Static sounding data acquisition quality control method and system fused with abnormal early warning

Abstract

The invention discloses a static sounding data acquisition quality management and control method and system fused with abnormal early warning, and relates to the field of data analysis and processing, and the method comprises the steps: an edge end collects original static sounding data through associated static sounding equipment, and carries out edge abnormal early warning through an embedded local geologic control model; based on the edge anomaly early warning result, performing transmission pre-processing on the original static sounding data in combination with a preset packaging strategy, obtaining transmission state static sounding data, and transmitting the transmission state static sounding data to the cloud; the cloud carries out depth anomaly identification on the transmission state static sounding data, and the depth anomaly identification comprises knowledge verification and mode verification; and performing structured storage on the transmission state static sounding data in combination with the depth anomaly recognition result and a preset continuity rule set. The problems that in the prior art, reliability of data transmission and storage is low, the data anomaly detection effect lags behind, and safety verification of data storage is lacked are solved.

Description

Technical Field

[0001] This invention relates to the field of data analysis and processing, specifically to a method and system for quality control of static cone penetration test data acquisition that integrates anomaly early warning. Background Technology

[0002] Data quality management involves monitoring and managing each stage of the data lifecycle to identify and resolve potential data quality issues, thereby enhancing the value of the data.

[0003] Traditional methods rely heavily on manual review after data collection or simple threshold checks, which cannot provide real-time warnings and interventions the moment data is generated. This leads to the continuous generation of abnormal data. There is a lack of an intelligent detection system that can coordinate at the edge and cloud and balance real-time performance and accuracy. This results in low reliability of data transmission and storage, delayed detection of data anomalies, and a lack of security verification for data storage.

[0004] Therefore, a security and quality control solution is needed that can achieve intelligent control from the source of data collection to cloud storage, enabling real-time and accurate monitoring of static cone penetration test data. Summary of the Invention

[0005] This application provides a method and system for quality control of static cone penetration test data acquisition that integrates anomaly early warning, addressing the problems of low reliability of data transmission and storage, delayed data anomaly detection, and lack of data storage security verification in existing technologies.

[0006] In view of the above problems, this application provides a method and system for quality control of static cone penetration test data acquisition that integrates anomaly early warning.

[0007] Firstly, this application provides a method for quality control of static cone penetration test data acquisition that integrates anomaly early warning, the method comprising: The edge end collects raw static cone penetration data through associated static cone penetration equipment and uses an embedded local quality control model to provide early warning of edge anomalies. Based on the edge anomaly warning results, the original static penetration data is preprocessed for transmission using a preset packaging strategy to obtain the static penetration data in the transmission state and transmit it to the cloud. The cloud performs deep anomaly identification on the transmitted static penetration data, wherein the deep anomaly identification includes knowledge verification and pattern verification; The transmission-state static penetration data is structured and stored by combining the depth anomaly identification results with a preset set of continuity rules.

[0008] Secondly, this invention provides a static cone penetration test data acquisition quality control system that integrates anomaly early warning, comprising: The anomaly warning module is used to collect raw static cone penetration data at the edge via associated static cone penetration equipment and to provide edge anomaly warnings through an embedded local quality control model. The data transmission processing module is used to perform pre-transmission processing on the original static cone penetration data based on the edge anomaly warning result and in combination with a preset packaging strategy, to obtain the static cone penetration data in the transmission state and transmit it to the cloud. A data depth identification module is used to perform depth anomaly identification on the transmitted static penetration data in the cloud, wherein the depth anomaly identification includes knowledge verification and pattern verification. The data storage module is used to perform structured storage of the transmitted static penetration data by combining the deep anomaly identification results with a preset set of continuity rules.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application firstly intercepts quality issues by collecting data in real time at the edge and using a local quality control model for early warning, thereby improving response speed and reducing reliance on a stable network. Pre-emptive intelligent screening filters out obviously abnormal data, reducing the processing burden and network transmission pressure on the cloud. Secondly, based on the early warning results, differentiated pre-transmission processing and workflows are executed to preserve localized diagnostic information for abnormal data and prepare for the secure and efficient transmission of normal data. Compression and encryption optimize network resources while ensuring data security, and intelligent decision-making based on data quality improves overall resource allocation efficiency. Thirdly, deep anomaly identification is performed in the cloud, integrating domain knowledge and data patterns. Regional knowledge bases and anomaly pattern libraries are used to perform multi-dimensional verification of data, discovering complex anomalies that are difficult to identify at the edge, ensuring the reliability and rationality of the data entering the database. Finally, qualified data is stored in a structured and versioned manner using continuity rules. Logical coherence verification is performed before data entry, ensuring the integrity and accuracy of the data sequence. Through structured storage and version management, high-quality data is transformed into standardized data assets that are easy to query and analyze. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the static cone penetration test data acquisition quality control method that integrates anomaly early warning, as described in this application. Figure 2 This is a schematic diagram of the static cone penetration test data acquisition quality control system that integrates anomaly early warning, as described in this application.

[0012] In the attached diagram, the components represented by each number are as follows: Anomaly warning module 11, data transmission processing module 12, data depth recognition module 13, data storage module 14. Detailed Implementation

[0013] This application provides a static cone penetration test data acquisition quality control method that integrates anomaly early warning, which solves the problems of low reliability of data transmission and storage, delayed data anomaly detection effect, and lack of data storage security verification.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0016] The present invention will now be described in detail with reference to the accompanying drawings.

[0017] Example 1, as Figure 1 As shown, this application provides a method for quality control of static cone penetration test data acquisition that integrates anomaly early warning, the method comprising: S10: The edge end collects raw static cone penetration data through the associated static cone penetration equipment and performs edge anomaly early warning through the embedded local quality control model; In this embodiment, the edge device refers to a computing device deployed at or near the static cone penetration test equipment, possessing data acquisition, preliminary calculation, and communication capabilities; the associated static cone penetration test equipment refers to the static cone penetration test host and sensor system that are physically connected to the edge device via wired or wireless means and communicate with it via data; the raw static cone penetration test data refers to the initial data stream from the static cone penetration test equipment sensors without any cloud processing; the embedded local quality control model refers to a software algorithm module that is pre-deployed and stored on the edge device and can perform real-time quality judgment on the raw data; the edge anomaly early warning refers to using the local quality control model to quickly analyze the raw data collected in real time at the source of data acquisition and immediately provide a risk judgment and early warning output regarding the existence of obvious anomalies.

[0018] Specifically, the edge device acts as the field control center, first driving the connected static cone penetration test device to begin penetration and data collection according to preset parameters. The data packets are then sent to the edge device's built-in local quality control model for real-time analysis, quickly screening the data to determine if any obvious anomalies exist.

[0019] Step S10 in the method provided in this application embodiment includes: The edge device sends preset acquisition parameters to the static cone penetration test device to trigger continuous sensing acquisition, and associates and stores the acquisition parameters with the acquired sensing data to obtain the original static cone penetration test data; A copy of the original static penetration data is created and cached in a buffer at the edge, wherein the buffer is configured in a ring cache mode; The raw static penetration data is input into the local quality control model to perform edge anomaly warning, wherein the local quality control model includes parallel rule channels and lightweight neural network channels; If the outputs of both the rule channel and the lightweight neural network channel are normal, then the output edge anomaly warning result is normal.

