Construction engineering quality witness sampling detection supervision system

Abstract

This invention discloses a construction engineering quality witnessing sampling and testing supervision system, belonging to the field of construction engineering quality testing informatization. The system includes: a distributed sampling terminal deployed at sampling points on the construction site, generating and attaching an RFID-QR code fusion tag when the specimen is formed; a blockchain evidence storage module connected to the terminal, writing sample birth information, sampler and witness facial features, and GNSS coordinates into a consortium blockchain, returning a globally unique TxID; an edge computing node module, completing sampling data verification, running a federated learning model to calculate the sample risk index, updating the judgment threshold, handling anomalies, and logging to the blockchain; a testing node module, reading the tag EPC and comparing it with the TxID to verify the sample's legality; and a supervision node module, triggering a smart contract to issue an electronic stop-work order when the risk index exceeds the threshold, locking the corresponding test report number. This invention achieves end-to-end data collaboration between sampling, testing, and supervision, ensuring data traceability and tamper-proofing, and meeting regulatory standards. It solves the problems of difficult traceability, easy tampering, and insufficient regulatory collaboration in construction engineering sampling and testing data.

Description

Technical Field

[0001] This invention relates to the field of information technology for quality inspection in building construction, and in particular to a monitoring system for witness sampling and inspection of quality in building construction. Background Technology

[0002] In the quality control of construction projects, witnessed sampling and testing is a crucial link in ensuring project quality, and the accuracy of its data directly affects the structural safety and performance of the project. Current witnessed sampling supervision mostly adopts a "QR code + photo + manual record" model, which has significant shortcomings in practical application:

[0003] QR codes are easily copied or reprinted. Criminals can replace QR codes to switch samples, which makes the test data unable to accurately reflect the quality of the project and poses a hidden danger to project safety. The test data is stored in a centralized database, which poses a single point of tampering risk. Some companies tamper with the test data through illegal means in order to avoid quality issues, thereby undermining the authenticity and seriousness of the test data. The problem of data silos among provinces and cities is serious. AI model training relies on a large amount of cross-regional data, but due to data privacy protection policies, raw data cannot be shared across regions. This results in poor generalization ability and low accuracy of AI risk identification models, making it difficult to meet the needs of cross-regional collaborative supervision.

[0004] Therefore, there is an urgent need for a construction engineering quality witnessing, sampling, testing and supervision system that can be cross-domain collaborative, tamper-proof and self-evolving in a zero-trust environment, in order to solve the pain points of existing technologies and improve the level of supervision. Summary of the Invention

[0005] The purpose of this invention is to propose a construction engineering quality witnessing sampling and testing supervision system, which aims to solve the problems of difficulty in tracing, easy tampering, and insufficient supervision and coordination of existing construction engineering sampling and testing data.

[0006] This invention is implemented as follows: a construction engineering quality witnessing sampling and testing supervision system, the system comprising: Distributed sampling terminals are deployed at various sampling points on construction sites to generate and attach RFID-QR code fusion tags the moment the specimen is formed. The edge computing node module is used to complete the verification of sampled data; run federated learning models to calculate sample risk index; aggregate multi-region model parameters and dynamically update sample risk judgment thresholds; adapt data formats and encrypt data transfer; process abnormal data and log it on the blockchain to provide data and technical support for regulatory decisions. The blockchain evidence storage module is connected to the distributed sampling terminal via a network. It is used to write the sample's birth information, the facial features of the sampler and witness, and GNSS coordinates into the consortium blockchain and return a globally unique TxID. The detection node module is used to receive sample association data pushed by the distributed sampling terminal. During the sample receiving process, it reads the EPC information of the RFID-QR code fusion tag and compares it with the TxID in the blockchain evidence storage module to verify the legality of the sample. The regulatory node module is used to automatically trigger the smart contract and issue an electronic stop-work order when the risk index is higher than the sample risk judgment threshold; at the same time, it locks the status position of the corresponding test report number, prohibiting the issuance of reports with that number.

[0007] Beneficial effects of the present invention This invention discloses a construction engineering quality witnessing sampling and testing supervision system. The system includes: a distributed sampling terminal deployed at sampling points on the construction site, which generates and affixes an RFID-QR code fusion tag when the specimen is formed; a blockchain evidence storage module connected to the terminal, which writes the sample's origin information, the facial features of the sampler and witness, and GNSS coordinates into the consortium blockchain and returns a globally unique TxID; an edge computing node module, which performs sampling data verification, runs a federated learning model to calculate the sample risk index, updates the judgment threshold, handles anomalies, and logs them on the blockchain; a testing node module, which reads the tag EPC and compares it with the TxID to verify the sample's legality; and a supervision node module, which triggers a smart contract to issue an electronic stop-work order when the risk index exceeds the threshold, locking the corresponding test report number. This invention achieves end-to-end data collaboration between sampling, testing, and supervision, ensuring data traceability and tamper-proofing, and meeting regulatory standards. It solves the problems of difficult traceability, easy tampering, and insufficient supervision collaboration in construction engineering sampling and testing data. Attached Figure Description

[0008] Figure 1 This is a structural diagram of a construction engineering quality witnessing sampling and testing supervision system according to a preferred embodiment of the present invention. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. For ease of explanation, only the parts related to the embodiments of this invention are shown. It should be understood that the specific embodiments described herein are merely for explaining this invention and are not intended to limit this invention.

