A detection laboratory equipment whole-process management method

Abstract

The present application belongs to the technical field of laboratory management, and specifically relates to a kind of detection laboratory equipment whole process management method.The method comprises: constructing multi-protocol heterogeneous synchronous bus, using unified clock gateway to realize cross-protocol timestamp alignment and timing synchronization;Deploy edge and cloud collaborative dual-brain decision system, real-time inference is carried out through edge node and combined with cloud to carry out multi-dimensional vector aggregation analysis;Establish a test piece digital passport system based on block chain, and realize the whole life cycle data evidence of the test piece from molding to sample using radio frequency identification technology;Configure graphical process chain editor, convert detection standards into executable scripts to realize codeless driving of process.The present application can improve the linkage accuracy and automation level of detection equipment, improve the detection efficiency through the closed loop of sensing and execution, and ensure the scientific nature of detection data and the traceability of whole process.

Description

Technical Field

[0001] This invention belongs to the field of laboratory management technology, specifically relating to a method for the whole-process management of testing laboratory equipment. Background Technology

[0002] With the ever-increasing demand for quality testing in construction projects, the digital and intelligent transformation of laboratories has become a core path for the industry to improve testing accuracy and management efficiency. Traditional laboratory management models are shifting from manual operation to highly integrated automated production lines, aiming to achieve closed-loop control of the entire process from specimen preparation and curing to testing and sample retention by introducing robotics, the Internet of Things, and cloud computing. In this context, building a comprehensive management system that covers multiple disciplines and coordinates various types of testing equipment is of great significance for ensuring the objectivity of testing data and full lifecycle traceability.

[0003] The end-to-end management of testing laboratory equipment focuses on the deep coupling across equipment, processes, and data flows, involving collaborative operations between composite robots, automated guided vehicles, intelligent testing cabinets, and cloud management platforms. The core of this technology lies in establishing a real-time communication mechanism based on heterogeneous protocols and a digital twin model to integrate previously isolated processes such as mechanical loading, sample transfer, and environmental monitoring into a seamless data pipeline, achieving standardized execution of testing processes and automated acquisition of experimental data.

[0004] Existing technologies primarily focus on automating individual devices, leading to severe equipment and information silos between different testing departments, hindering cross-process collaborative scheduling. Due to the inconsistent communication protocols used by different subsystems and the lack of a unified clock synchronization benchmark, clock drift and data misalignment easily occur during the fusion of mechanical loading data, robot motion commands, and transportation status parameters. Furthermore, traditional management methods lack a reliable traceability mechanism for the entire lifecycle of test specimens, posing a significant risk of data fragmentation when specimens are transferred across departments. The testing process also heavily relies on pre-set static scripts or manual observation and judgment, failing to provide dynamic parameter compensation and real-time quality warnings for complex experimental conditions. Summary of the Invention

[0005] The purpose of this invention is to provide a method for the whole-process management of testing laboratory equipment, which can solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a method for full-process management of testing laboratory equipment, comprising the following specific steps: Step 1: Construct a global perception architecture based on a multi-protocol heterogeneous synchronous bus. Through a unified clock gateway, align the Ethernet, data distribution service, and message queue telemetry transmission protocols used for control technology across protocols to achieve time synchronization of the mechanical loading subsystem, robot collaboration subsystem, and automated guided transportation scheduling subsystem. Step 2: Deploy a dual-brain decision-making system that integrates edge and cloud computing. Utilize edge computing nodes deployed at the detection site to perform real-time reasoning on the acquired raw images. Synchronize the reasoning results and environmental parameters to the cloud via a stream processing platform. The stream processing engine then identifies quality risk patterns based on multi-dimensional vector aggregation analysis and triggers tiered alarms. Step 3: Establish a digital passport system for specimens based on blockchain technology. During the specimen molding stage, a unique identification code is written through RFID tags and QR code tags, and an initial block recording the mix ratio and test standards is generated on the blockchain sidechain to realize full life cycle data storage of specimens from molding to sample retention. Step 4: Configure the execution logic based on the graphical process chain orchestrator, encapsulate the inspection standards into executable scripts, and publish them to the corresponding inspection equipment through the industrial automation standard protocol. The edge computing node automatically parses and generates a process data object mapping table to achieve code-free driving of the inspection process. Step 5: Implement closed-loop control and result backtracking management. Write the correction parameters generated by the decision-making level back into the process data object of the device's programmable logic controller to realize dynamic compensation of the detection parameters. Write the detection results with hash verification values ​​together with the original sampling data into the digital passport to complete the archiving.

[0007] Preferably, in step 1, the process of constructing a multi-protocol heterogeneous synchronous bus involves physical-level clock calibration of protocols with different communication cycles. The unified clock gateway adopts a precise time protocol standard, unifying the millisecond-level mechanical control cycle corresponding to Ethernet used for control technology, the microsecond-level robot coordination cycle corresponding to data distribution services, and the hundred-millisecond-level scheduling cycle corresponding to message queue telemetry transmission protocols within a deviation range of microseconds. By embedding hardware timestamp generators in each communication node, it is ensured that the pressure data and displacement data collected by the sensors are encapsulated with the robot's spatial coordinate data under the same time reference.

[0008] Preferably, in step 1, a deterministic communication link is established using the Ethernet protocol for control technology to control the mechanical loading subsystem. On this link, the master station device sends control commands to the slave station device according to a preset sampling frequency, and the slave station device provides real-time feedback of the values ​​from the load sensor and displacement sensor. By embedding a specific formatted sequence number within the data frame, the problem of data packet out-of-order delivery that may occur under network congestion is resolved, ensuring the continuity and authenticity of the mechanical test curve.

[0009] Preferably, in step 2, the edge computing node's hardware configuration includes a high-performance multi-core processor and a large-capacity random access memory, and its running embedded real-time operating system supports multi-task parallel processing. In the processing flow, the edge computing node acquires real-time video streams of the specimen during the inspection process via a camera, and uses a pre-trained target detection model to identify the fracture morphology of the reinforcing steel bars or cracks on the concrete surface. The identification process is completed within a preset millisecond time, and the detected feature coordinates, crack width, and pressure value at the moment of bursting are correlated and mapped.