[0020] In this embodiment, the edge device first sends preset acquisition parameters to the static cone penetrometer to trigger continuous sensing acquisition, and then associates and stores the acquisition parameters with the acquired sensing data to obtain the original static cone penetrometer data. The preset acquisition parameters refer to a set of instructions pre-set at the edge device to control the operation of the static cone penetrometer before acquisition begins, ensuring data consistency and comparability. Sending parameters refers to the process by which the edge device, acting as the control host, sends parameter instructions to the main controller of the static cone penetrometer through a communication interface. Association and storage refer to the edge device simultaneously recording the complete set of preset acquisition parameters used to trigger this acquisition when receiving and storing the data packets returned by the sensors, and binding the parameters as metadata with the original sensing data.

[0021] After the equipment is ready, the operator loads the preset acquisition parameters for the task from the edge device. The edge device accurately sends the preset acquisition parameters to the static cone penetration test device, instructing the device to start and operate according to the established procedures. During device operation, the sensors synchronously generate continuous raw voltage or digital signals such as cone tip resistance and sidewall friction resistance, and transmit them back to the edge device in real time. While receiving the raw sensor data, the edge device binds and stores it with the sent preset acquisition parameters. This yields raw static cone penetration test data with complete acquisition context information.

[0022] For example, during a highway subgrade survey, the edge terminal sends preset acquisition parameters to the static cone penetration testing (CPPT) equipment according to the task requirements. The sampling mode is equal depth spacing, 2 cm apart; the calibrated penetration rate is 1.8 cm / s; and the cone tip resistance sensor range is 50 MPa. After the equipment is started, penetration and acquisition are performed strictly according to the parameters. Each data packet received by the edge terminal, containing depth, cone tip resistance, and sidewall friction, is automatically tagged with metadata, collectively forming the raw static cone penetration data.

[0023] Secondly, a copy of the original static probe data is created and cached in a buffer at the edge device, configured as a circular cache. The copy is identical to the original static probe data; the buffer is a dedicated area in the edge device's memory or high-speed storage medium used to temporarily store the data copy; the circular cache mode is an efficient cache management strategy, typically implemented using read / write pointers, and is a fixed-size first-in-first-out buffer implementation.

[0024] Specifically, after acquiring the raw static penetration data at the edge, it is immediately executed in parallel, automatically creating a copy of the data and sending it to a buffer operating in a circular cache mode. This ensures that the latest data copy is appended to the end of the circular cache. If the cache space is full, when a new data copy arrives, it will overwrite the first part of the data in the circular cache, achieving a rolling update of the cache.

[0025] For example, during highway subgrade surveying, for each point collected at the edge, a copy is simultaneously sent to the local quality control model and stored in a circular buffer with a capacity equal to the nearest 5 meters of data. When the penetration depth reaches 7.0 meters, data copies from depth 2.0 meters to 7.0 meters are saved. When the penetration depth reaches 12.0 meters, data copies from depth 7.0 meters to 12.0 meters are saved, with the earliest 5 meters of data automatically overwritten.

[0026] Next, the raw static penetration test data is input into the local quality control model to perform edge anomaly early warning. The local quality control model includes a parallel rule-based channel and a lightweight neural network channel. The local quality control model is a quality judgment algorithm module deployed at the edge, employing a dual-channel parallel hybrid intelligent architecture. The rule-based channel is a channel that makes judgments based on explicit logical rules. The lightweight neural network channel is a compressed and optimized deep learning model channel that has learned complex feature patterns of anomaly data through training on a large amount of historical anomaly data.

[0027] Specifically, the raw static cone penetration test data is synchronously fed to two channels of the local quality control model for edge-end quality control. The two channels operate independently and make preliminary judgments separately. The rule channel applies logical rules to screen the data; the lightweight neural network channel takes a short continuous data segment as input, extracts features, and outputs probability values ​​to determine whether the data performance is similar to normal samples in historical data. The judgment processes of the two channels do not interfere with each other and are judged independently.

[0028] Furthermore, if both the rule-based channel and the lightweight neural network channel output no anomalies, the edge anomaly warning result is output as "no anomaly". The judgment results of the rule-based channel and the lightweight neural network channel on the current data are collected. Only when the outputs of both channels are clearly no anomalies is the edge warning successfully passed, and the edge anomaly warning result is officially output as "no anomaly". If either channel outputs an anomaly, the final warning result is "anomaly", which maximizes the initial quality of the uploaded data.

[0029] In step S10 of the method provided in this application embodiment, the step of constructing the local quality control model includes: Obtain the task description for the static penetration test, analyze the task description to extract control clauses, and perform regularization processing on the control clauses to obtain the rule channels. Obtain a historical static penetration anomaly dataset, and perform binarization simplification on the anomaly discrimination results in the historical static penetration anomaly dataset to obtain a sample static penetration anomaly dataset. Based on the sample static probing anomaly dataset, a lightweight neural network model is constructed and trained. The trained lightweight neural network model is output as the lightweight neural network channel and integrated with the rule channel to obtain the local quality control model.

[0030] In this embodiment, the task specification for the static cone penetration test is first obtained, the task specification is analyzed to extract control clauses, and the control clauses are then regularized to obtain rule channels. The task specification for the static cone penetration test specifies the technical requirements, operational standards, and quality objectives for the static cone penetration exploration operation. Control clauses refer to the technical requirement statements in the task specification that relate to the data acquisition process and quality, and can be transformed into specific inspection rules. Regularization is the process of converting the control clauses described in natural language into logical rules or regular expression patterns that a computer program can precisely understand and execute.

[0031] Specifically, the implementation personnel first need to obtain the static cone penetration test task specification. By analyzing the text of the task specification, all specific clauses related to data quality control are extracted. Then, the clauses described in language are regularized. All the regularized rules are then integrated to form a rule channel module that can be called by edge applications.

[0032] Secondly, a historical static cone penetration test anomaly dataset is obtained, and the anomaly identification results in the historical static cone penetration test anomaly dataset are simplified by binarization to obtain a sample static cone penetration test anomaly dataset. The historical static cone penetration test anomaly dataset refers to the collection of static cone penetration test data accumulated in the past that were indeed considered to have quality problems; the anomaly identification results refer to the detailed anomaly descriptions labeled for each sample in the dataset, which may include complex information such as anomaly category, degree, and cause.

[0033] Specifically, a large number of known abnormal data cases are collected from the historical project database to form a historical static cone penetration anomaly dataset. The labeling of the historical static cone penetration anomaly dataset is simplified by uniformly setting the label of all data segments marked as any type of anomaly to "abnormal". At the same time, a large number of normal data segments confirmed to be without any problems are selected from the historical data and their labels are set to "normal". This results in a sample static cone penetration anomaly dataset with clear correspondence between features and labels.