[0010] This invention proposes a construction engineering quality witnessing sampling and testing supervision system. The system includes: a distributed sampling terminal deployed at sampling points on the construction site, which generates and affixes an RFID-QR code fusion tag when the specimen is formed; a blockchain evidence storage module connected to the terminal, which writes the sample's origin information, the facial features of the sampler and witness, and GNSS coordinates into the consortium blockchain, returning a globally unique TxID; an edge computing node module, which performs sampling data verification, runs a federated learning model to calculate the sample risk index, updates the judgment threshold, handles anomalies, and logs them to the blockchain; a testing node module, which reads the tag EPC and compares it with the TxID to verify the sample's legality; and a supervision node module, which triggers a smart contract to issue an electronic stop-work order when the risk index exceeds the threshold, locking the corresponding test report number. This invention achieves end-to-end data collaboration between sampling, testing, and supervision, ensuring data traceability and tamper-proofing, and meeting regulatory standards. It solves the problems of difficult traceability, easy tampering, and insufficient supervision collaboration in construction engineering sampling and testing data.

[0011] Figure 1 This is a structural diagram of a preferred embodiment of a construction engineering quality witnessing sampling and testing supervision system. The system includes a distributed sampling terminal, an edge computing node module, a blockchain evidence storage module, a testing node module, and a supervision node module.

[0012] Distributed sampling terminals are deployed at various sampling points on construction sites to generate and attach radio frequency-QR code fusion tags at the moment the specimen is formed. The specimens are cementitious material specimens that need to undergo a setting and hardening process, such as concrete specimens, masonry mortar specimens, and grouting material specimens, which are suitable for various construction projects such as residential buildings, commercial buildings, and public facilities. Furthermore, the distributed sampling terminal also includes a vibration signal detection module, a tag generation and attachment module, and a three-element acquisition module. The vibration signal detection module is used to collect characteristic vibration signals of building engineering specimens during the molding stage in real time, identify the initial setting point of the specimens, trigger the label generation command, and provide a signal reference for the subsequent sampling process. The tag generation and attachment module receives trigger commands from the vibration signal detection unit, generates RF-Q tag conforming to technical specifications, and attaches the tag to the surface of the specimen with a preset strength to achieve physical binding between the tag and the specimen. The RF-Q tag has a built-in unmodifiable EPC area and an expandable user area, and the attachment strength is ≥ a preset attachment strength threshold (e.g., 50N), which physically prevents the tag from falling off and the specimen from being swapped. The three-element data collection module is used to simultaneously collect the identity information of the sampler and witness, as well as the spatiotemporal information of the sampling scenario, and generate the basic data required for three-element verification to ensure that the identities of the sampling participants are authentic and that the sampling behavior is traceable in time and space. The three-factor verification refers to: 1) comparing the facial feature vectors of the sampler and the witness within a Trusted Execution Environment (TEE), with a similarity ≥ a similarity threshold (e.g., 95%) (the terminal can have a built-in camera and TEE module to extract and temporarily store facial feature vectors within the TEE), ensuring the authenticity of the personnel's identity; 2) the GNSS coordinate difference between the sampling terminal and the witnessing terminal ≤ a coordinate difference threshold (e.g., 2m), with the terminal integrating a GNSS module (positioning accuracy down to the meter level), ensuring that the sampling and witnessing activities are within the same spatial range; 3) the difference between the timestamp attached to the tag and the on-chain timestamp ≤ a timestamp difference threshold (e.g., 30s), ensuring data time consistency; only when all three conditions are met can the data be written to the blockchain.

[0013] Furthermore, the vibration signal detection module also includes a signal acquisition unit, a preliminary coagulation determination unit, and... Instruction triggering unit, The signal acquisition unit is used to collect vibration data related to specimen molding in real time (such as vibration frequency, vibration waveform, amplitude and other parameters at the concrete mixer outlet and during the mold pouring process) through a built-in high-sensitivity vibration sensor (such as a piezoelectric vibration sensor). The sampling frequency is not lower than the sampling frequency threshold (such as 100Hz) to ensure that instantaneous vibration changes are captured. The initial setting determination unit is used to analyze the collected vibration data in real time based on the preset initial setting vibration characteristic threshold of the specimen (for concrete specimens, the determination criterion is "the vibration frequency of the concrete at the mixer outlet changes by more than 50 Hz"). By calculating the vibration frequency difference within a unit time (e.g., 100 ms), when the difference is greater than the preset fluctuation threshold (e.g., 10 Hz, which is determined to be a frequency change) and the final frequency is greater than the preset change critical threshold (e.g., 50 Hz, which is the quantitative boundary for distinguishing between "not yet set" and "entering the initial setting stage"), the specimen is automatically determined to have entered the initial setting stage.