[0010] Preferably, in step 2, the stream processing platform employs a compression algorithm when transmitting data. This algorithm encodes image feature descriptors and sensor data streams using a global dictionary, thereby reducing bandwidth usage while ensuring data integrity. Within a preset time window, the cloud-based stream processing engine performs complex event processing on pressure, displacement, image, and environmental temperature and humidity data aggregated from multiple edge nodes. Based on preset logical criteria, such as a pressure change rate exceeding a specific threshold and a duration reaching a predetermined value, it automatically determines that the specimen has failed or is leaking, and generates an abnormal alarm signal.

[0011] Preferably, in step 3, the generation process of the specimen digital passport combines physical tags and logical blocks. At the specimen molding site, the mix proportion information, molding time, geographical location information, and construction team information of the specimen are written into the RFID tag using a handheld terminal. The blockchain sidechain uses an asymmetric encryption algorithm to sign the data, ensuring the immutability of the initial block information. Each circulation node, including the curing room, laboratory, and sample retention room, is equipped with a fixed or mobile RFID reader / writer.

[0012] Preferably, in step 3, when the automated guided vehicle carrying the specimen passes through the RFID gate, the reader reads the unique identification code within a preset millisecond time and encapsulates the current position coordinates and the temperature and humidity parameters collected by the environmental sensors into a data transfer data packet. This data packet is uploaded to the sample transfer smart contract through an encrypted channel. The contract automatically determines whether the specimen meets the admission conditions for entering the next testing stage based on preset logical conditions, such as whether the age has reached a predetermined number of days and whether the cumulative temperature under the same curing conditions has reached a predetermined threshold. If the conditions are met, the system automatically sends a grasping permission command to the robot control system.

[0013] Preferably, in step 4, the graphical process chain orchestrator provides a visual logic modeling interface. Test engineers can construct a complete testing flowchart by dragging and dropping different functional components, such as loading modules, pressure holding modules, photo evidence collection modules, and unloading modules. The orchestrator automatically converts the flowchart into an executable script in a linked data format. This executable script contains the execution parameters, logical decision branches, and exception handling mechanisms for each process step.

[0014] Preferably, in step 4, when parsing the executable script, the edge computing node establishes communication with the specific testing equipment through an industrial automation standard protocol. For concrete impermeability testing equipment, the edge computing node converts the pressurization step size and holding time in the executable script into a process data object mapping table controlling the solenoid valve and servo motor. During execution, the edge computing node monitors the feedback values ​​of the pressure sensor in real time and dynamically adjusts the pressurization frequency or terminates the test based on preset criteria algorithms, such as sudden changes in the slope of the leakage pressure versus time curve, thus achieving unmanned and intelligent testing.

[0015] Preferably, in step 5, closed-loop control is achieved by feeding back the optimized parameters output by the cloud-based decision layer to the actuator in real time. For example, in a steel bar tensile test, if the edge computing node detects through image analysis that the tensile speed causes an abnormal fracture shape, the cloud-based decision model will calculate the corrected displacement rate and send it to the device controller via a high-speed communication link. The controller adjusts the output pulse of the motor driver in real time according to the corrected parameters, thereby achieving real-time compensation for the physical execution process.

[0016] Preferably, in step 5, the result backtracking process requires all test data to have a complete chain of evidence. The final report after testing not only includes result parameters such as maximum force and intensity level, but also uses a hash algorithm to calculate a digest of the pressure curve, displacement data, and key frame videos of the test process. The generated hash digest, along with the storage path of the original data, is written into subsequent blocks of the blockchain. Third-party institutions, holding the corresponding public key, can verify the integrity of the data in the report, ensuring that every step from specimen molding to result output is traceable and tamper-proof.

[0017] Preferably, for intelligent testing of concrete impermeability, the management method is implemented by configuring an intelligent impermeability cabinet. The intelligent impermeability cabinet contains multiple independent sealing units, each using a die-cast specimen container. The inner wall of the specimen container is embedded with a composite material air bladder, which is made of reinforced nylon and natural rubber laminated together. By filling the air bladder with compressed air at a preset pressure, a flexible seal is achieved on the sides of the concrete specimen. The pressure during the sealing process is monitored in real time by a high-precision pressure sensor, and the sampling interval is set to a preset value to ensure that no side leakage occurs during pressurized testing.

[0018] Preferably, the intelligent anti-seepage cabinet is also equipped with a lifting mechanism driven by a servo motor and a ball screw. When the composite robot performs loading and unloading operations, the lifting mechanism moves the specimen bucket to a predetermined height, achieving a repeatability accuracy at the micron level. The end effector of the composite robot is equipped with a pneumatic gripper integrating both pressure and displacement sensors. Real-time feedback of the gripping force data prevents damage to the specimen during handling and ensures that the flatness of the specimen within the sealing plane meets preset threshold requirements.

[0019] Preferably, at the detection algorithm level, a sudden change in the slope of the leakage pressure versus time curve is used as the criterion. The system acquires the real-time water pressure in each specimen barrel through independent pressure sensors. When the pressure value is detected to decrease by more than a predetermined threshold within a preset time step, and the corresponding slope change rate satisfies a specific logical relationship, the system automatically determines that the specimen has leaked. The edge computing node then triggers an industrial camera to capture an image of the bottom of the current specimen, confirms the leakage height or leakage point distribution through image recognition algorithms, and updates this information to the specimen's digital passport in real time.

[0020] Preferably, the scheduling layer of the automated guided vehicle (AGV) employs a dual-channel communication mechanism. The message queue telemetry transmission protocol receives task instructions and path planning parameters from the management platform, while the distributed stream processing platform uploads high-frequency sensor data streams, including LiDAR scan data, ultrasonic obstacle avoidance data, and battery status parameters. By binding geographical location information with a unique identification code, the system can present the precise spatial coordinates and current circulation status of each specimen in real time within the digital twin model.