[0034] Finally, based on the sample static probing anomaly dataset, a lightweight neural network model is constructed and trained. The trained lightweight neural network model is output as a lightweight neural network channel, which is then integrated with the rule channel to obtain a local quality control model. The lightweight neural network channel refers to the final model file that, after sufficient training, achieves high accuracy on the validation set and meets the real-time requirements of edge computing.

[0035] The following is a specific example of designing a lightweight neural network architecture: Model Structure: A small network consisting of convolutional layers, pooling layers, and fully connected layers is designed. It receives a sample static anomaly dataset as input. The hidden layers are responsible for extracting features from the input sample static anomaly dataset. The fully connected layers use the ReLU activation function to integrate and transform the features extracted by the hidden layers.

[0036] Model Training: The network was trained using a sample static probing anomaly dataset. Mean squared error was used as the loss function, and the Adam optimizer was used for parameter updates. An initial learning rate was set, and a learning rate decay strategy was employed. Model performance was evaluated on the validation set after each training epoch. Training was terminated when the validation set loss no longer decreased for 10 consecutive epochs. After training, a final evaluation was performed on the test set to obtain the lightweight neural network channel.

[0037] During training, the network learns to automatically extract discriminative features from normal and abnormal data segments. After training, the model is compressed and optimized to ensure fast inference on edge devices. Finally, the optimized neural network model is deployed as a lightweight neural network channel. Simultaneously, the rule channel code module is integrated into a unified framework through software programming, deploying it to a local quality control model at the edge.

[0038] In this embodiment, preset collection parameters are distributed and associated with storage at the edge, ensuring the standardization and consistency of data collection from the source and providing an immutable original basis for subsequent quality judgment. By establishing a circular cache copy, local data recovery capability is provided for network fluctuations or interruptions during data transmission, which is the foundation for reliable transmission and avoids the loss of information in abnormal situations. By constructing a local quality control model, integrating rule channels and lightweight neural network channels, rule review based on explicit knowledge and intelligent pattern screening based on data-driven methods improves the reliability of rule judgment and reduces cloud processing pressure and review burden.

[0039] S20: Based on the edge anomaly warning result, combined with the preset packaging strategy, perform pre-transmission processing on the original static penetration data, obtain the transmission state static penetration data, and transmit it to the cloud; In this embodiment, the preset packaging strategy is a set of rules on how to compress, encrypt, and encapsulate data in preparation for network transmission; pre-transmission processing refers to sending the data to the previous processing operation based on the warning result and the packaging strategy; transmission-state static probe data refers to data in a form that is suitable for efficient transmission through network security after compression, encryption, and formatting; the cloud refers to a remote central server cluster.

[0040] First, based on the preset packaging strategy and the edge anomaly warning results, different action paths are taken to preprocess the original static penetration data. After compression, encryption, and formatting, the data is stored locally or processed into abnormal data.

[0041] Step S20 in the method provided in this application embodiment includes: If the edge anomaly warning result is abnormal, the static penetration test task is terminated, and the edge anomaly warning result, the original static penetration test data and the corresponding context data are packaged and stored as abnormal static penetration test data. If the edge anomaly warning result is no anomaly, the original static cone penetration data is processed before transmission based on a preset packaging strategy, including data compression and data encryption, to obtain the transmitted static cone penetration data. The abnormal static penetration data is stored in the local storage area at the edge, and the transmission static penetration data is transmitted to the cloud through a dedicated encrypted channel.

[0042] In this embodiment of the application, if the edge anomaly warning result is abnormal, the static penetration test task is terminated, and the edge anomaly warning result, the original static penetration test data and the corresponding context data are packaged and stored as abnormal static penetration test data.

[0043] Specifically, when the edge device detects an anomaly in the edge anomaly warning result, a response procedure is immediately triggered. First, the current data acquisition task is terminated to prevent the device from continuing to generate meaningless or even misleading data in the abnormal state. Next, the diagnostic information packaging process is initiated, collecting and packaging all key evidence related to the abnormal event, the edge anomaly warning result, and contextual data serving as environmental clues. The resulting data packet is tagged and stored as abnormal static probing data in the local storage area of ​​the edge device.

[0044] Furthermore, if the edge anomaly warning result is no anomaly, the original static cone penetration test data is preprocessed for transmission based on a preset packaging strategy, including data compression and data encryption, to obtain the transmitted static cone penetration test data. The preset packaging strategy is a standardized operating procedure for preparing data transmission; data compression refers to the technical process of reducing the volume of the original data using a certain algorithm; and data encryption refers to encoding the data using cryptographic algorithms.

[0045] Specifically, after confirming the data's initial compliance, if the local quality control model outputs an edge anomaly warning result indicating no anomalies, the batch of data is deemed suitable for uploading. Subsequently, a preset packaging strategy is invoked to process the raw static cone penetration test data, including data compression, data encryption, and the addition of data packet headers. Completing all these pre-transmission processing steps transforms the raw data into standardized, transportable static cone penetration test data.

[0046] Abnormal static cone penetration test data is stored in the local storage area at the edge, while the transmitting static cone penetration test data is transmitted to the cloud through a dedicated encrypted channel. The dedicated encrypted channel refers to the encrypted communication link for transmitting static cone penetration test data. Specifically, abnormal static cone penetration test data is stored in the local storage area at the edge, remaining on the field device and not uploaded to the network to avoid contaminating the main cloud database; simultaneously, the transmitting static cone penetration test data is transmitted to the cloud through a dedicated encrypted channel for in-depth processing.

[0047] In this embodiment, different processing paths are executed based on the edge anomaly warning results. For data with an anomaly warning, the task is immediately terminated and all diagnostic information is packaged and saved. For data with no anomalies warning, access to the cloud channel is permitted. Data compression and encryption are performed using a preset packaging strategy. This reduces the amount of data transmitted over the network, lowers the reliance on bandwidth and traffic costs for field operations, ensures that data can reach the cloud securely and efficiently, and provides a structured guarantee for reliable operation.

[0048] S30: The cloud performs deep anomaly identification on the transmitted static penetration data, wherein the deep anomaly identification includes knowledge verification and pattern verification; In this embodiment, deep anomaly identification refers to the cloud using its powerful computing capabilities and rich prior knowledge base to perform a more intelligent and complex second-round quality analysis on the uploaded static cone penetration test data; knowledge verification refers to comparing and verifying the data with professional knowledge and regional experience patterns in the field of geological engineering; pattern verification refers to using data mining and pattern recognition technologies to detect whether there are typical abnormal waveforms or feature patterns in the data sequence that characterize equipment failure or operational errors.

[0049] Once the cloud receives the static penetration test data uploaded from the edge device, it first decrypts and decompresses it to recover the data content. Then, it initiates the deep anomaly detection process, performing knowledge verification and pattern verification.

[0050] Step S30 in the method provided in this application embodiment includes: Based on the mission objective zoning information of the static cone penetration test, a regional static cone penetration knowledge base is extracted; The transport-state static penetration data is subjected to multi-dimensional knowledge verification using the regional static penetration knowledge base. If each dimension satisfies the regional static penetration knowledge base, the knowledge verification is output as passed. Based on the regional static cone penetration test knowledge base, obtain multiple data content anomaly patterns corresponding to the target area, and traverse multiple data content anomaly patterns to perform pattern matching on the transmitted static cone penetration test data. If the pattern matching result is empty, output that the pattern verification is successful. If both knowledge verification and pattern verification pass, the output deep anomaly identification result is "no anomaly". The regional static cone penetration test knowledge base includes at least an empirical relationship map of soil properties and penetration parameters.