[0014] The instruction triggering unit is used to immediately generate an "RF-QR code fusion tag generation trigger instruction" after confirming the initial solidification of the specimen, and send it to the tag generation and attachment unit (through the terminal's internal data bus); thereby achieving synchronous connection between the sampling process and the initial solidification node of the specimen, avoiding deviations in tag attachment timing caused by delays in manual judgment. Furthermore, the tag generation and locking module also includes a tag parameter generation unit, a physical locking execution unit, and a locking verification unit. The tag parameter generation unit is used to generate dual-module data for RFID-QR code fusion tags according to preset tag technology standards. In some embodiments, the label parameter generation unit further includes an RF unit and a QR code unit; the RF unit is used to write an SM3 hash value (96 bits) of "sample ID + project number" in the EPC area and to write sample birth information (such as specimen type, strength grade, raw material information, etc., 256 bits) encrypted with SM4 in the user area; the operating frequency can be set to 920-925MHz; the QR code unit is used to generate a visual pattern based on the hash data in the EPC area of ​​the RF unit through an XOR operation to ensure that the QR code corresponds one-to-one with the data in the EPC area; in specific implementations, the pattern resolution can reach 600dpi, and it can still be parsed even if 30% damaged. The physical locking execution unit is used to lock the generated RFID-QR code fusion tag (with a laser-coated fragile film) onto a preset position on the surface of the specimen (such as the center area of ​​the side of the mold) through a mechanical locking mechanism (such as an electric snap-on locking component) built into the distributed sampling terminal. In specific implementation, the locking tension is monitored in real time during the locking process to ensure that the locking strength is ≥ the preset locking strength threshold (such as 50N) to prevent the tag from falling off or being maliciously peeled off during transportation and inspection. The tag-attachment verification unit is used to perform a secondary detection of the tag-attachment strength using a tensile sensor after tagging is completed. If the strength meets the standard, a "tag tag-attachment qualified signal" is generated and synchronized to the blockchain evidence storage module (which can be received via a front-end data interface), providing a basis for the tag tag-attachment status for subsequent three-factor verification. If the strength does not meet the standard, an audible and visual alarm is triggered on the terminal, prompting the operator to re-attach the tag. In specific implementations, a tensile sensor (such as a miniature tensile sensor) can be embedded inside the RFID-QR code fusion tag. When the tag is attached, the sensor directly detects the interfacial tensile force between the tag and the surface of the specimen, simultaneously achieving tag-attachment strength monitoring and tag integrity protection to prevent reuse after peeling. The tensile sensor can also be deployed in the mechanical tagging mechanism (such as an electric snap-on tagging assembly) built into the distributed sampling terminal.

[0015] Furthermore, the three-element acquisition module also includes The information acquisition unit is used to acquire facial images of samplers and witnesses through the high-definition camera (resolution not less than 2 million pixels) and local TEE (Trusted Execution Environment) module built into the distributed sampling terminal, and to extract and temporarily store facial feature vectors within the TEE (to avoid leakage of original facial data) to generate facial feature data that can be compared. The spatiotemporal information acquisition unit is used to acquire the current GNSS coordinates of the terminal (i.e., the spatial location of the sampling / witnessing site) through the GNSS positioning module integrated in the distributed sampling terminal (supporting Beidou + GPS dual-mode positioning, with a positioning accuracy of ≤1m); at the same time, it obtains the on-chain standard time of the blockchain evidence storage module through the time synchronization module of the distributed sampling terminal, records the timestamp of the tag locking operation, and ensures that the spatiotemporal information is consistent with the on-chain time reference. The data association and upload unit is used to associate and bind the collected "sampler's facial features + witness's facial features + GNSS coordinates + timestamp of tag attachment operation" with the sample ID of the current specimen to form a "personnel-spatiotemporal-sample" associated data group, which is temporarily stored in the local encrypted storage area of ​​the terminal. After the tag generation and attachment unit confirms that the attachment is qualified, the data group is pushed to the edge computing node module to provide complete data support for the three-element verification (facial similarity comparison, coordinate difference judgment, and timestamp difference judgment).