[0021] Preferably, the cloud-based digital twin is built upon a stream processing engine. This engine has multiple pre-set anomaly pattern matching models, such as a temperature and humidity instability mode for the maintenance environment. When the average temperature fluctuation of multiple sampling points in the maintenance room deviates from the standard curve within a predetermined time, the system automatically triggers an alarm and sends an airflow adjustment command to the maintenance control system. This predictive maintenance and closed-loop adjustment based on real-time data ensures the constancy of the testing environment and the reliability of the testing results.

[0022] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a multi-protocol heterogeneous synchronization bus and uses a unified clock gateway to achieve microsecond-level timestamp alignment of detection devices, robots and transportation systems under different communication protocols. This solves the problem of data misalignment and coordination failure caused by clock drift of each subsystem in traditional laboratories, and significantly improves the accuracy and reliability of cross-device linkage.

[0023] 2. This invention employs a dual-brain decision-making system that coordinates edge and cloud computing, leaving high-frequency video stream inference at the edge while handling complex multi-dimensional vector aggregation analysis in the cloud. This reduces network bandwidth load while ensuring real-time detection. By transforming artificial intelligence criteria into correction parameters for physical execution, closed-loop control from perception to execution is achieved, improving the automation level and anomaly handling capabilities of the detection process.

[0024] 3. This invention introduces a blockchain-based digital passport for test specimens, utilizing radio frequency identification and encryption algorithms to store the entire lifecycle data of the specimens on the blockchain, eliminating information silos and data fragmentation during cross-departmental transfers. Hash verification and distributed storage mechanisms ensure the authentic correspondence between test data, video recordings, and physical samples, achieving reliable traceability throughout the entire testing process and enhancing industry credibility.

[0025] 4. This invention achieves code-free driving of testing processes through a graphical process chain orchestrator, transforming complex testing standards into machine-readable script logic and reducing the system's reliance on manual programming. This flexible process configuration method can quickly adapt to different types of testing needs, improving the laboratory's flexible production capabilities and management efficiency in handling diverse engineering tasks.

[0026] 5. This invention features an intelligent anti-seepage sealing structure and a high-precision motion control module designed for specific testing scenarios. Through the optimization of the physical structure and the coupling of sensing algorithms, it achieves automatic sealing, real-time leakage detection, and precise handling of concrete specimens. This not only reduces the workload of testing personnel but also eliminates the interference of human factors on the test results through precise parameter control, ensuring the scientific validity and accuracy of the test data. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall technical solution architecture according to the present invention; Figure 2 This is a schematic diagram illustrating the core principle framework of the dual-brain decision-making system that integrates edge and cloud collaboration according to the present invention. Figure 3 The following is a logical flow diagram illustrating the construction of a multi-protocol heterogeneous synchronization bus to achieve cross-protocol timing synchronization according to the present invention. Figure 4 This is a flowchart illustrating the logical process of generating digital passports for test specimens based on a blockchain sidechain and achieving full lifecycle evidence storage according to the present invention. Figure 5 This is a logic flow diagram of the present invention, which uses a graphical process chain orchestrator to encapsulate the detection standard and drive the device. Figure 6This is a schematic diagram of the multi-level interaction and data flow between the reverse writing of decision-level correction parameters and the archiving of detection results with hash verification according to the present invention. Detailed Implementation

[0028] Example 1: Please refer to the appendix Figure 1 To be continued Figure 6 To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0029] In implementing the full-process management method for testing laboratory equipment, the first step is to construct a global perception architecture based on a multi-protocol heterogeneous synchronous bus. The physical foundation of this architecture lies in establishing a unified time coordinate system that spans different communication standards. In the complex electromagnetic environment of a construction engineering quality testing laboratory, there are three subsystems: a mechanical loading subsystem, a robot collaboration subsystem, and an automated guided transportation scheduling subsystem. These three subsystems employ protocols with drastically different communication characteristics. Step 1, the process of constructing the multi-protocol heterogeneous synchronous bus, involves physical clock calibration of protocols with different communication cycles. The unified clock gateway, as the core hardware node, adopts a precise time protocol standard, namely, hardware-level timestamp marking technology, to forcibly align the originally isolated clock sources.

[0030] In one specific embodiment, the unified clock gateway adopts a method based on... The hardware-assisted clock synchronization mechanism of the precision time protocol. Its core synchronization process consists of two stages: offset measurement and delay measurement.

[0031] First, the master clock (gateway) periodically sends Sync messages to all slave clocks (each protocol node) and records the precise timestamp of the transmission. Receive Sync messages from the clock and record the precise timestamp of the received message. .

[0032] Immediately afterwards, the master clock sent a message carrying... The slave clock then sends a Follow_Up message to the master clock, and records the sending timestamp. The master clock receives Delay_Req messages and records the received timestamp. and through the Delay_Resp message It is sent back to the slave clock.

[0033] Using the four timestamps mentioned above, the clock offset between the clock and the master clock can be calculated from the slave clock. With network transmission delay The calculation formula is as follows: in, The timestamp of the master clock sending the Sync message. For the timestamp of the Sync message received from the clock, The timestamp for sending the Delay_Req message from the clock. The timestamp for receiving Delay_Req messages from the master clock.

[0034] The local clock dynamically adjusts its frequency and phase via hardware registers based on the calculated offset value, achieving nanosecond-level synchronization with the master clock. For different protocol periods, the gateway's internal scheduling unit, using a unified nanosecond-level time base, adds a global timestamp to each protocol's communication event via hardware interrupts. For example, for a data distribution service protocol node operating on a microsecond period, each data packet it sends is captured by the hardware circuitry and appended with the current global timestamp the instant it is sent at the physical layer. This ensures the physical alignment of timing across protocols, ultimately controlling the time deviation of all protocols within a preset microsecond range.

[0035] Ethernet protocols used for control technology typically handle mechanical control tasks with millisecond-level cycles, their internal clock cycle set to 1 millisecond, primarily responsible for closed-loop feedback from load and displacement sensors. Data distribution service protocols operate at microsecond-level cycles, ensuring the multi-joint coordination accuracy of six-axis composite robots or collaborative robots in spatial motion. Message queue telemetry transmission protocols operate at hundreds of millisecond-level cycles, mainly responsible for global path planning and task command distribution for automated guided vehicles. A unified clock gateway extracts synchronization messages from Ethernet frames and generates synchronization pulses at the physical layer, unifying these three different cycle scales within a tolerance range.