[0051] In this embodiment, a regional static cone penetration test (SPT) knowledge base is first extracted based on the task objective zoning information of the SPT mission. Here, a static cone penetration test refers to a specific static cone penetration engineering survey operation, typically with a clearly defined geographical location and objective; the task objective zoning information refers to the specific geographical area or geological unit identifier involved in the static cone penetration test; and the regional static cone penetration test knowledge base refers to a pre-established set of data and rules stored in the cloud, containing empirical relationships between typical soil characteristics and static cone penetration parameters in the region.

[0052] Specifically, when the cloud receives static cone penetration test data uploaded from the edge, it first needs to delineate the target area. Then, it locates the regional static cone penetration test knowledge base for that area from the cloud's vast geological and static penetration test experience database.

[0053] Secondly, multi-dimensional knowledge verification is performed on the transport-state static penetration data using a regional static penetration knowledge base. If each dimension satisfies the regional static penetration knowledge base, the knowledge verification is considered successful. Multi-dimensional knowledge verification refers to checking the consistency of the data using rules from the regional knowledge base across multiple different feature dimensions or indicators.

[0054] Specifically, the uploaded static cone penetration test data is compared point-by-point or segment-by-segment with the extracted regional static cone penetration test knowledge base. The key parameters in the data are checked to see if they contradict the general understanding of the region, including checks on numerical ranges and physical relationships. Knowledge verification is considered successful only when the transmitted data conforms to regional empirical knowledge in all checked dimensions.

[0055] Next, based on the regional static cone penetration test knowledge base, multiple data content anomaly patterns corresponding to the target area are obtained. These patterns are then used to perform pattern matching on the transmitted static cone penetration test data. If the pattern matching result is empty, the pattern verification is considered successful. Here, the data content anomaly pattern refers to an abstract pattern describing common anomaly features in the static cone penetration test data; pattern matching uses algorithms to detect whether there are feature segments in the data sequence that match a certain anomaly pattern.

[0056] Specifically, common anomaly patterns applicable to the current mission target zoning are retrieved from the regional static cone penetration test (SPPT) knowledge base. Then, the SPPT data curves in the transport state are scanned one by one using these anomaly patterns. If the data pattern within a certain depth range highly matches the description of a certain anomaly pattern, a match is successful, and the anomaly segment is recorded. Only when all patterns have been scanned and no matches are found is the pattern verification considered successful.

[0057] For example, taking engineering site data as an example, the cloud loads anomaly patterns applicable to this area, including single-point spike noise, continuous constant values ​​greater than 0.5 meters, and data jumps exceeding 50% between adjacent points. During pattern matching: scanning the static cone penetration test data curve in the transmission state reveals that all data points change continuously, with no constant value segments longer than 0.5 meters. Checking the differences between adjacent points shows that the maximum rate of change is within 30%, with no jumps exceeding 50%. No anomalies were matched in any of the predefined patterns, therefore the pattern verification passed.

[0058] If both knowledge verification and pattern verification pass, the output depth anomaly identification result will be "no anomalies." The regional static cone penetration test knowledge base must include at least an empirical relationship map of soil properties and penetration parameters. Specifically, only when all verification steps pass can the uploaded data be considered to have passed the depth quality check, and the depth anomaly identification result will be "no anomalies."

[0059] Step S30 in the method provided in this application embodiment further includes: If at least one of the knowledge verification and pattern verification fails, the deep anomaly identification result will be output as an anomaly. The cloud clears the transmitted static penetration data, terminates the static penetration task, and sends a data anomaly receipt to the edge, marking the transmitted static penetration data as abnormal static penetration data.

[0060] In this embodiment, if at least one of the knowledge verification and pattern verification fails, the deep anomaly identification result is output as an anomaly. Specifically, after the knowledge verification and pattern verification are executed in parallel or sequentially, two independent verification results are received. These results are judged according to the set rules. If both the knowledge verification and pattern verification results are passed, the process proceeds to storage; if either verification result fails, the final result of this deep anomaly identification is determined to be an anomaly.

[0061] Secondly, the cloud clears the static cone penetration data in transit, terminates the static cone penetration task, and sends a data anomaly receipt to the edge, marking the transit static cone penetration data as abnormal static cone penetration data. Here, terminating the static cone penetration task refers to the cloud acting as the central control node; the data anomaly receipt refers to a notification message generated by the cloud and sent to the edge.

[0062] Specifically, after the depth anomaly identification result is determined to be abnormal in the cloud, the transmitted static cone penetration test data temporarily stored in the cloud cache is first cleared. Then, a remote control command is generated to terminate the task, and a data anomaly receipt is generated simultaneously. Finally, the uploaded data packet is marked as abnormal static cone penetration test data in the cloud task management log. Upon receiving the receipt at the field edge, an alarm sounds, the operator suspends work, and can focus on checking equipment calibration or verifying borehole records based on the prompts in the receipt, thus enabling a rapid response.

[0063] In this embodiment, a deep anomaly identification verification method is constructed through knowledge verification and pattern verification. By invoking a regional static probing knowledge base strongly correlated with the task area for multi-dimensional comparison, the method detects the presence of known fault characteristics from the data curve shape itself, achieving deep verification and ensuring detection accuracy. When deep identification detects an anomaly, the verification fails, the cloud rejects data entry into the database, proactively clears the data, terminates the remote task, and sends a detailed anomaly receipt to the edge, forcibly interrupting the source of the problematic data. Simultaneously, the data anomaly receipt feeds back the cloud's professional judgment to staff for targeted investigation.

[0064] S40: Combine the depth anomaly identification results with the preset continuity rule set to perform structured storage of the transmission state static penetration data.

[0065] In this embodiment, the continuity rule set is a set of business rules used to determine whether the static cone penetration test data is continuous, complete and logically consistent in terms of depth and acquisition process; structured storage refers to the persistent storage of verified and qualified data according to a predefined data model, which typically includes the data itself, metadata and quality tags.

[0066] Based on the deep anomaly identification results, if the deep anomaly is identified as an anomaly, the cloud will refuse to store it and trigger a cleanup and feedback process. If the deep anomaly identification result is not an anomaly, the data will be processed through a continuity rule set to ensure that the stored data is not only reliable in content but also complete and error-free in logical sequence.

[0067] Step S40 in the method provided in this application embodiment includes: When the depth anomaly identification result is no anomaly, the static penetration data in the transmission state is temporarily stored in the cloud cache. Based on the aforementioned continuous rule set, the acquisition and temporary storage of the static penetration data in the transmission state are performed iteratively. The data length in the cloud cache is continuously monitored until the data length meets the preset collection window. Then, all the data in the cloud cache is transferred to the static cone penetration database for structured and versioned storage. The continuity rule set includes physical continuity rules for adjacent points and continuity rules for the experimental process.

