where the formula is:

[0028] Among them, is the vibration signal waveform feature coding vector, represents sorting the vector elements, is the baseline feature vector, is the vibration signal waveform feature sorted coding vector, is the baseline feature sorted coding vector.

[0029] Next, the vibration signal waveform feature ordered arrangement coding vector and the baseline feature ordered arrangement coding vector are subjected to equal-granularity feature segmentation to obtain a sequence of vibration signal waveform local feature ordered coding vectors and a sequence of baseline local feature ordered coding vectors. It should be understood that although the vibration signal waveform feature ordered arrangement coding vector and the baseline feature ordered arrangement coding vector have eliminated the randomness of the original arrangement through eigenvalue sorting, their overall dimensions are still relatively high and cover a wide intensity range, making it difficult to directly capture the subtle offsets of the equipment state within a specific intensity range. Therefore, in order to establish a hierarchical fine-grained comparison framework to focus on the vibration mode differences in different intensity ranges, in the technical solution of this application, the vibration signal waveform feature ordered arrangement coding vector and the baseline feature ordered arrangement coding vector are subjected to equal-granularity feature segmentation to obtain a sequence of vibration signal waveform local feature ordered coding vectors and a sequence of baseline local feature ordered coding vectors.

[0030] In the specific implementation process, the system synchronously cuts the vibration signal waveform feature ordered arrangement coding vector and the baseline feature ordered arrangement coding vector along the feature dimension into multiple consecutive sub-vector segments of equal length, forming a sequence of vibration signal waveform local feature ordered coding vectors and a sequence of baseline local feature ordered coding vectors. For example, if the total dimension of the ordered vector is 100 and the segmentation granularity is 10, each coding vector will be evenly divided into 10 sub-vectors, and each sub-vector corresponds to an eigenvalue intensity range (such as the highest 10% intensity segment, the second highest 10% intensity segment, etc.). It is worth mentioning that the finer the segmentation granularity, the stronger the ability to capture local anomalies, but the computational complexity increases accordingly; on the contrary, the coarser granularity reduces resource consumption but may miss weak anomaly signals in key intervals. By dynamically adjusting the segmentation granularity, the system can adapt to the monitoring requirements under different working conditions. Through equal-granularity segmentation, the global feature comparison is transformed into multi-level local intensity range comparisons, enabling the system to accurately locate the feature intensity range where anomalies occur (such as high-frequency vibration anomalies being concentrated in high-value sections), enhancing the sensitivity to local offsets.

[0031] In a specific example of this application, the following formula is used to perform equal-granularity feature segmentation on the vibration signal waveform feature ordered arrangement coding vector and the baseline feature ordered arrangement coding vector to obtain a sequence of vibration signal waveform local feature ordered coding vectors and a sequence of baseline local feature ordered coding vectors; where the formula is:

[0032] Among them, represents the feature segmentation function, are respectively the 1st, 2nd, th, and An ordered coding vector of local features of a vibration signal waveform, which are respectively the 1st, 2nd, th, and th baseline local feature ordered coding vectors in the sequence of the baseline local feature ordered coding vectors.

[0033] Subsequently, each pair of corresponding vibration signal waveform local feature ordered coding vectors and baseline local feature ordered coding vectors in the sequence of vibration signal waveform local feature ordered coding vectors and the sequence of baseline local feature ordered coding vectors are input into the transfer response inference unit to obtain a sequence of initial vibration waveform offset baseline local transfer response coding matrices. It should be understood that although the sequences of vibration signal waveform local feature ordered coding vectors and baseline local feature ordered coding vectors have achieved alignment of intensity intervals through ordered arrangement and equal-granularity segmentation, the dynamic association patterns between feature elements within their corresponding segments are still hidden in the high-dimensional data. Traditional methods using global similarity calculation or simple difference statistics are difficult to capture the non-linear response relationship between vibration signals and baseline features in local intervals. For example, weak anomalies in high-frequency intervals may manifest as co-offset of eigenvalue or suppression effect of specific frequency band energy, and such complex patterns cannot be fully characterized by scalar metrics. Therefore, in the technical solution of this application, to reveal the dynamic offset law of vibration signals relative to the baseline, each pair of corresponding vibration signal waveform local feature ordered coding vectors and baseline local feature ordered coding vectors in the sequence of vibration signal waveform local feature ordered coding vectors and the sequence of baseline local feature ordered coding vectors are input into the transfer response inference unit to obtain a sequence of initial vibration waveform offset baseline local transfer response coding matrices.