[0036] In step 1 above, a deterministic communication link is established using the Ethernet protocol for control technology to control the mechanical loading subsystem. On this link, the master station sends control commands to the slave stations distributed at various testing stations at a preset sampling frequency, such as 1000 Hz. Upon receiving the command, the slave station triggers a hardware timestamp generator to encapsulate the pressure and displacement data collected by the sensors. To ensure that the data packet order remains intact under network congestion or data fluctuations, the system embeds a specific formatted sequence number within the data frame. This sequence number has a linear correspondence with the global nanosecond-level timestamp distributed by the clock gateway. When the slave station feeds back the values ​​from the load and displacement sensors, the master station reassembles them according to the sequence number and timestamp, resolving the potential data packet disorder problem under high network load and ensuring the continuity and accuracy of the mechanical test curve on the time axis.

[0037] Step 2 is then executed, deploying a dual-brain decision-making system that integrates edge and cloud computing. This system relies on edge computing nodes deployed at the testing site. These nodes utilize high-performance multi-core processors, such as high-computing modules with 64GB of random access memory, and their internal embedded real-time operating system supports multi-task parallel processing. In the processing flow, the edge computing nodes acquire real-time video streams of specimen testing captured by industrial cameras via gigabit Ethernet ports. For rebar tensile tests, the edge computing nodes use pre-trained deep residual network models or target detection models to perform real-time inference on the necking morphology and fracture location of the rebar during the tensile process.

[0038] Specifically, the pre-trained object detection model adopts the YOLOv5s network architecture. This network consists of three parts: a backbone network, a neck network, and a head network.

[0039] The backbone network adopts a CSPDarknet53 structure to extract multi-scale feature maps from the input image. The input image resolution is... The tensor. The backbone network downsamples through convolutional layers and cross-stage local network modules, sequentially outputting feature maps. , and The height, width, and number of channels are respectively expressed as: .

[0040] The neck network employs a path aggregation network to process the multi-scale feature maps output by the backbone network. Feature fusion is performed. It combines high-level semantic features with low-level localization features through upsampling and lateral connections, generating three fused feature maps with rich semantic information. , , .

[0041] The head network performs object detection on each fused feature map. For each scale, the model predicts three prior boxes with different aspect ratios. The final output tensor has a dimension of [missing value]. ,in The feature map grid size, The number of anchor frames per grid. Corresponding bounding box coordinates and confidence level , To detect the number of categories (e.g., necking, cracks, fractures, etc.), post-processing is performed on the model output during the inference phase. First, a threshold is used to filter out predicted boxes with low confidence. Then, a non-maximum suppression algorithm is used to remove overlapping boxes. The execution logic of the non-maximum suppression algorithm is as follows: for predicted boxes of the same category, they are sorted from high to low confidence. The predicted box with the highest confidence is selected, and its intersection-union ratio (IUR) is calculated with the remaining predicted boxes. When the IUR exceeds a preset threshold... If the prediction fails, the remaining bounding box is deleted. The final output is the detected target category, bounding box coordinates, and corresponding confidence score.

[0042] In step 2, the inference process of the edge computing node is completed within a preset 2 milliseconds, and the coordinates of the detected feature points, the crack width, and the pressure value at the moment of specimen bursting are correlated and mapped. This localized processing avoids the latency caused by the transmission of high-definition images over the network. Simultaneously, the stream processing platform employs a specific compression algorithm when transmitting data to the cloud. This algorithm is not a simple general compression, but rather encodes the feature descriptors of the image and the data stream generated by the sensor by establishing a global dictionary. During transmission, the data stream is divided into a metadata layer and a raw sampling layer, and only the encoded feature vectors are synchronized to the cloud through the stream processing platform, thereby reducing bandwidth consumption while ensuring data integrity.

[0043] Within a preset 5-second sliding time window, the cloud-based stream processing engine performs complex event processing on pressure, displacement, image, and environmental temperature and humidity data aggregated from multiple edge nodes. The engine incorporates multi-dimensional vector aggregation analysis criteria. When the pressure change rate exceeds a specific threshold and the duration on the time axis reaches a predetermined value, combined with image recognition of water seepage marks or fracture irregularities on the specimen surface, the system automatically determines that the specimen has failed or is leaking prematurely. At this point, the decision-making layer triggers a tiered alarm mechanism, sending push notifications to mobile terminals or directly issuing emergency stop commands to on-site actuators based on the risk level.

[0044] As a preferred implementation method, the specific implementation process of the multidimensional vector aggregation analysis is as follows: First, define the multidimensional feature vector. ,in, This is the statistical feature vector of the pressure time series, including the mean, variance, and peak value of the rate of change; This is the statistical feature vector of the displacement time series; The feature vector output for image recognition, such as the maximum crack width and the encoding vector of fracture morphology; It is a vector composed of ambient temperature and humidity.

[0045] The maintenance length of the cloud stream processing engine is A sliding time window is used to perform aggregation analysis on the converged vectors within each window. A comprehensive risk index is defined. The calculation formula is as follows: in, Representing vectors The Middle The current value of each feature. and These are the historical mean and standard deviation of this feature under normal operating conditions (obtained through offline statistics). The weight coefficient corresponding to this feature (can be determined by the analytic hierarchy process based on expert experience). For the total number of features, The Sigmoid activation function maps the output to... Interval.

[0046] When the calculated comprehensive risk index Exceeding the preset risk threshold When this occurs, the system determines that a quality risk pattern exists and triggers a tiered alarm. For example, if... If so, a yellow alert will be triggered; if If this occurs, a red alarm will be triggered and an emergency stop command will be sent to the actuator.