[0068] In this embodiment, when the depth anomaly identification result is no anomaly, the static penetration test data in the transmission state is temporarily stored in the cloud cache area. The cloud cache area refers to a temporary data storage area in the memory of a cloud server or high-speed solid-state storage; temporary storage means that the data is temporarily stored in the cloud cache area, indicating that the data is in a transitional state and will be processed only after certain conditions are met.

[0069] Specifically, after the cloud completes the deep anomaly identification of a batch of static penetration test data and concludes that there are no anomalies, in order to perform continuous verification and possible batch optimization storage, the data packets that have passed the content quality inspection are first moved from the temporary processing queue to the cloud cache for temporary storage. They are then transferred to permanent storage in batches once the accumulated data reaches the collection window.

[0070] Secondly, the acquisition and temporary storage of static cone penetration test data in the transmission state are iteratively performed using a continuity rule set. The continuity rule set is a set of business rules used to verify the completeness and coherence of the static cone penetration test data in terms of acquisition logic. It mainly includes two categories: physical continuity rules for adjacent points and continuity rules for the test process. Iteration refers to the cyclical process of repeatedly acquiring new data, temporary storage, and checking against the rule set. Specifically, after a data packet is temporarily stored, the end depth is recorded. During temporary storage, a continuity check is triggered to see if the continuity rules are met. If they are met, temporary storage is allowed; if not, it means that data loss has occurred, violating the continuity rules, and the data packet will not be temporarily stored normally. It may be marked as a continuity anomaly and trigger corresponding processing.

[0071] Next, continuously monitor the data length in the cloud cache until it meets the preset acquisition window. Then, transfer all data in the cloud cache to the static penetration database for structured and versioned storage. The continuity rule set includes adjacent point physical continuity rules and experimental process continuity rules. Continuous monitoring refers to the background monitoring process periodically or in real-time checking the total amount of data accumulated in the cloud cache for a specific task. Data length is an indicator of the accumulated data in the cache, typically referring to the number of data packets. The preset acquisition window is a pre-configured threshold parameter that triggers batch data persistence operations. Adjacent point physical continuity rules detect the continuity of data in data packets through physical laws. Experimental process continuity rules verify whether data arrives in chronological order, whether there are unexpected interruptions, and whether data packet sequence numbers are consecutive.

[0072] Specifically, during iterative temporary data storage, the data length of the task data in the cache is continuously monitored. When the cumulative depth reaches or exceeds the preset collection window, a batch storage operation is triggered. All data from the accumulated window in the cache is transferred to the formal static probing database. During data import, the data is parsed and entered in a structured manner, and version numbers are assigned or updated. After completion, the cache space is cleared or released to receive data from the next window.

[0073] The method provided in this application embodiment, which involves acquiring and transmitting static penetration data in a transmission state to the cloud, further includes: Iteratively replace the original static penetration data in the edge ring buffer with the transmitted static penetration data; After each replacement is completed, the transmission of the static penetration data in the transmission state is triggered, and the cloud's transfer completion receipt is monitored; If the transfer completion receipt is received within a consecutive preset acquisition window, then all transfer-state static penetration data in the edge ring buffer is cleared. If the transfer completion receipt is not received within the preset acquisition window, a breakpoint resume instruction is triggered, and an attempt is made to access the cloud to obtain the breakpoint location code of the transmitted static penetration data, and re-upload according to the breakpoint location code; If the breakpoint location code is not obtained within the preset timeout constraint, the data acquisition task is terminated, and a network anomaly warning is executed accordingly.

[0074] In this embodiment, the original static probe data in the edge ring buffer is first replaced iteratively with the transport-state static probe data. The edge ring buffer is a specific type of buffer configured at the edge, with a ring-shaped data structure and a fixed size. When new data is written, it overwrites the oldest data, forming a dynamically scrolling data window. The original static probe data is the raw data without any packing processing. Replacement refers to removing the copy of the original static probe data stored in the ring buffer and storing the newly generated transport-state static probe data in the same buffer location.

[0075] Specifically, once the raw static cone penetration test (SPT) data passes the local quality control model and is deemed anomaly-free, it is processed into transport-state SPT data according to the packaging strategy. While preparing to send this transport-state data, a copy of the original SPT data corresponding to this data is located in the edge ring buffer, and then replaced with the newly generated transport-state SPT data. This replacement is performed every time a batch of data to be transmitted is generated, ensuring that the ring buffer always contains the latest version of the transport-state data.

[0076] Secondly, after each replacement is completed, the transmission state static cone penetration data is sent, and the cloud listens for the transfer completion receipt. Sending refers to the network communication process of sending the prepared transmission state static cone penetration data to the cloud server through the physical network interface. Listening refers to the edge program continuously monitoring the network port or message queue in the background, waiting for and capturing specific feedback information from the cloud. The receipt refers to the cloud server sending an acknowledgment message to the edge after successfully receiving, decrypting, and temporarily storing the transmission state data in its buffer.

[0077] Once the circular buffer completes a data replacement, it immediately triggers a network transmission action, pushing the data to the cloud. Simultaneously, the edge device starts a listening process, waiting for feedback from the cloud regarding specific data packets. The feedback received is a data transfer completion receipt, which is an acknowledgment signal informing the edge device that the data has been received and properly stored.

[0078] If a transfer completion receipt is received within a continuously preset acquisition window, all transmission-state static probe data in the edge ring buffer is cleared. Clearing refers to deleting all transmission-state static probe data stored in the edge ring buffer, freeing up buffer space. Specifically, the edge device dynamically evaluates transmission stability during continuous sending and listening. A continuously preset acquisition window is set; if a corresponding transfer completion receipt is received from the cloud, it indicates a stable network connection and normal cloud service. Therefore, a clearing operation is performed, deleting all transmission-state data from the ring buffer.

[0079] If the transfer completion receipt is not received within the preset acquisition window, a breakpoint resume instruction is triggered. The system attempts to access the cloud to obtain the breakpoint location code of the transmitted static penetration data and re-uploads it based on the breakpoint location code. Here, "not received" means that after sending data within the preset acquisition window, no corresponding cloud receipt was received; the breakpoint resume instruction refers to the process command initiated by the edge device to resume uploading after detecting a transmission failure or timeout; the breakpoint location code is an identifier provided by the cloud to the edge device, indicating the exact location where the cloud successfully received the last data packet; and re-uploading means that the edge device retrieves the data after the breakpoint location code from its local circular buffer and re-executes the transmission action.

[0080] Specifically, when the edge device detects that a data packet has been sent without receiving a confirmation within a preset time or quantity window, it determines that the transmission may have failed. It then triggers a resume command, first attempting to access the cloud to obtain the breakpoint location code. Upon receiving the query, the cloud returns the location of the latest successful data in its record. Once the edge device obtains this code, it immediately locates it in the circular buffer, finds all transmission status data stored after that code, and then, based on the breakpoint location code, re-uploads only this unconfirmed data.