[0034] In the specific implementation process, the system pairs the sequences of vibration signal waveform local feature ordered coding vectors and baseline local feature ordered coding vectors group by group according to the intensity interval, and each pair of corresponding local vectors (such as the high-intensity segment of the vibration signal and the high-intensity segment of the baseline) are input into the transfer response inference unit; this unit deeply models the element-level association between the two vectors through a multi-layer non-linear transformation network, for example, analyzing whether the eigenvalue of the vibration signal shows abnormal activation in a specific intensity interval, and whether there is conditional dependence or phase shift with the baseline feature. During this process, by calculating the similarity difference between the two vectors, the hidden transfer law during their interaction (such as the conduction suppression or co-resonance of vibration energy in a certain interval) is captured, and then the response intensity of each feature element to the other vector is recorded in matrix form. This transfer response mechanism based on deep interaction modeling significantly improves the detection rate and classification accuracy of complex abnormal patterns in the intelligent water service scenario, and at the same time provides high-confidence feature support for the optimization of dynamic data compression strategies in the edge computing environment.

[0035] In a specific example of the present application, the sequences of the local feature ordered coding vectors of the vibration signal waveform and the local feature ordered coding vectors of the baseline are input into the transfer response inference unit for each corresponding local feature ordered coding vector of the vibration signal waveform and the local feature ordered coding vector of the baseline in the sequences to obtain a sequence of initial vibration waveform offset baseline local transfer response coding matrices; where the formula is:

[0036] Wherein, represents function, is between the initial initial vibration waveform offset baseline local transfer response coding matrix, represents matrix multiplication, represents the weight matrix.

[0037] Furthermore, a spatial measure regularity constraint based on a Poisson-like process is performed on each initial vibration waveform offset baseline local transfer response coding matrix in the sequence of initial vibration waveform offset baseline local transfer response coding matrices to obtain a sequence of vibration waveform offset baseline local transfer response coding matrices. It should be understood that although the sequence of initial vibration waveform offset baseline local transfer response coding matrices has captured the interaction pattern between local features through the transfer response inference unit, the directly generated matrix may have insufficient model generalization ability due to over-sensitivity in local regions or noise interference. For example, the vibration signal in a certain high-intensity section may generate false strong transfer responses due to random fluctuations, but there is actually no abnormality. If such matrices are directly used for global analysis, it is easy to cause misjudgment or resource waste. Therefore, in the preferred example of the present application, a spatial measure regularity constraint based on a Poisson-like process is performed on each initial vibration waveform offset baseline local transfer response coding matrix in the sequence of initial vibration waveform offset baseline local transfer response coding matrices to obtain a sequence of vibration waveform offset baseline local transfer response coding matrices. That is, by introducing statistical prior constraints, the expression intensity and stability of the local transfer response matrix are balanced, the risk of overfitting is suppressed, and at the same time, the abnormal offset patterns in different intensity intervals have interpretable statistical regularities.