[0047] Following step 3, a digital passport system for test specimens based on blockchain technology is established. The generation process of the digital passport combines physical tag identification with logical block storage. At the specimen molding site, workers use handheld terminals to write the specimen's mix proportion information, specific molding time, geographical coordinates, and construction team information into RFID tags. In step 3, each specimen is assigned a globally unique identification code. The blockchain sidechain employs an asymmetric encryption algorithm, using a private key to sign the initial block data, ensuring that the initial block information recording the mix proportion and experimental standards cannot be tampered with in all subsequent circulation stages.

[0048] Fixed RFID readers are installed at various points in the laboratory, including the constant temperature and humidity curing room, the mechanical laboratory, and the sample retention room. When an automated guided vehicle carrying a specimen passes through the RFID access control, the reader quickly reads the unique identification code and encapsulates the current location coordinates and real-time temperature and humidity parameters collected by environmental sensors into a data transfer data packet. This data packet is uploaded to the sample transfer smart contract via an encrypted channel. The smart contract has strict pre-set access logic criteria, such as whether the curing age of the specimen is greater than or equal to 28 days, and whether the cumulative temperature under the same curing conditions has reached a predetermined threshold of 1200 degrees Celsius per day. Only when all logical conditions are met will the system automatically send a grasping permission command to the robot control system. This data-driven process permission mechanism eliminates detection errors and violations caused by human intervention, ensuring that every specimen entering the testing stage meets the standard specifications.

[0049] Step 4 then proceeds to configure the execution logic based on the graphical process chain orchestrator. The graphical process chain orchestrator provides test engineers with a visual logic modeling interface. Engineers can drag and drop different functional components, such as loading modules, holding modules, video monitoring modules, and unloading modules, to build a complete testing flowchart, much like assembling building blocks. The orchestrator automatically converts this graphical logic into executable scripts that conform to the linked data format. These executable scripts define in detail the execution parameters for each step, such as the pressurization rate, holding time, sensor sampling frequency, and handling paths for abnormal situations.

[0050] In step 4, the edge computing node parses the executable script and establishes bidirectional communication with the specific testing equipment through industrial automation standard protocols, such as the Open Platform Communication Unified Architecture Protocol. Taking concrete seepage resistance testing as an example, the edge computing node converts the pressurization step size and pressure holding cycle parameters in the script into a process data object mapping table that controls the solenoid valves and servo motors in the intelligent seepage resistance cabinet. During execution, the edge computing node monitors the feedback values ​​of the pressure sensor in real time and dynamically adjusts the opening frequency of the pressurization solenoid valve according to a preset slope change criterion algorithm. If an abnormal change in the slope of the leakage pressure versus time curve is detected, i.e., the pressure drop rate exceeds the preset critical value, the system will automatically terminate the test and record the current pressure point. This no-code process-driven approach enables the laboratory to quickly respond to testing requirements of different standards and improves the flexibility of equipment management.

[0051] Finally, step 5 is executed to implement closed-loop control and result feedback management. The essence of closed-loop control lies in feeding back the optimization parameters output by the cloud-based decision-making layer to the underlying physical actuators in real time. In the example of rebar tensile testing, if the edge computing node detects through real-time image analysis that the current tensile speed is causing the fracture shape to deviate from the standard, the cloud-based decision model will calculate the corrected displacement rate in real time and send this parameter to the device controller via a high-speed communication link. After receiving the corrected parameter, the controller immediately adjusts the output pulse frequency of the motor driver to achieve real-time dynamic compensation for the physical execution process.

[0052] During the result backtracking process in step 5, the system requires all test data to have a complete chain of evidence. The final report after testing not only includes result parameters such as maximum force and intensity level, but also uses a hash algorithm to calculate a digest of the original pressure curve, displacement sampling data, and key frame video of the test process. The generated hash digest, along with the distributed storage path of the original data, is written into subsequent blocks of the blockchain. Any third-party institution holding the corresponding public key can compare and verify the digest in the report. If the original data has undergone any byte-level displacement during storage or transmission, its generated hash value will be completely different from the digest recorded on the blockchain. This traceability mechanism ensures that every step from specimen molding, curing, and transfer to the final result output is under control, achieving truly reliable end-to-end management.

[0053] In the result backtracking process of step 5, the hash algorithm specifically adopts the SHA-256 algorithm. Before calculating the digest, the multi-source data needs to be preprocessed and concatenated. The specific process is as follows: First, convert the time-series data, such as the raw pressure curve data and displacement sampling data, into standard JSON format text. For the keyframe video of the experiment, the keyframe images were first extracted and compressed into JPEG format binary data. .

[0054] Then, according to the preset splicing order (such as...) Metadata is used to sequentially concatenate the data blocks to form the original message. .in This indicates a data concatenation operation, and the metadata includes text information such as the test number and timestamp.

[0055] For the original message The specific steps for performing hash calculations include: Padding: in the message Add a 1 and multiple 0 bits to the end to make its length meet the requirement. Then, a 64-bit binary number representing the original message length is appended. The padded message is an integer multiple of 512 bits in length.

[0056] Blocking and Processing: Divide the padded message into multiple 512-bit message blocks. Initialize eight 32-bit link variables. These are predefined constants. For each message block The compression function performs 64 rounds of computation. Each round updates the linker variables using a specific logical function, round constants, and words derived from the current message block. The core computation of the compression function can be summarized as follows: in, , This is a circular right shift XOR function. , For bitwise logic functions, For the first wheel constant, This refers to the word obtained by expanding from the current message block. The update rule is: , , , , , , , .

[0057] After processing all message blocks, the final result is The concatenation results in a 256-bit hash digest value. This digest value, along with the distributed storage path of the original data, is written into subsequent blocks of the blockchain for subsequent integrity verification.

[0058] In a further refined application scenario, the management method is implemented through the configuration of an intelligent anti-seepage cabinet. The intelligent anti-seepage cabinet contains multiple independent sealing units, each using a single-piece die-cast specimen container. The inner wall of the specimen container is embedded with a composite material airbag, which is made of reinforced nylon and natural rubber laminations, possessing high pressure resistance and good sealing flexibility. Before testing begins, the system fills the airbag with compressed air at a preset pressure to achieve a flexible seal on the sides of the concrete specimen. The pressure during the sealing process is monitored in real time by a high-precision pressure sensor. To ensure the reliability of the seal, the system is set to a total sealing time of less than or equal to 5 minutes per test, and the airbag pressure is kept constant during the pressurization test.