[0016] The edge computing node module is used to complete the verification of sampled data; run federated learning models to calculate sample risk index; aggregate multi-region model parameters and dynamically update sample risk judgment thresholds; adapt data formats and encrypt data transfer; process abnormal data and log it on the blockchain to provide data and technical support for regulatory decisions. The edge computing node module includes: The data interaction and format adaptation unit is used to connect the distributed sampling terminal, testing agency node, regulatory node, and blockchain evidence storage module bidirectionally via the 5G-SA slicing network. On the one hand, it receives the raw data uploaded by the front end (including the distributed sampling terminal and testing agency node), completes the format conversion, encryption and decryption (adapting to system security specifications); on the other hand, it pushes the business processing results and dynamic rule parameters (including but not limited to verification results, sample risk index, and sample risk judgment threshold) to the regulatory node module (supporting decision-making) and the blockchain evidence storage module (achieving data traceability), ensuring low latency and accuracy of data flow throughout the system. The data verification unit is used to perform three-element judgment locally based on the "personnel facial features + GNSS coordinates + tag attachment timestamp" associated data group processed by the data interaction and format adaptation unit (such as the similarity between the facial features of the sampler and the witness ≥ similarity threshold 95%, the difference between the GNSS coordinates of the sampling terminal and the witness terminal ≤ coordinate difference threshold 2m, and the difference between the tag attachment timestamp and the on-chain timestamp ≤ timestamp difference threshold 30s). If the verification passes, a "data on-chain permission signal" is generated; if it fails, an alarm (such as an audible and visual alarm) is triggered and the violation short video is uploaded to the blockchain evidence storage module. The sample risk prediction unit is used to predict the sample risk index based on the federated learning model. Specifically, it receives the test data hash (such as the SM3 summary of the load-displacement curve) and basic sample information (cement type, water-cement ratio, etc.) uploaded by the testing agency node, and calculates and outputs the sample risk index in real time based on the pre-trained federated learning model. Specifically, the federated learning model adopts a horizontal federation + homomorphic encryption architecture. Testing institutions in various provinces and cities participate by training the model locally, encrypting only the model parameters before uploading them to the federated learning center (i.e., the federated learning management module) to prevent leakage of raw data. The loss function incorporates a domain-adaptive term to reduce the impact of data distribution differences across provinces and cities on model performance (difference reduction ≥30%), thereby improving the model's cross-domain generalization ability. The model dynamically updates the risk assessment threshold with an update cycle of ≤24 hours to ensure the timeliness and accuracy of risk identification.

[0017] The Federated Learning Management Unit receives encrypted local model parameters uploaded by testing agency nodes, aggregates these parameters using horizontal federation and homomorphic encryption, incorporates a domain-adaptive term into the loss function to reduce cross-regional data distribution differences (reducing differences by ≥30%), and generates a globally optimized model (i.e., the optimized federated learning model). Simultaneously, it dynamically updates the sample risk assessment threshold at a preset period (e.g., ≤24h) and synchronizes it to the regulatory node to ensure the timeliness and cross-domain adaptability of risk identification. The local model refers to the sub-model independently trained by each testing agency node based on its own local testing data; its core function is to extract local data features and generate encrypted model parameters for upload. The globally optimized model generated by the Federated Learning Management Unit after aggregating all local model parameters from testing agencies is used to dynamically update the sample risk assessment threshold. The abnormal data processing and logging unit is used to locally store and perform preliminary analysis on received abnormal data (such as incomplete facial features uploaded by the terminal and abnormal hash values ​​of the testing agency), and generate abnormal data logs (including data source, abnormal type, and timestamp); at the same time, it records the operation logs (such as the number of verifications, model calculation time, and parameter aggregation results). The logs are uploaded to the blockchain after being hashed by SM3, providing traceable evidence for regulatory audits and problem investigation. The blockchain evidence storage module is connected to the distributed sampling terminal via a 5G-SA slicing network. It is used to write the sample's birth information, the facial features of the sampler and witness, and GNSS coordinates into the consortium blockchain and return a globally unique TxID, thereby achieving distributed data storage and tamper-proof data. Specifically, the blockchain evidence storage module is built on consortium blockchain technology and connects to the distributed sampling terminal, detection node module, and supervision node module through a 5G-SA slicing network to ensure the real-time performance and security of data transmission. It receives sample birth information (including specimen type, strength grade, raw material information, etc.), facial features of samplers and witnesses, GNSS coordinates, and other data uploaded by the distributed sampling terminal. After passing the three-factor verification, the data is written into the consortium blockchain in one go and a globally unique TxID is generated, realizing full-process traceability and immutability of data. The detection node module receives sample-related data (including sample ID and on-chain TxID) pushed by the distributed sampling terminal. During the sample receiving stage, it reads the EPC information of the RFID-QR code fusion tag and compares it with the TxID in the blockchain evidence storage module to verify the sample's legality. Specifically, during the detection process, the hash calculation module (which can be built into the detection node module) generates SM3 hash digests in real time for raw data such as load-displacement curves and detection results (retaining the raw data locally to protect privacy) and uploads the digests to the blockchain evidence storage module. At the same time, it receives the status instructions of the detection report number issued by the regulatory node (such as prohibiting the issuance of the corresponding report when it is "locked"), and provides raw detection data and supporting materials to meet regulatory review requirements. Ultimately, it realizes data collaboration between the detection stage, the sampling end, and the regulatory end, ensuring that the detection data is traceable, tamper-proof, and compliant with regulatory standards.