[0038] In the specific implementation process, the edge computing center maps the row vector length of each initial transfer response matrix into a local size parameter, constructs a Poisson-like distribution model with the row vector length as the intensity factor, and presets the expected connection probability distribution of the pipe network vibration mode in the stable state. By dynamically adjusting the spatial measure constraints of each matrix row vector, the transfer response intensity in the local area is probabilistically aligned with the global statistical law. At the same time, the mean expectation of the connection probability is iteratively corrected to suppress the abnormally high response values caused by accidental noise or segmentation granularity deviation, ensuring that the difference expression between the pipe vibration and the baseline in different intensity intervals not only retains local specificity but also conforms to the prior law of the overall pipe network operation mode. Through the regularization mechanism of the Poisson-like process, the persistent low-frequency vibration difference caused by slight pipe leakage can penetrate the random noise interference and be stably characterized in the transfer response matrix. In addition, the dynamic correction of the spatial measure automatically filters out the pseudo-abnormal signals caused by instantaneous working conditions such as sudden changes in water flow velocity, while the progressive feature shift of real equipment failures is strongly captured.

[0039] In this example, the following formula is used to perform spatial measure regularity constraint based on the Poisson-like process on each initial vibration waveform offset baseline local transfer response coding matrix in the sequence of the initial vibration waveform offset baseline local transfer response coding matrix to obtain the sequence of the vibration waveform offset baseline local transfer response coding matrix; where, the formula is:

[0040] Where, means flattening the matrix into a vector, is the th initial vibration waveform offset baseline local transfer response coding vector obtained after flattening , means norm operation, means exponential operation, is the number of initial vibration waveform offset baseline local transfer response coding matrices in the sequence of the initial vibration waveform offset baseline local transfer response coding matrix, is the row vector length of the initial initial vibration waveform offset baseline local transfer response coding matrix, means mean expectation, is the edge connection representation of the spatial measure, means transfer response connection probability, is the optimized expected degree.

[0041] Subsequently, a transfer response inference sequence transfer is performed on the sequence of the vibration waveform baseline-offset local transfer response encoding matrices to obtain a vibration waveform baseline-offset response encoding vector. It should be understood that although the sequence of local transfer response encoding matrices has retained the interaction details in different intensity intervals through regularization constraints, its scattered local features cannot directly reflect the overall offset trend between the vibration signal and the baseline. For example, the persistent anomaly in a certain low-frequency interval may present a compensatory fluctuation in the transfer response matrix in the medium and high-frequency intervals. If a single matrix is analyzed in isolation, it may be misjudged as a normal fluctuation or local noise, resulting in the omission of key abnormal events. Therefore, in the technical solution of this application, a transfer response inference sequence transfer is performed on the sequence of the vibration waveform baseline-offset local transfer response encoding matrices to obtain a vibration waveform baseline-offset response encoding vector.

[0042] In the specific implementation process, the system inputs the sequence of the vibration waveform baseline-offset local transfer response encoding matrices into the sequence modeling module in the order of feature intensity intervals (such as from low frequency to high frequency or from high value to low value). For example, a Transformer structure based on the attention mechanism is adopted to perform adaptive weight assignment on each element in the matrix sequence, focusing on the intervals with significant abnormal features (such as sudden offsets in the high-frequency band or persistent fluctuations in the low-frequency band), while weakening the contribution of the noise interference regions. During this process, the model captures the temporal dependence or intensity gradient relationship between the matrices, learns the abnormal propagation path of the vibration signal in different intensity intervals (such as the energy diffusion from high-frequency anomalies to medium frequency), and dynamically aggregates the local response information; finally, the sequence information is compressed into a single high-dimensional encoding vector through multiple non-linear transformations to obtain a vibration waveform baseline-offset response encoding vector. In this way, the limitation of local analysis is broken through, and complex abnormal patterns across intensity intervals (such as the coupling effect of high-frequency pulses and low-frequency pressure fluctuations caused by pipeline leakage) can be recognized, significantly reducing the false omission risk caused by local misjudgment.

[0043] In a specific example of this application, the following formula is used to perform a transfer response inference sequence transfer on the sequence of the vibration waveform baseline-offset local transfer response encoding matrices to obtain a vibration waveform baseline-offset response encoding vector; where the formula is:

[0044] where, and are respectively the 1st and the th initial vibration waveform baseline-offset local transfer response encoding matrices in the sequence of the initial vibration waveform baseline-offset local transfer response encoding matrices, represents function, is the vibration waveform baseline-offset response encoding vector.