[0081] If the breakpoint location code is not obtained within the preset timeout constraint, the data acquisition task is terminated, and a network anomaly warning is issued accordingly. The preset timeout constraint refers to a maximum waiting time specifically set for attempting to access the cloud to obtain the breakpoint location code; "no acquisition" means that within the timeout constraint period, the breakpoint resume query request sent by the edge device does not receive any response from the cloud; "termination of data acquisition task" means that when the edge device determines that a serious network or cloud failure has occurred, it actively stops the penetration operation of the static probe device, and the network anomaly warning means that a high-level alarm notification is issued to the operator interface.

[0082] Specifically, after triggering the breakpoint resume command, if the edge device contacts the cloud to obtain the breakpoint location and receives no response within the preset timeout constraint, it indicates that the network connection may have been completely interrupted. Therefore, the highest level of fault response is executed: first, the current data acquisition task is terminated, stopping the device from operating. Simultaneously, a clear network anomaly warning is issued to on-site operators, informing them that communication has been completely interrupted and immediate manual intervention is required.

[0083] For example, after triggering the resume command, the edge device begins waiting for a response from the cloud. If no response is received after a preset timeout of 90 seconds, the system determines that the cloud connection has been lost. It immediately sends a stop command to the static cone penetrometer, terminating the data acquisition task. An alarm is also triggered.

[0084] In the method provided in this application embodiment, a local storage area is configured at the edge, and the local storage area is used for: According to the preset differentiated backup rules, rolling backups are performed on the static cone penetration data in the transmission state and the static cone penetration data in the abnormal state, respectively. The differentiated backup rules include a first rolling cycle and a second rolling cycle. When the static probe data in the transmission state exceeds the first rolling cycle, the static probe data in the transmission state is cleared. When the abnormal static cone penetration data exceeds the second rolling cycle, the preset packaging strategy packages the abnormal static cone penetration data and sends the packaging result to the cloud to update the local quality control model.

[0085] In this embodiment, rolling backups are first performed on the static probe data in the transmission state and the static probe data in the abnormal state according to preset differentiated backup rules. The differentiated backup rules include a first rolling period and a second rolling period. The preset differentiated backup rules refer to rules that pre-set different backup strategies; rolling backup refers to overwriting old backups with new backups within a fixed storage space, according to time or sequence; the first rolling period and the second rolling period are two different time length parameters. The first rolling period is suitable for transmission state data and is usually shorter; the second rolling period is suitable for abnormal state data and is usually longer.

[0086] Specifically, static cone penetration test data in transit is managed according to the first rolling cycle; static cone penetration test data in abnormal state is managed according to the second rolling cycle; at the same time, the storage area is scanned periodically, and based on the type and generation time of each data item, it is determined whether it has exceeded the corresponding rolling cycle, and subsequent cleanup or packaging actions are triggered.

[0087] Secondly, when the static probe data in the transfer state exceeds the first rolling cycle, the static probe data in the transfer state is cleared. Here, "exceeding" means that the storage life of the data file has exceeded the rolling cycle threshold set for its type; "clearing" means directly deleting the specified data file from the file system in the local storage area.

[0088] Specifically, the system periodically checks all data files marked as being in transit in the local storage area. For each file, it calculates the number of days since its creation or last modification. If the number exceeds the first rolling cycle set by the data set, it is determined that the data has been successfully uploaded to the cloud and has never been accessed due to interrupted resume. Subsequently, an automatic cleanup operation is performed, deleting the expired transit-state data files from the disk.

[0089] Finally, when abnormal static penetration test data exceeds the second rolling cycle, a preset packaging strategy packages the abnormal static penetration test data and sends the packaging results to the cloud for feedback updates to the local quality control model. Specifically, when an abnormal data file is found to have exceeded its designated second rolling cycle, firstly, the expired abnormal data file is reprocessed using a preset packaging strategy to ensure format consistency and security; then, the packaging results are sent via the network to a dedicated feedback data receiving area in the cloud; finally, after receiving the abnormal case packages from each edge terminal in the cloud, they are centrally analyzed, classified, and labeled. Historical static penetration test anomaly datasets are added to train the lightweight neural network channel. Subsequently, based on a richer training set, the local quality control model is updated to improve the anomaly detection capability of the entire system network.

[0090] In this embodiment, a continuity rule set is introduced to check the physical continuity of adjacent points and the continuity of the experimental process, ensuring the integrity of the data entering the database. A temporary storage-monitoring-batch transfer strategy is adopted, where data that passes the test is temporarily stored in a high-speed cache, and then written to the database in batches after accumulating to a preset collection window size, reducing the storage pressure on the database. Structured storage makes data fields clear and formatted uniformly, facilitating direct querying, analysis, and visualization by various professional software.

[0091] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: In this embodiment, preset collection parameters are first distributed and associated with storage at the edge, ensuring the standardization and consistency of data collection from the source and providing an immutable original basis for subsequent quality judgment. By establishing a circular cache copy, local data recovery capability is provided for network fluctuations or interruptions during data transmission, which is the foundation for reliable transmission and avoids the loss of abnormal situation information. By constructing a local quality control model and integrating rule channels and lightweight neural network channels, the reliability of rule judgment is improved, reducing cloud processing pressure and review burden.

[0092] Secondly, different processing paths are executed based on the edge anomaly warning results. For data with an anomaly warning, the task is immediately terminated and all diagnostic information is packaged and saved, achieving rapid isolation of the problem and preservation of on-site evidence. Data with no anomalies warnings is allowed to enter the cloud channel. Data compression and encryption are performed through a preset packaging strategy. This reduces the amount of data transmitted over the network, lowers the dependence on bandwidth and traffic costs for field operations, ensures that data can safely and efficiently reach the cloud, and provides a structured guarantee for reliable operation.

[0093] Furthermore, a deep anomaly identification verification method is constructed through knowledge verification and pattern verification. This involves multi-dimensional comparison by invoking a regional static probing knowledge base strongly correlated with the task area. The method detects the presence of known fault characteristics from the data curve shape itself, achieving deep verification and ensuring detection accuracy. If verification fails, the cloud rejects data entry, proactively clears the data, terminates the remote task, and sends a detailed anomaly report to the edge, forcibly interrupting the source of the problematic data. Simultaneously, the data anomaly report feeds back the cloud's professional judgment to staff for targeted investigation.

[0094] Finally, a continuity rule set was introduced to check the physical continuity of adjacent points and the continuity of the experimental process, ensuring the integrity of the data entering the database. A temporary storage-monitoring-batch transfer strategy was adopted, where data that passed the test was temporarily stored in a high-speed cache and then written to the database in batches after accumulating to a preset collection window size, reducing the storage pressure on the database. Structured storage makes data fields clear and formatted uniformly, facilitating direct querying, analysis, and visualization by various professional software.

[0095] Example 2, as Figure 2 As shown, based on the same inventive concept as the static cone penetration test data acquisition quality control method with fusion anomaly early warning provided in Embodiment 1, this embodiment of the invention also provides a static cone penetration test data acquisition quality control system with fusion anomaly early warning, including: The anomaly warning module 11 is used to collect raw static cone penetration data at the edge through the associated static cone penetration equipment, and to provide edge anomaly warnings through the embedded local quality control model. The data transmission processing module 12 is used to perform pre-transmission processing on the original static penetration data based on the edge anomaly warning result and in combination with a preset packaging strategy, to obtain the static penetration data in the transmission state and transmit it to the cloud. The data depth identification module 13 is used to perform depth anomaly identification on the transmitted static penetration data in the cloud, wherein the depth anomaly identification includes knowledge verification and pattern verification. The data storage module 14 is used to perform structured storage of the transmission-state static penetration data by combining the depth anomaly identification results with a preset continuity rule set.