[0018] The regulatory end refers to the technical deployment environment corresponding to the hierarchical entities that exercise the function of supervising the quality of construction projects. Specifically, it is the regulatory infrastructure cluster of the housing and construction authorities (including municipal and provincial authorities), including regulatory hardware equipment, system platforms and related network environments, and is the physical and logical location of the edge computing nodes on the regulatory side.

[0019] Furthermore, the hash calculation module adopts the SM3 national cryptographic hash algorithm to calculate the load-displacement curve, detection results and other data in real time during the detection process, and generate a 256-bit hash digest. The first 16 bits of the digest are used as a short code for the detection report for public query. The public can query the detection data on the blockchain platform through the short code, thereby improving the transparency of the detection.

[0020] The regulatory node module is used to receive the sample risk index output by the sample risk prediction unit. When the risk index is higher than the sample risk judgment threshold, the smart contract is automatically triggered to issue an electronic stop-work order. At the same time, the status of the corresponding test report number is set to "locked", prohibiting the testing agency from issuing the report with that number.

[0021] Furthermore, the regulatory node can also manually review the report number in the "locked" state, and unlock it after the review is passed, ensuring regulatory flexibility; the sample risk judgment threshold is dynamically updated by the federated learning model, with an update cycle of ≤24h.

[0022] Furthermore, the smart contract automatically verifies and stores the sampling data, test data hash, and risk index that meet the three-factor standard; when a violation signal is received (such as sampling verification failure or sample risk index exceeding the standard), it automatically triggers the corresponding compliance handling and traceability operations (including but not limited to alarm notification, on-chain locking of violation certificates, and identification of responsible parties); it synchronously updates the risk judgment threshold of the federated learning management module and makes it effective in real time to ensure the uniformity of judgment rules across the entire system; at the same time, it records key nodes in the entire data flow chain (such as data on-chain time and verification pass mark), providing tamper-proof technical evidence for subsequent regulatory traceability and responsibility determination, and realizing a regulatory closed loop.

[0023] The electronic work stoppage order is automatically pushed to the digital wallets of the general contractor, the supervision unit, and the housing and construction authority via smart contracts; and the receipt status is recorded on the blockchain. If it is not signed for within a preset period (e.g., 24 hours), it will be automatically upgraded to the provincial supervision node to ensure that the work stoppage order is effectively communicated and supervised. The status bits of the test report number include: normal, locked, and retest; when the status bit is "locked", no testing agency node can issue a report with the corresponding number until it is unlocked after manual review by the regulatory node, forming a closed loop for the whole process of test report supervision.

[0024] The following example, using a provincial highway bridge engineering concrete test block quality supervision implementation case, details the implementation process of the construction engineering quality witness sampling and testing supervision system of the present invention.

[0025] Taking the sampling and supervision of concrete bridge column test blocks in the "G107 Expressway Reconstruction and Expansion Project (K120-K150 section)" in a certain province as an example, the complete collaborative operation process of each module is presented as follows: Distributed sampling terminals perform sampling and data acquisition (construction site stage): At 10:00 AM on August 10, 2025, a sampler from the construction unit used a distributed sampling terminal (a handheld smart terminal with a built-in RFID tag printer, face camera, and GNSS positioning module) deployed at the concrete pouring site of the bridge's No. 3 column to perform operations. The moment the C50 concrete test block (150mm×150mm×150mm) is poured and formed, the terminal automatically generates an RFID-QR code fusion label (EPC code: G107-20250810-3LZ-001, the QR code contains the project ID, test block number, and strength grade). The label is then physically attached to the surface of the test block using the terminal's built-in label locking device (anti-tampering buckle design, removing the label will damage it). The terminal synchronously collects the facial features of the sampler (128-dimensional feature vector, encrypted storage) and the facial features of the witness (personnel from the supervision unit, pre-verified in real time with the provincial transportation supervision bureau's filing database), and obtains the current GNSS coordinates (WGS84 coordinate system: 116°32′45.78″E, 38°15′22.31″N, positioning accuracy ±0.5m), and records the tag attachment timestamp (2025-08-10 10:00:18.352). The terminal packages the "test block birth information (molding time, strength grade, component number) + facial features + GNSS coordinates + tag EPC code" into a raw data group, and pushes it synchronously to the edge computing node module and the blockchain evidence storage module through the 5G-SA slicing network (dedicated engineering supervision frequency band).