[0045] More specifically, the S313 obtains the degree of deviation based on the vibration waveform offset baseline response coding feature. It should be understood that although the vibration waveform offset baseline response coding vector has integrated the interaction pattern across the intensity range through global sequence transfer, its essence is still a high-dimensional abstract feature and cannot be directly used to quantify the actual degree of deviation of the equipment state. For example, a certain coding vector may imply both high-frequency instantaneous noise and low-frequency continuous anomalies. If it only relies on manual experience or simple threshold judgment, it is difficult to accurately distinguish the actual impact level on water supply safety. Therefore, in the technical solution of the present application, the vibration waveform offset baseline response coding vector is decoded and regressed to obtain the degree of deviation. In the specific implementation process, the system first inputs the vibration waveform offset baseline response coding vector into a pre-trained decoding regression model (such as a multi-layer perceptron or a lightweight neural network). The model deconstructs the key information in the high-dimensional features layer by layer by learning the nonlinear mapping relationship between the coding vector and the true deviation value in the historical data. In this process, the vibration waveform offset baseline response encoding vector is transmitted layer by layer through the weight matrix and nonlinear activation function, and finally a scalar deviation value (such as a normalized value from 0 to 1) is output, which is then passed to the data classification module to trigger the corresponding compression and transmission strategy. The degree of deviation after decoding regression can intuitively reflect whether the operating status of the equipment has changed substantially during the current monitoring period, providing a scientific basis for the subsequent multi-level label division such as highly suspected abnormality, slightly suspected abnormality or normal.

[0046] Specifically, the S32 determines the data classification label based on the degree of deviation, and the data classification labels include highly suspected abnormality, slightly suspected abnormality and normal. Here, the degree of deviation reflects the comprehensive distance between the current vibration mode and the normal state of the equipment in the multi-dimensional feature space. In the specific implementation, the system automatically maps the degree of deviation into three types of data classification labels: highly suspected abnormality, slightly suspected abnormality and normal according to the preset multi-level threshold or dynamic discrimination model. In actual operation, if the degree of deviation of a certain vibration signal far exceeds the historical baseline range, that is, it shows significant mutation or abnormal characteristics, it will be marked as "highly suspected abnormality"; if the degree of deviation is slightly higher than the normal fluctuation but does not reach the serious threshold, it is classified as "mildly suspected abnormality"; and when the degree of deviation is within the historical fluctuation range, it is considered "normal". This classification mechanism not only fully taps the deep-level information in the original vibration signal, but also greatly improves the flexibility and accuracy of the system's response to various events.

[0047] Specifically, in S33, based on the data classification label, a data transmission strategy and a data compression strategy for the original vibration signal are specified. In a specific example of the present application, when the data classification label is highly suspected of being abnormal, the data transmission strategy is to retain the original vibration signal with high fidelity and the vibration signal waveform coding features, and the data compression strategy is to use a near-lossless or high-quality lossy compression algorithm; when the data classification label is mildly suspected of being abnormal, the data transmission strategy is to retain the vibration signal waveform coding features, and the data compression strategy is to use a high-rate lossy compression algorithm to compress the vibration signal waveform coding features; when the data classification label is normal, the data transmission strategy is to retain the principal components of the features of the vibration signal waveform coding features, and the data compression strategy is to use a high-rate lossy compression algorithm to compress the principal components of the features of the vibration signal waveform coding features.