[0059] The intelligent anti-seepage cabinet is also equipped with a lifting mechanism driven by a servo motor and a ball screw. The lifting mechanism achieves extremely high repeatability, ensuring precise docking of the specimen barrel with the robot's end effector during loading and unloading. The composite robot's end effector is equipped with a pneumatic gripper integrating pressure and displacement sensors. Through real-time feedback of gripping force data, the system can precisely control the gripping force, preventing damage to the specimen due to excessive force or slippage due to insufficient force. At the detection algorithm level, the system uses a slope abrupt change criterion based on the leakage pressure versus time curve. When the water pressure value collected by the independent pressure sensor decreases by more than a set threshold within a preset time step, and the corresponding slope change rate satisfies a logically decreasing relationship, the system determines that the specimen is leaking. The edge computing node then triggers an industrial camera to capture images of the seepage at the bottom of the specimen. Image recognition algorithms are used to confirm the location distribution of the seepage points, and this information is updated in real-time to the specimen's digital passport, achieving a deep integration of physical detection and digital evidence storage.

[0060] In the intelligent detection scenario of concrete seepage resistance, the image recognition algorithm confirms the distribution of seepage points through a semantic segmentation model. In one specific embodiment, a U-Net network architecture is adopted.

[0061] The U-Net network contains one encoder path and one decoder path. The encoder consists of four downsampling stages, each containing two... A convolutional layer (followed by batch normalization and ReLU activation function) and a Max pooling layer. Input image size is... After passing through the encoder, the feature map size is gradually halved, and the number of channels becomes 64, 128, 256, and 512 respectively, ultimately yielding... The bottleneck characteristics.

[0062] The decoder path also contains four upsampling stages. Each stage first passes through... The transposed convolution doubles the feature map size and halves the number of channels, then performs a skip connection with the corresponding feature map size in the encoder, and then passes through two... The convolutional layers process the data. Finally, the decoder outputs... The feature map, where channel dimension 2 corresponds to the background and the seepage area, respectively.

[0063] When training this segmentation model, a weighted Dice coefficient loss function is used to address the issue of small pixel proportions in the water seepage area. Loss function Defined as: in, Represents pixels The true label (0 for background, 1 for water seepage). This indicates the probability that the model predicts this pixel to be a water seepage area. Minimum value (e.g.) This is to prevent the denominator from being zero.

[0064] During inference, the bottom image of the specimen captured by an industrial camera is input into the trained U-Net model, which outputs a probability map of the seepage area. By thresholding the probability map (e.g., probability greater than 0.5) and performing connected component analysis, parameters such as the coordinates of the seepage point location and the seepage area can be obtained and updated in the specimen's digital passport.

[0065] Example 2: In another specific implementation scenario, the full-process management method for testing laboratory equipment described in this invention is applied to an automated testing line for steel bar tensile and bending tests. In this example, the multi-protocol heterogeneous synchronization bus constructed in step 1 primarily addresses the issue of cycle synchronization between multiple testing machines and the central composite robot. Because the steel bar tensile test requires extremely high precision in displacement control, the system utilizes the Ethernet protocol for control technology to establish a master-slave synchronization model. The master station clock gateway sends a synchronization heartbeat message to all slave stations every 1 microsecond, ensuring that the control systems of each testing machine and the robot controller have a completely consistent time reference.

[0066] In step 2, edge computing nodes are deployed at the front end of each universal testing machine. To address the fracture spatter and severe vibrations generated during the tensile testing of the reinforcing steel, the high-frequency sampling program running on the edge computing nodes acquires load data at a rate of 5000 times per second. At the instant the reinforcing steel undergoes brittle fracture, the sensor captures a massive stress drop signal. The edge computing node immediately locks the timestamp of that moment and instructs a high-speed camera to capture a real-time image of the fracture surface. This high-frequency sampled data and image features are pushed to the cloud via the Kafka streaming platform. Large-scale language models or multimodal analysis models in the cloud automatically classify the metallurgical characteristics and defect distribution of the fracture surface, identifying potential quality defects such as slag inclusions and porosity.

[0067] In the specimen digital passport system of step 3, ultra-high frequency radio frequency identification (UHF) technology is used to address the characteristics of large-volume samples like steel bars. During the steel bar forming or cutting stage, electronic tags containing information such as material, furnace number, and specifications are attached to the binding straps. When the steel bar samples enter the laboratory's automated storage system, RFID readers simultaneously read multiple tags within a 3-meter range and bind the read UUIDs to the spatial coordinates of the storage location. Whenever a specimen is removed or moved by a robot, a new transaction is recorded on the blockchain sidechain, detailing the specimen's displacement path. This transparent flow record reduces the probability of sample confusion during the testing process to zero.

[0068] Driven by the process chain in step 4, the graphical programmable controller pre-sets various tensile curve templates for steel bars of different strength grades. The testing engineer only needs to select the corresponding steel bar specification, and the programmable controller automatically generates a control command stream including the elastic stage, yield stage, hardening stage, and necking stage. These command streams are sent to the testing machine's programmable logic controller (PLC) via an open platform communication protocol. The PLC precisely controls the opening degree of the hydraulic servo valve based on the stress increase rate defined in the script, using a proportional-integral-derivative (PID) adjustment algorithm. This no-code configuration method not only shortens the cycle for implementing new standards but also ensures a high degree of consistency in execution logic across different batches of tests.

[0069] In step 5, the system establishes a comprehensive closed-loop mechanism for abnormal test results. If the cloud-based decision-making level detects significant dispersion in the tensile strength of a batch of steel bars, the system will automatically trigger a re-inspection process chain. The correction parameters issued by the decision-making level will change the sampling frequency and loading step size of the next set of samples. All original tensile displacement curves, load curves, and high-definition videos of fracture surfaces are stored in a distributed storage system after hash calculation. The final output electronic test report includes a digital signature and an anti-counterfeiting QR code. Third-party regulatory agencies can retrieve the entire lifecycle data summary of the specimen from the blockchain—from steel entry, sampling, processing, curing, testing to scrapping—by scanning the QR code, achieving fully transparent supervision of the testing industry.