[0026] The blockchain evidence storage module completes data uploading and TxID generation, and returns the following: After receiving the raw data pushed by the terminal, the blockchain evidence storage module deployed by the Provincial Transportation Supervision Bureau (the consortium blockchain nodes include four types of authorized nodes: the Provincial Supervision Bureau, municipal testing institutions, construction parties, and supervision parties) then: The system automatically uses SM4 symmetric encryption on facial features (the key is jointly managed by the consortium blockchain nodes), performs SM3 hashing on the birth information and GNSS coordinates of the test block, and generates an immutable evidence storage data packet. The encrypted evidence data is written into the consortium blockchain distributed ledger (each authorized node records the data synchronously). After the writing is completed, a globally unique TxID (transaction identifier) ​​is returned: Tx20250810G1073LZ0010001 (including timestamp and node signature information). This TxID is then sent back to the distributed sampling terminal (stored locally on the terminal for use in subsequent testing).

[0027] Edge computing node modules perform data processing and risk analysis: After receiving raw data from the terminal, the edge computing node modules (equipped with GPU computing cards and supporting federated learning inference) deployed in the city-level edge data centers along the project route process the data as follows: Data verification: The built-in verification algorithm is called to verify the similarity between the sampler's facial features and the filing database (calculated to be 98.2% ≥ 95% threshold), the compliance of the witness's facial features (verification passed), the deviation between the GNSS coordinates and the project filing sampling area (1.2m ≤ 2m threshold), and the difference between the tag attachment timestamp and the test block forming time (5s ≤ 30s threshold), and to determine that "the sampling data is compliant"; Format adaptation and encrypted transfer: The standardized data that has passed the verification (coordinates are uniformly in WGS84 format and timestamps are converted to UTC standard) is processed with homomorphic encryption and pushed to the provincial regulatory bureau's regulatory node module, while retaining an encrypted copy for subsequent traceability; Federated learning model operation and risk index calculation: Load the previously aggregated "provincial testing agency federated learning model" (this model was generated by the provincial transportation supervision bureau's federated learning management unit by aggregating local model parameters from 13 municipal testing agencies, and the loss function includes a domain adaptive term, which has been adapted to the differences in aggregate gradation in different regions), input the basic information of the test block (strength grade C50, pouring environment temperature 28℃) and sampling compliance data, and calculate the quality risk index of the test block as 72 points (out of 100 points, below 80 points is low risk). Model parameter aggregation and threshold update: At 18:00 on the same day, the edge computing node module receives local model parameters (trained based on the test block test data of each institution on the same day) uploaded by the encrypted data of the three surrounding municipal testing institutions. The parameters are aggregated through horizontal federated learning technology, and the "sample risk judgment threshold" is dynamically updated from the original 80 points to 82 points (due to the recent fluctuation of aggregate quality in a certain area, it is necessary to improve the sensitivity of risk warning). The updated threshold is synchronized to the regulatory node module and each testing node module. Abnormal data handling: At 10:15 on the same day, the GNSS coordinate deviation of the test block uploaded by another sampling point terminal reached 3.5m (exceeding the 2m threshold). The edge computing node module determined that the "sampling data was abnormal", automatically recorded the abnormal log (including terminal device ID, abnormal type, and timestamp), encrypted it and wrote it into the consortium blockchain evidence storage module, and pushed the abnormal warning to the supervision node module at the same time.

[0028] The detection node module performs sample legality verification (testing institution stage): At 9:00 AM on August 12th, the construction unit delivered the test block for column No. 3 to the municipal-level testing agency (a third-party testing agency authorized by the Provincial Transportation Supervision Bureau, which deploys a dedicated testing terminal): The testing personnel used the radio frequency reader of the testing node module to read the EPC information (G107-20250810-3LZ-001) of the test block label, and the terminal automatically extracted the TxID (Tx20250810G1073LZ0010001) embedded in the EPC. The detection node module queries the evidence storage data corresponding to the TxID in the blockchain evidence storage module through the consortium link port, compares the consistency of "Label EPC code, test block number, project ID" with the evidence storage information, and confirms that "the binding relationship between EPC and TxID is legal and the source of the test block is traceable"; If the comparison results are consistent (no tampering or forgery), the detection node module records the "sample is legal" status and allows it to enter the compressive strength testing process; if "EPC and TxID do not match" (e.g., the label has been replaced), the sample is automatically rejected and a "suspected sample forgery" alarm is pushed to the regulatory node module.