[0096] In one embodiment, the anomaly warning module 11 is used for: The edge device sends preset acquisition parameters to the static cone penetration test device to trigger continuous sensing acquisition, and associates and stores the acquisition parameters with the acquired sensing data to obtain the original static cone penetration test data; A copy of the original static penetration data is created and cached in a buffer at the edge, wherein the buffer is configured in a ring cache mode; The raw static penetration data is input into the local quality control model to perform edge anomaly warning, wherein the local quality control model includes parallel rule channels and lightweight neural network channels; If the outputs of both the rule channel and the lightweight neural network channel are normal, then the output edge anomaly warning result is normal.

[0097] The construction steps of the local quality control model include: Obtain the task description for the static penetration test, analyze the task description to extract control clauses, and perform regularization processing on the control clauses to obtain the rule channels. Obtain a historical static penetration anomaly dataset, and perform binarization simplification on the anomaly discrimination results in the historical static penetration anomaly dataset to obtain a sample static penetration anomaly dataset. Based on the sample static probing anomaly dataset, a lightweight neural network model is constructed and trained. The trained lightweight neural network model is output as the lightweight neural network channel and integrated with the rule channel to obtain the local quality control model.

[0098] In one embodiment, the data transmission processing module 12 is used for: If the edge anomaly warning result is abnormal, the static penetration test task is terminated, and the edge anomaly warning result, the original static penetration test data and the corresponding context data are packaged and stored as abnormal static penetration test data. If the edge anomaly warning result is no anomaly, the original static cone penetration data is processed before transmission based on a preset packaging strategy, including data compression and data encryption, to obtain the transmitted static cone penetration data. The abnormal static penetration data is stored in the local storage area at the edge, and the transmission static penetration data is transmitted to the cloud through a dedicated encrypted channel.

[0099] In one embodiment, the data depth recognition module 13 is used for: Based on the mission objective zoning information of the static cone penetration test, a regional static cone penetration knowledge base is extracted; The transport-state static penetration data is subjected to multi-dimensional knowledge verification using the regional static penetration knowledge base. If each dimension satisfies the regional static penetration knowledge base, the knowledge verification is output as passed. Based on the regional static cone penetration test knowledge base, obtain multiple data content anomaly patterns corresponding to the target area, and traverse multiple data content anomaly patterns to perform pattern matching on the transmitted static cone penetration test data. If the pattern matching result is empty, output that the pattern verification is successful. If both knowledge verification and pattern verification pass, the output deep anomaly identification result is "no anomaly". The regional static cone penetration test knowledge base includes at least an empirical relationship map of soil properties and penetration parameters.

[0100] In one embodiment, the data depth recognition module 13 is further configured to: If at least one of the knowledge verification and pattern verification fails, the deep anomaly identification result will be output as an anomaly. The cloud clears the transmitted static penetration data, terminates the static penetration task, and sends a data anomaly receipt to the edge, marking the transmitted static penetration data as abnormal static penetration data.

[0101] In one embodiment, the data storage module 14 is used for: When the depth anomaly identification result is no anomaly, the static penetration data in the transmission state is temporarily stored in the cloud cache. Based on the aforementioned continuous rule set, the acquisition and temporary storage of the static penetration data in the transmission state are performed iteratively. The data length in the cloud cache is continuously monitored until the data length meets the preset collection window. Then, all the data in the cloud cache is transferred to the static cone penetration database for structured and versioned storage. The continuity rule set includes physical continuity rules for adjacent points and continuity rules for the experimental process.

[0102] This includes acquiring and transmitting static penetration data in the transmission state to the cloud, as well as: Iteratively replace the original static penetration data in the edge ring buffer with the transmitted static penetration data; After each replacement is completed, the transmission of the static penetration data in the transmission state is triggered, and the cloud's transfer completion receipt is monitored; If the transfer completion receipt is received within a consecutive preset acquisition window, then all transfer-state static penetration data in the edge ring buffer is cleared. If the transfer completion receipt is not received within the preset acquisition window, a breakpoint resume instruction is triggered, and an attempt is made to access the cloud to obtain the breakpoint location code of the transmitted static penetration data, and re-upload according to the breakpoint location code; If the breakpoint location code is not obtained within the preset timeout constraint, the data acquisition task is terminated, and a network anomaly warning is executed accordingly.

[0103] The edge terminal is equipped with a local storage area, which is used for: According to the preset differentiated backup rules, rolling backups are performed on the static cone penetration data in the transmission state and the static cone penetration data in the abnormal state, respectively. The differentiated backup rules include a first rolling cycle and a second rolling cycle. When the static probe data in the transmission state exceeds the first rolling cycle, the static probe data in the transmission state is cleared. When the abnormal static cone penetration data exceeds the second rolling cycle, the preset packaging strategy packages the abnormal static cone penetration data and sends the packaging result to the cloud to update the local quality control model.

[0104] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: In this embodiment, the anomaly warning module 11 first sends preset collection parameters to the edge and stores them, ensuring the standardization and consistency of data collection from the source and providing an immutable original basis for subsequent quality judgment. By establishing a circular cache copy, local data recovery capability is provided for network fluctuations or interruptions during data transmission, which is the foundation for reliable transmission and avoids the loss of anomaly information. By constructing a local quality control model and integrating rule channels and lightweight neural network channels, the reliability of rule judgment is improved, and the processing pressure and review burden on the cloud are reduced.

[0105] Secondly, the data transmission processing module 12 executes different processing paths based on the edge anomaly warning results. For data with an anomaly warning, the task is immediately terminated and all diagnostic information is packaged and saved, achieving rapid isolation of the problem and preservation of on-site evidence. For data with no anomalies warning, it is allowed to enter the cloud channel. Data compression and encryption are performed through a preset packaging strategy. This reduces the amount of data transmitted over the network, lowers the dependence on bandwidth and traffic costs for field operations, ensures that data can reach the cloud safely and efficiently, and provides a structured guarantee for reliable operation.

[0106] Secondly, through the data deep identification module 13, knowledge verification and pattern verification are performed to construct a deep anomaly identification verification method. This involves multi-dimensional comparison by calling a regional static probing knowledge base strongly correlated with the task area, detecting the presence of known fault characteristics from the data curve shape itself, thus achieving deep verification and ensuring detection accuracy. Rigid handling measures are implemented when deep identification detects anomalies. If verification fails, the cloud refuses to accept the data into the database, proactively clears the data, terminates the remote task, and sends a detailed anomaly report to the edge, forcibly interrupting the source of the problematic data. Simultaneously, the data anomaly report feeds back the cloud's professional judgment to staff for targeted investigation.