[0070] At the scheduling level of the automated guided vehicle (AGV), the system employs a dual-channel communication mechanism to ensure real-time performance and stability. The message queue telemetry transmission protocol serves as the command channel, using a publish / subscribe model to distribute scheduling commands from the management platform to each transport terminal. Meanwhile, the distributed stream processing platform acts as the data channel, responsible for real-time transmission of the massive point cloud data streams generated by the lidar. By introducing a geospatial indexing algorithm at the scheduling layer, the system can present the precise spatial location of each specimen in the digital twin model with centimeter-level accuracy in real time. When the transport vehicle detects personnel or obstacles ahead, the edge computing unit makes obstacle avoidance decisions through rapid reasoning and simultaneously records the time and reason for the obstacle avoidance action in the flow log, ensuring the safety and reliability of the transportation process.

[0071] The cloud-based digital twin, built on a stream processing engine, is not only a mirror of the physical world but also the core of predictive maintenance. The engine has pre-set models for matching various anomaly patterns, such as those for temperature and humidity instability in the environmental curing room. When the average fluctuation of temperature sensor values ​​at different heights within the curing room deviates from the standard curve within a predetermined time, even if individual values ​​remain within acceptable ranges, the system will determine an anomaly in the air circulation system through trend analysis. Subsequently, the cloud-based decision-making layer automatically sends airflow adjustment commands and dehumidifier operating power correction parameters to the curing control system. This closed-loop adjustment, driven by real-time data, ensures extreme stability of the monitored environment and reduces equipment failure rates through preventative compensation.

[0072] Example 3: In an example concerning the testing of the physical and mechanical properties of cement, the full-process management method described in this invention demonstrates its advantages in controlling minute physical quantities. The synchronous bus constructed in step 1 tightly couples the automatic cement mixer, flexural strength testing machine, compressive strength testing machine, and robotic handling system. Since cement specimens are extremely sensitive to time points during testing, any delay in any step will affect the strength determination. The unified clock gateway ensures that the time span from the end of mixing to molding and warehousing is precisely controlled.

[0073] In the decision-making system of step 2, edge computing nodes analyze the displacement amplitude of cement specimens on the vibration table to determine the density of the specimens in real time. The flow processing engine aggregates the mix proportion, water-cement ratio, and current fluctuation curves during the mixing process in the cloud. Through complex event processing, the system can predict the strength development trend of cement specimens after 3 or 28 days. If the prediction results indicate insufficient strength, the alarm system will intervene in advance to remind the test personnel to check the quality of raw materials, thus realizing the forward shift of quality control.

[0074] In step 3, considering the small size of the cement samples, a scheme using QR code labels and RFID tags as backups was adopted. The digital passport records in detail the cement manufacturer, type, strength grade, and sampling records after entry. During the transfer process, each operator's employee number and operation time are recorded on the blockchain. When the specimen enters the flexural testing machine, the force sensor at the robot's end senses the placement angle of the specimen. If the angle deviates from the centerline by more than 0.5 mm, the system automatically corrects the deviation and stores the correction action as quality metadata on the blockchain.

[0075] In step 4, the graphical process chain orchestrator encapsulates the logic for automatic gear shifting and automatic range switching, catering to different national standards for cement flexural and compressive strength testing. After the edge computing nodes parse the script, they directly drive the hydraulic system of the testing machine to apply load at a constant rate. At the moment the specimen breaks, the system automatically records the peak force and calculates the strength. All these operations require no manual intervention, truly achieving unmanned operation of the testing process.

[0076] Finally, in step 5, the archiving process of the test results includes complete original waveform data. Unlike traditional methods that only record numerical results, this embodiment uses digital hash storage to preserve the full stress-strain curve during the cement's compressive crushing process. This means that even several years after the test, as long as the specimen's digital passport identification code is held, the original test environment and stress process can be completely reproduced. This deep data backtracking capability provides irrefutable technical support for resolving engineering quality disputes.

[0077] The entire system abstracts the construction engineering quality testing laboratory into a super-large measuring machine capable of self-sensing, self-decision-making, and self-execution. Its level of intelligence is not simply reflected in the automation of individual devices, but rather in the use of a comprehensive perception layer, a dual-brain decision-making layer, a blockchain passport layer, a no-code execution layer, and a closed-loop control layer to connect previously isolated professional departments into a seamless, auditable data pipeline. The technological breakthrough lies in solving the microsecond-level time synchronization problem across protocols, transforming the visual criteria of artificial intelligence into control parameters that can reverse-correct the actions of physical equipment. The digital passport for test specimens established through blockchain technology completely eliminates information silos in cross-departmental workflows.

[0078] Furthermore, in terms of specific hardware implementation, the edge computing nodes adopt an advanced system-on-a-chip architecture, integrating an artificial intelligence acceleration unit, enabling them to complete complex video stream parsing tasks with extremely low power consumption. The stream processing platform employs a distributed architecture, possessing high throughput and low latency characteristics, ensuring that every raw record generated from massive sensors can be processed in real time. The cloud-based digital twin model, by continuously absorbing data from on-site feedback, enables preventative maintenance of experimental equipment and continuous optimization of testing processes.

[0079] In summary, this invention constructs a closed-loop laboratory management system covering the entire lifecycle of test specimens through the deep integration of a series of technologies, including multi-protocol synchronous bus, edge cloud collaboration, blockchain notarization, and no-code orchestration. This management approach not only significantly reduces reliance on human experience and eliminates the interference of human factors on test results, but also improves the scientific rigor, accuracy, and authenticity of test data through high-precision parameter control and full-process data notarization. It addresses the pain points of traditional laboratory data being scattered, difficult to coordinate, and prone to falsification, providing a novel technical means based on digital twins and trust chains for the quality supervision of construction projects.