[0029] Regulatory node module implements risk control and reporting lockout (regulatory decision-making stage): On August 15th, the testing agency completed the compressive strength test of the No. 3 column test block (measured strength 52.3 MPa, meeting the C50 standard). However, another batch of test blocks from the No. 4 abutment had an "abnormal aggregate gradation data," and the edge computing node module calculated its risk index to be 85 points (exceeding the updated threshold of 82 points). Automatic triggering of smart contracts and electronic work stoppage orders: After receiving the high-risk signal, the regulatory node module immediately triggers the "risk disposal smart contract" on the consortium blockchain, automatically generating an electronic work stoppage order (number TG20250815001), which includes the scope of work stoppage (subsequent concrete pouring of bridge abutment No. 4), rectification requirements (re-verification of aggregate supplier qualifications), and resumption verification process. After the electronic work stoppage order is stored on the blockchain, it is simultaneously pushed to the construction unit, supervision unit, and local traffic law enforcement team. Test report number locked: The regulatory node module queries the test report number (JCB20250815003) corresponding to the high-risk test block and sets its status to "locked" in the system, prohibiting the testing agency from issuing test reports with this number (if the testing agency tries to generate a report, the system will pop up a message "Report number is locked and needs to be unlocked after regulatory review"). Once the construction unit completes the aggregate qualification verification and submits the rectification report, and the regulatory node module approves it, the smart contract will be triggered to lift the work stoppage order, unlock the corresponding test report number, and resume the normal testing process.

[0030] This embodiment achieves trusted and automated control of the entire chain of "concrete test block sampling - evidence storage - testing - supervision" through the collaboration of various modules. It not only avoids the leakage of raw data (federated learning + encryption), but also eliminates the problems of "false reports and illegal construction" through smart contracts and report locking mechanisms, which meets the actual needs of construction project quality supervision.

[0031] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A construction project quality witnessing sampling and testing supervision system, characterized in that, The system includes: Distributed sampling terminals are deployed at various sampling points on construction sites to generate and attach RFID-QR code fusion tags the moment the specimen is formed. The edge computing node module is used to complete the verification of sampled data; run federated learning models to calculate sample risk index; aggregate multi-region model parameters and dynamically update sample risk judgment thresholds; adapt data formats and encrypt data transfer; process abnormal data and log it on the blockchain to provide data and technical support for regulatory decisions. The blockchain evidence storage module is connected to the distributed sampling terminal via a network. It is used to write the sample's birth information, the facial features of the sampler and witness, and GNSS coordinates into the consortium blockchain and return a globally unique TxID. The detection node module is used to receive sample association data pushed by the distributed sampling terminal. During the sample receiving process, it reads the EPC information of the RFID-QR code fusion tag and compares it with the TxID in the blockchain evidence storage module to verify the legality of the sample. The regulatory node module is used to automatically trigger the smart contract and issue an electronic stop-work order when the risk index is higher than the sample risk judgment threshold; at the same time, it locks the status position of the corresponding test report number, prohibiting the issuance of reports with that number.

2. The construction engineering quality witnessing sampling and testing supervision system as described in claim 1, characterized in that, The distributed sampling terminal also includes: The vibration signal detection module is used to collect characteristic vibration signals of building engineering specimens during the molding stage in real time, identify the initial setting point of the specimens, and trigger the label generation command. The tag generation and attachment module is used to receive the trigger command from the vibration signal detection unit, generate radio frequency-QR code fusion tags that conform to the technical specifications, and attach the tags to the surface of the specimen with a preset strength to achieve physical binding between the tags and the specimen. The radio frequency-QR code fusion tag has a built-in unmodifiable EPC area and an expandable user area, and the locking strength is greater than or equal to a preset locking strength threshold. The three-element data collection module is used to simultaneously collect the identity information of the sampler and witness, as well as the spatiotemporal information of the sampling scenario, and generate the basic data required for three-element verification.

3. The construction engineering quality witnessing sampling and testing supervision system as described in claim 2, characterized in that, The three-factor verification refers to: the facial feature vectors of the sampler and the witness are compared in the local trusted execution environment, and the similarity is greater than or equal to the similarity threshold; the GNSS coordinate difference between the sampling terminal and the witness terminal is less than or equal to the coordinate difference threshold; and the difference between the tag attachment timestamp and the on-chain timestamp is less than or equal to the timestamp difference threshold. Only when all three conditions are met can the data be written into the blockchain.

4. The construction engineering quality witnessing sampling and testing supervision system as described in claim 2, characterized in that, The vibration signal detection module also includes The signal acquisition unit is used to collect vibration data related to specimen molding in real time through a built-in high-sensitivity vibration sensor. The initial setting determination unit is used to analyze the collected vibration data in real time based on the preset initial setting vibration characteristic threshold of the specimen. By calculating the vibration frequency difference per unit time, when the difference is greater than the preset fluctuation threshold and the final frequency is greater than the preset sudden change critical threshold, the specimen is automatically determined to have entered the initial setting stage. The instruction triggering unit is used to immediately generate an RF-QR code fusion tag generation trigger instruction after confirming the initial solidification of the specimen, and send it to the tag generation and attachment unit.