[0048] Particularly, in S4, the edge computing center processes the original vibration signal based on the data transmission strategy and the data compression strategy to obtain an adaptive compression data packet, and transmits it to the cloud platform. Here, it should be understood that uploading all the original vibration signals without distinction will not only cause a great waste of network bandwidth and cloud storage resources, but may also lead to key abnormal information being submerged by a large amount of normal data, affecting the monitoring response efficiency. In the technical solution of the present application, after the data classification label is determined in the previous steps, the edge computing center can adopt a differentiated processing strategy for data of different levels, thereby maximizing the system performance and economy. In this process, the edge computing center will call the corresponding software module to perform format conversion, feature extraction, intelligent coding, and multi-level compression on the original time series signal, and upload the finally generated data packet to the cloud platform safely and efficiently according to the preset protocol. In this way, it is ensured that key abnormal events can be delivered to the cloud in the highest priority and the most complete form in a timely manner, providing strong support for remote operation and maintenance, fault tracing, etc.; in addition, by completing preliminary intelligent analysis and screening on the edge side, the overall response speed can be greatly improved, making the intelligent water service platform have stronger real-time performance, scalability, and sustainable development capabilities.

[0049] After the edge computing center processes these signals with data transmission and compression strategies based on data classification labels, the system can achieve true data intelligent monitoring. Specifically, on the cloud platform, all uploaded data packets are automatically archived, indexed, and aggregated according to their classification labels and feature content. For data highly suspected of being abnormal, the system will prioritize in-depth analysis and visualization, and link the alarm mechanism to promptly notify the operation and maintenance personnel to achieve rapid response and traceability tracking for major faults or emergencies. For data mildly suspected of being abnormal and normal data, methods such as batch processing and trend analysis are used to support equipment health assessment, water usage behavior modeling, and long-term operation and maintenance optimization. In addition, by continuously comparing historical data with real-time new data, the cloud platform can continuously optimize the abnormal discrimination model, improve self-learning and self-adaptive capabilities, and make the entire monitoring system have a stronger intelligent level.

[0050] This collaborative architecture with the edge computing as the outpost and the cloud platform as the decision-making center enables the large-scale and diverse vibration signals generated by ultrasonic water meters to be managed orderly and efficiently. The system can not only accurately identify various water usage abnormalities, equipment failures, and pipeline network risks, but also dynamically adjust the monitoring strategy to adapt to different scenario requirements, thus achieving truly efficient, safe, economical, and scalable data intelligent monitoring in the field of smart water.

[0051] In summary, the data intelligent monitoring method based on ultrasonic water meters according to the embodiments of the present application is clarified. It determines the data classification label by comparing the vibration characteristics of the ultrasonic sensor vibration data within a predetermined time period with the recent historical feature baseline, which enables the system to intelligently identify the abnormal degree of the current water meter operation state. Further, based on the data classification label, a data transmission strategy and a data compression strategy for the original vibration signal are specified, so as to maximize the optimization of data transmission and storage efficiency on the premise of ensuring that key information is not lost. This adaptive strategy not only solves the problem of low data transmission and storage efficiency in traditional solutions, but more importantly, it ensures that sufficient detailed data can be obtained at critical moments, while efficiently utilizing resources during normal operation, achieving the best balance between data integrity, accuracy, and transmission and storage costs.

[0052] Furthermore, a data intelligent monitoring system based on ultrasonic water meters is also provided.

[0053] Figure 5 It is a block diagram of the data intelligent monitoring system based on ultrasonic water meters according to the embodiments of the present application. As Figure 5As shown, the data intelligent monitoring system 300 based on an ultrasonic water meter according to an embodiment of the present application includes: a raw vibration signal acquisition module 310, configured to receive, by an edge computing center, raw vibration signals of a preset time period collected by a vibration sensor of a target ultrasonic water meter; a vibration feature extraction module 320, configured to extract vibration features of the raw vibration signals by the edge computing center to obtain vibration signal waveform coding features; a data processing strategy generation module 330, configured to, based on a comparison result between the vibration signal waveform coding features and a recent historical feature baseline, specify a data transmission strategy and a data compression strategy for the raw vibration signals by the edge computing center; and a data processing module 340, configured to process the raw vibration signals by the edge computing center based on the data transmission strategy and the data compression strategy to obtain an adaptive compression data packet and transmit the same to a cloud platform.