[0107] Finally, through data storage module 14, a continuity rule set is introduced to check the physical continuity of adjacent points and the continuity of the experimental process, ensuring the integrity of the data entering the database. A temporary storage-monitoring-batch transfer strategy is adopted, temporarily storing the verified data in a high-speed cache until it accumulates to a preset acquisition window size before being batch-written into the database, reducing the database's storage pressure. Structured storage makes data fields clear and formatted uniformly, facilitating direct querying, analysis, and visualization by various professional software.

[0108] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0109] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0110] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for quality control of static cone penetration test data acquisition that integrates anomaly early warning, characterized in that, include: The edge end collects raw static cone penetration data through associated static cone penetration equipment and uses an embedded local quality control model to provide early warning of edge anomalies. Based on the edge anomaly warning results, the original static penetration data is preprocessed for transmission using a preset packaging strategy to obtain the static penetration data in the transmission state and transmit it to the cloud. The cloud performs deep anomaly identification on the transmitted static penetration data, wherein the deep anomaly identification includes knowledge verification and pattern verification; The transmission-state static penetration data is structured and stored by combining the depth anomaly identification results with a preset set of continuity rules.

2. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, The edge end collects raw static cone penetration data through associated static cone penetration testing equipment and uses an embedded local quality control model to provide early warnings of edge anomalies, including: The edge device sends preset acquisition parameters to the static cone penetration test device to trigger continuous sensing acquisition, and associates and stores the acquisition parameters with the acquired sensing data to obtain the original static cone penetration test data; A copy of the original static penetration data is created and cached in a buffer at the edge, wherein the buffer is configured in a ring cache mode; The raw static penetration data is input into the local quality control model to perform edge anomaly warning, wherein the local quality control model includes parallel rule channels and lightweight neural network channels; If the outputs of both the rule channel and the lightweight neural network channel are normal, then the output edge anomaly warning result is normal.

3. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 2, characterized in that, The steps for constructing the local quality control model include: Obtain the task description for the static penetration test, analyze the task description to extract control clauses, and perform regularization processing on the control clauses to obtain the rule channels. Obtain a historical static penetration anomaly dataset, and perform binarization simplification on the anomaly discrimination results in the historical static penetration anomaly dataset to obtain a sample static penetration anomaly dataset. Based on the sample static probing anomaly dataset, a lightweight neural network model is constructed and trained. The trained lightweight neural network model is output as the lightweight neural network channel and integrated with the rule channel to obtain the local quality control model.

4. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, Based on the edge anomaly warning results, and combined with a preset packaging strategy, the original static penetration test data is preprocessed for transmission to obtain the transmission-state static penetration test data and transmit it to the cloud, including: If the edge anomaly warning result is abnormal, the static penetration test task is terminated, and the edge anomaly warning result, the original static penetration test data and the corresponding context data are packaged and stored as abnormal static penetration test data. If the edge anomaly warning result is no anomaly, the original static cone penetration data is processed before transmission based on a preset packaging strategy, including data compression and data encryption, to obtain the transmitted static cone penetration data. The abnormal static penetration data is stored in the local storage area at the edge, and the transmission static penetration data is transmitted to the cloud through a dedicated encrypted channel.

5. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, The cloud performs deep anomaly identification on the transmitted static penetration data, including: Based on the mission objective zoning information of the static cone penetration test, a regional static cone penetration knowledge base is extracted; The transport-state static penetration data is subjected to multi-dimensional knowledge verification using the regional static penetration knowledge base. If each dimension satisfies the regional static penetration knowledge base, the knowledge verification is output as passed. Based on the regional static cone penetration test knowledge base, obtain multiple data content anomaly patterns corresponding to the target area, and traverse multiple data content anomaly patterns to perform pattern matching on the transmitted static cone penetration test data. If the pattern matching result is empty, output that the pattern verification is successful. If both knowledge verification and pattern verification pass, the output deep anomaly identification result is "no anomaly". The regional static cone penetration test knowledge base includes at least an empirical relationship map of soil properties and penetration parameters.

6. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 5, characterized in that, Also includes: If at least one of the knowledge verification and pattern verification fails, the deep anomaly identification result will be output as an anomaly. The cloud clears the transmitted static penetration data, terminates the static penetration task, and sends a data anomaly receipt to the edge, marking the transmitted static penetration data as abnormal static penetration data.

7. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, Combining the depth anomaly identification results with a preset continuity rule set, the transmitted static penetration data is structured and stored, including: When the depth anomaly identification result is no anomaly, the static penetration data in the transmission state is temporarily stored in the cloud cache. Based on the aforementioned continuous rule set, the acquisition and temporary storage of the static penetration data in the transmission state are performed iteratively. The data length in the cloud cache is continuously monitored until the data length meets the preset collection window. Then, all the data in the cloud cache is transferred to the static cone penetration database for structured and versioned storage. The continuity rule set includes physical continuity rules for adjacent points and continuity rules for the experimental process.

8. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, Acquiring and transmitting static penetration data in transit state to the cloud also includes: Iteratively replace the original static penetration data in the edge ring buffer with the transmitted static penetration data; After each replacement is completed, the transmission of the static penetration data in the transmission state is triggered, and the cloud's transfer completion receipt is monitored; If the transfer completion receipt is received within a consecutive preset acquisition window, then all transfer-state static penetration data in the edge ring buffer is cleared. If the transfer completion receipt is not received within the preset acquisition window, a breakpoint resume instruction is triggered, and an attempt is made to access the cloud to obtain the breakpoint location code of the transmitted static penetration data, and re-upload according to the breakpoint location code; If the breakpoint location code is not obtained within the preset timeout constraint, the data acquisition task is terminated, and a network anomaly warning is executed accordingly.

9. The static cone penetration test data acquisition quality control method with integrated anomaly early warning as described in claim 1, characterized in that, The edge is equipped with a local storage area, which is used for: According to the preset differentiated backup rules, rolling backups are performed on the static cone penetration data in the transmission state and the static cone penetration data in the abnormal state, respectively. The differentiated backup rules include a first rolling cycle and a second rolling cycle. When the static probe data in the transmission state exceeds the first rolling cycle, the static probe data in the transmission state is cleared. When the abnormal static cone penetration data exceeds the second rolling cycle, the preset packaging strategy packages the abnormal static cone penetration data and sends the packaging result to the cloud to update the local quality control model.

10. A static cone penetration test data acquisition quality control system integrating anomaly early warning, characterized in that, The system is used to implement the static cone penetration test data acquisition quality control method with fusion anomaly early warning as described in any one of claims 1-9, the system comprising: The anomaly warning module is used to collect raw static cone penetration data at the edge via associated static cone penetration equipment and to provide edge anomaly warnings through an embedded local quality control model. The data transmission processing module is used to perform pre-transmission processing on the original static cone penetration data based on the edge anomaly warning result and in combination with a preset packaging strategy, to obtain the static cone penetration data in the transmission state and transmit it to the cloud. A data depth identification module is used to perform depth anomaly identification on the transmitted static penetration data in the cloud, wherein the depth anomaly identification includes knowledge verification and pattern verification. The data storage module is used to perform structured storage of the transmitted static penetration data by combining the deep anomaly identification results with a preset set of continuity rules.

Images