5. The construction engineering quality witnessing sampling and testing supervision system as described in claim 2, characterized in that, The label generation and locking module also includes The tag parameter generation unit is used to generate dual-module data for RFID-QR code fusion tags according to preset tag technology standards. The physical locking execution unit is used to lock the generated radio frequency-QR code fusion tag to a preset position on the surface of the specimen through the mechanical locking mechanism built into the distributed sampling terminal; The tag attachment verification unit is used to perform a secondary detection of the tag attachment strength using a tension sensor after the tag is attached. If the strength meets the standard, a tag attachment qualification signal is generated and synchronized to the blockchain evidence storage module. If the strength does not meet the standard, a terminal alarm is triggered, prompting the operator to re-attach the tag.

6. The construction engineering quality witnessing sampling and testing supervision system as described in claim 5, characterized in that, The tag parameter generation unit further includes an radio frequency unit and a QR code unit; The radio frequency unit is used to write the SM3 hash value of sample ID + project number in the EPC area and to write the sample birth information encrypted with SM4 in the user area. The QR code unit is used to generate a visual pattern based on the hash data of the EPC area of ​​the radio frequency unit through an XOR operation, ensuring that the QR code corresponds one-to-one with the data in the EPC area.

7. The construction engineering quality witnessing sampling and testing supervision system as described in claim 2, characterized in that, The three-element acquisition module also includes: The information acquisition unit is used to acquire facial images of samplers and witnesses through the high-definition camera and local trusted execution environment (TEE) module built into the distributed sampling terminal, extract and temporarily store facial feature vectors within the TEE, and generate facial feature data that can be compared. The spatiotemporal information acquisition unit is used to acquire the current GNSS coordinates of the terminal through the GNSS positioning module integrated in the distributed sampling terminal; at the same time, it obtains the on-chain standard time of the blockchain storage module through the time synchronization module of the distributed sampling terminal, records the timestamp of the tag locking operation, and ensures that the spatiotemporal information is consistent with the on-chain time reference. The data association and upload unit is used to associate and bind the collected sampler's facial features, witness's facial features, GNSS coordinates, and the timestamp of the tag attachment operation with the sample ID of the current specimen, forming a personnel-spatiotemporal-sample associated data group, which is temporarily stored in the local encrypted storage area of ​​the terminal. After the tag generation and attachment unit confirms that the attachment is qualified, the data group is pushed to the edge computing node module to provide complete data support for the three-factor verification.

8. The construction engineering quality witnessing sampling and testing supervision system as described in claim 1, characterized in that, The edge computing node module includes: The data interaction and format adaptation unit is used to connect the distributed sampling terminal, testing agency node, regulatory node and blockchain evidence storage module bidirectionally through the network, receive the raw data uploaded by the front end, complete the format conversion and encryption / decryption; and push the business processing results and dynamic rule parameter module to the regulatory node and blockchain evidence storage module. The data verification unit is used to determine the three elements locally based on the data group associated with the personnel's facial features, GNSS coordinates, and tag-locked timestamps after processing by the data interaction and format adaptation unit. If the verification passes, a data upload permission signal is generated; if it fails, an alarm is triggered and the illegal short video is uploaded to the blockchain evidence storage module. The sample risk prediction unit is used to predict the sample risk index based on the federated learning model; it calculates and outputs the sample risk index in real time based on the pre-trained federated learning model. The federated learning management unit receives encrypted local model parameters uploaded by each testing agency node, aggregates the parameters through horizontal federation and homomorphic encryption technology, introduces a domain adaptive term into the loss function to reduce cross-regional data distribution differences, and generates an optimized federated learning model; at the same time, it dynamically updates the sample risk judgment threshold according to a preset period and synchronizes it to the regulatory node. The abnormal data processing and logging unit is used to perform localized temporary storage and preliminary analysis on the received abnormal data and generate abnormal data logs; at the same time, it records the operation logs, which are then uploaded to the blockchain after being hashed by SM3.

9. The construction engineering quality witnessing sampling and testing supervision system as described in claim 8, characterized in that, The federated learning model adopts a horizontal federation + homomorphic encryption architecture. Each detection agency, as a participant, completes model training locally and only encrypts and uploads the model parameters to the federated learning center. The loss function incorporates a domain-adaptive term; the model dynamically updates the risk assessment threshold, with an update cycle of ≤24h.

10. The construction engineering quality witnessing sampling and testing supervision system as described in claim 9, characterized in that, The smart contract automatically verifies and stores the sampling data, detection data hash, and risk index that meet the three-factor standard; when a violation signal is received, it automatically triggers the corresponding compliance handling and tracing operation for the violation event; The risk assessment thresholds are updated synchronously with the federated learning management module and take effect in real time; at the same time, key nodes in the entire data flow chain are recorded.

Images