[0054] As described above, the data intelligent monitoring system 300 based on an ultrasonic water meter according to an embodiment of the present application can be implemented in various wireless terminals, such as a server having a data intelligent monitoring algorithm based on an ultrasonic water meter. In a possible implementation manner, the data intelligent monitoring system 300 based on an ultrasonic water meter according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the data intelligent monitoring system 300 based on an ultrasonic water meter can be a software module in an operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the data intelligent monitoring system 300 based on an ultrasonic water meter can also be one of many hardware modules of the wireless terminal.

[0055] Alternatively, in another example, the data intelligent monitoring system 300 based on an ultrasonic water meter and the wireless terminal can also be separate devices, and the data intelligent monitoring system 300 based on an ultrasonic water meter can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0056] The embodiments of the present disclosure have been described above. The above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art in the technical field without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technologies in the market, or to enable other ordinary skill in the art in the technical field to understand the disclosed embodiments.

Claims

1. A data intelligent monitoring method based on an ultrasonic water meter, characterized in that, Including: The edge computing center receives the original vibration signals of a preset time period collected by the vibration sensor of the target ultrasonic water meter; The edge computing center extracts the vibration characteristics of the original vibration signals to obtain the vibration signal waveform coding characteristics; Based on the comparison result between the vibration signal waveform coding characteristics and the recent historical feature baseline, the edge computing center designates the data transmission strategy and data compression strategy for the original vibration signals; The edge computing center processes the original vibration signals based on the data transmission strategy and data compression strategy to obtain the adaptive compression data packets and transmits them to the cloud platform.

2. The data intelligent monitoring method based on an ultrasonic water meter according to claim 1, wherein The edge computing center extracts the vibration characteristics of the original vibration signals to obtain the vibration signal waveform coding characteristics, including: Performing vibration feature extraction based on one-dimensional convolutional coding on the original vibration signals to obtain the vibration signal waveform feature coding vectors as the vibration signal waveform coding characteristics.

3. The data intelligent monitoring method based on an ultrasonic water meter according to claim 1, wherein Based on the comparison result between the vibration signal waveform coding characteristics and the recent historical feature baseline, the edge computing center designates the data transmission strategy and data compression strategy for the original vibration signals, including: The edge computing center compares the vibration signal waveform coding characteristics with the recent historical feature baseline to obtain the deviation degree; Based on the deviation degree, determine the data classification label, and the data classification label includes highly suspected anomaly, mildly suspected anomaly, and normal; Based on the data classification label, designate the data transmission strategy and data compression strategy for the original vibration signals.

4. The data intelligent monitoring method based on an ultrasonic water meter according to claim 3, wherein The edge computing center compares the vibration signal waveform coding characteristics with the recent historical feature baseline to obtain the deviation degree, including: Obtaining the baseline feature vector of the recent historical vibration signals as the recent historical feature baseline; Inputting the vibration signal waveform feature coding vectors and the baseline feature vectors into the fine-grained waveform feature difference modeling network to obtain the vibration waveform offset baseline response coding vectors as the vibration waveform offset baseline response coding characteristics; Based on the vibration waveform offset baseline response coding characteristics, obtain the deviation degree.

5. The data intelligent monitoring method based on an ultrasonic water meter according to claim 4, wherein, Inputting the vibration signal waveform feature coding vectors and the baseline feature vectors into the fine-grained waveform feature difference modeling network to obtain the vibration waveform offset baseline response coding vectors, including: Performing ordered arrangement and feature segmentation on the vibration signal waveform feature coding vectors and the baseline feature vectors to obtain the sequence of vibration signal waveform local feature ordered coding vectors and the sequence of baseline local feature ordered coding vectors; Performing vibration waveform-baseline transfer response coding analysis on the sequence of vibration signal waveform local feature ordered coding vectors and the sequence of baseline local feature ordered coding vectors to obtain the vibration waveform offset baseline response coding vectors.

6. The data intelligent monitoring method based on an ultrasonic water meter according to claim 5, characterized in that Performing ordered arrangement and feature segmentation on the vibration signal waveform feature coding vectors and the baseline feature vectors to obtain the sequence of vibration signal waveform local feature ordered coding vectors and the sequence of baseline local feature ordered coding vectors, including: Performing ordered arrangement on the vibration signal waveform feature coding vectors and the baseline feature vectors based on the eigenvalue size to obtain the vibration signal waveform feature ordered arrangement coding vectors and the baseline feature ordered arrangement coding vectors; Perform equal - granularity feature segmentation on the ordered - arrangement coded vector of the vibration signal waveform features and the ordered - arrangement coded vector of the baseline features to obtain a sequence of ordered - coded vectors of local vibration signal waveform features and a sequence of ordered - coded vectors of local baseline features.

7. The data intelligent monitoring method based on an ultrasonic water meter according to claim 6, characterized in that, Perform vibration waveform - baseline transfer response coding analysis on the sequence of ordered - coded vectors of local vibration signal waveform features and the sequence of ordered - coded vectors of local baseline features to obtain a vibration waveform offset baseline response coded vector, including: Input each group of corresponding ordered - coded vectors of local vibration signal waveform features and ordered - coded vectors of local baseline features in the sequence of ordered - coded vectors of local vibration signal waveform features and the sequence of ordered - coded vectors of local baseline features into the transfer response inference unit to obtain a sequence of initial vibration waveform offset baseline local transfer response coding matrices; Perform spatial measure regularity constraint based on the Poisson - like process on each initial vibration waveform offset baseline local transfer response coding matrix in the sequence of initial vibration waveform offset baseline local transfer response coding matrices to obtain a sequence of vibration waveform offset baseline local transfer response coding matrices; Perform transfer response inference sequence transfer on the sequence of vibration waveform offset baseline local transfer response coding matrices to obtain a vibration waveform offset baseline response coded vector.

8. The data intelligent monitoring method based on an ultrasonic water meter according to claim 7, characterized in that, Based on the vibration waveform offset baseline response coding features, obtain the deviation degree, including: Decode and regress the vibration waveform offset baseline response coded vector to obtain the deviation degree.

9. The data intelligent monitoring method based on an ultrasonic water meter according to claim 3, wherein Based on the data classification label, specify the data transmission strategy and data compression strategy for the original vibration signal, including: When the data classification label is highly suspected of being abnormal, the data transmission strategy is to retain the original vibration signal with high - fidelity and the vibration signal waveform coding features, and the data compression strategy is to use a near - lossless or high - quality lossy compression algorithm; When the data classification label is mildly suspected of being abnormal, the data transmission strategy is to retain the vibration signal waveform coding features, and the data compression strategy is to use a high - magnification lossy compression algorithm to compress the vibration signal waveform coding features; When the data classification label is normal, the data transmission strategy is to retain the feature principal components of the vibration signal waveform coding features, and the data compression strategy is to use a high - magnification lossy compression algorithm to compress the feature principal components of the vibration signal waveform coding features.

10. An intelligent data monitoring system based on an ultrasonic water meter, characterized in that, Include: An original vibration signal acquisition module, which is used for the edge computing center to receive the original vibration signal of a preset time period collected by the vibration sensor of the target ultrasonic water meter; A vibration feature extraction module, which is used for the edge computing center to extract the vibration features of the original vibration signal to obtain the vibration signal waveform coding features; A data processing strategy generation module, which is used for the edge computing center to specify the data transmission strategy and data compression strategy for the original vibration signal based on the comparison result between the vibration signal waveform coding features and the recent historical feature baseline; A data processing module, which is used for the edge computing center to process the original vibration signal based on the data transmission strategy and data compression strategy to obtain an adaptive compression data packet and transmit it to the cloud platform.

Images