[0080] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for managing the entire process of testing laboratory equipment, characterized in that, Includes the following steps: Step 1: Construct a global perception architecture based on a multi-protocol heterogeneous synchronous bus. Through a unified clock gateway, align the Ethernet, data distribution service, and message queue telemetry transmission protocols used for control technology across protocols to achieve time synchronization of the mechanical loading subsystem, robot collaboration subsystem, and automatic guidance transportation scheduling subsystem. Step 2: Deploy a dual-brain decision-making system that integrates edge and cloud computing. Utilize edge computing nodes deployed at the detection site to perform real-time reasoning on the acquired raw images. Synchronize the reasoning results and environmental parameters to the cloud via a stream processing platform. The stream processing engine then identifies quality risk patterns based on multi-dimensional vector aggregation analysis and triggers tiered alarms. Step 3: Establish a digital passport system for specimens based on blockchain technology. During the specimen molding stage, a unique identification code is written through RFID tags and QR code tags, and an initial block recording the mix ratio and test standards is generated on the blockchain sidechain to realize full life cycle data storage of specimens from molding to sample retention. Step 4: Configure the execution logic based on the graphical process chain orchestrator, encapsulate the inspection standards into executable scripts, and publish them to the corresponding inspection equipment through the industrial automation standard protocol. The edge computing node parses and generates a process data object mapping table to realize code-free driving of the inspection process. Step 5: Implement closed-loop control and result backtracking management. Write the correction parameters generated by the decision-making level back into the process data object of the device's programmable logic controller to realize dynamic compensation of the detection parameters. Write the detection results with hash verification values ​​together with the original sampling data into the digital passport to complete the archiving.

2. The method for full-process management of testing laboratory equipment according to claim 1, characterized in that, In step 1, constructing a multi-protocol heterogeneous synchronization bus includes: adopting a precise time protocol standard and using a unified clock gateway to unify the millisecond-level mechanical control cycle corresponding to Ethernet for control technology, the microsecond-level robot collaboration cycle corresponding to data distribution service, and the hundred-millisecond-level scheduling cycle corresponding to message queue telemetry transmission protocol to a preset microsecond-level deviation range. By embedding hardware timestamp generators at each communication node, it is ensured that the pressure data and displacement data collected by the sensors are encapsulated with the robot's spatial coordinate data under the same time reference; For the control of the mechanical loading subsystem, a deterministic communication link is established using the Ethernet protocol used for control technology. The master station device sends control commands to the slave station device according to a preset sampling frequency, and the slave station device provides real-time feedback of the values ​​from the load sensor and displacement sensor.

3. The method for full-process management of testing laboratory equipment according to claim 2, characterized in that, In step 1, the process of achieving time synchronization further includes: embedding a formatted sequence number within the data frame, wherein the sequence number is linearly correlated with a global nanosecond-level timestamp distributed by the clock gateway; When the slave device feeds back sensor values ​​to the master device, the master device performs logical reassembly of the received data packets according to the sequence number and the global nanosecond-level timestamp. By restoring the original sending order of the data packets, the disorder of data packets caused by network congestion is resolved, and the continuity and authenticity of the mechanical test curve on the time axis are maintained. Through a high-precision time synchronization mechanism, the robot can accurately record spatial coordinates at the moment when the mechanical loading reaches a preset threshold, thus achieving deep coupling between physical and spatial quantities in the time dimension.

4. The method for full-process management of testing laboratory equipment according to claim 3, characterized in that, In step 2, the real-time inference of the acquired raw images using edge computing nodes includes: using the high-performance multi-core processor and random access memory of the edge computing nodes to run an embedded real-time operating system to support multi-task parallel processing. By acquiring real-time video streams of specimen inspection captured by industrial cameras, a pre-trained target detection model is used to identify the necking morphology and fracture location of steel bars during the tensile process, or cracks on the concrete surface. The identification process is completed within a preset millisecond time, and the coordinates of the detected feature points, the crack width, and the pressure value at the moment of specimen bursting are correlated and mapped. By using localized inference, the latency caused by transmitting high-definition images over the network is avoided, ensuring real-time dynamic monitoring of the detection process.

5. The method for full-process management of testing laboratory equipment according to claim 4, characterized in that, In step 2, synchronizing the inference results and environmental parameters to the cloud via a stream processing platform includes: using a compression algorithm to encode image feature descriptors and sensor data streams by establishing a global dictionary; The data stream is divided into a metadata layer and a raw sampling layer. Only the encoded feature vectors are uploaded to the cloud, reducing bandwidth consumption while maintaining data integrity. The cloud-based stream processing engine performs complex event processing on pressure data, displacement data, image feature data, and environmental temperature and humidity data that converge from multiple edge nodes within a preset sliding time window. Based on multidimensional vector aggregation analysis criteria, the stream processing engine determines the failure of the specimen when the pressure change rate exceeds a preset threshold and the duration on the time axis reaches a predetermined value, combined with the specimen surface features identified by image recognition, and sends an emergency stop command to the actuator.

6. The method for full-process management of testing laboratory equipment according to claim 5, characterized in that, In step 3, establishing a digital passport system for the specimens includes: at the specimen molding site, using a handheld terminal to write the specimen's mix ratio information, molding time, geographical location information, and construction team information into an RFID tag; An asymmetric encryption algorithm is used to sign the initial block data with a private key to ensure that the initial block information cannot be tampered with; Radio frequency identification (RFID) readers are installed in the constant temperature and humidity curing room, the mechanical laboratory, and the sample retention room. When the automated guided vehicle carrying the specimen passes through the RFID access control, the reader reads the unique identification code within a preset time and encapsulates the current position coordinates and the real-time temperature and humidity parameters collected by the environmental sensors into a data transfer data package.

7. The method for full-process management of testing laboratory equipment according to claim 6, characterized in that, In step 3, achieving full lifecycle data storage also includes: uploading the transfer data packet to the sample transfer smart contract through an encrypted channel; The smart contract automatically determines whether the specimen meets the admission conditions for entering the next testing stage based on preset logical conditions. The logical conditions include whether the curing age of the specimen has reached the predetermined number of days and whether the cumulative temperature of curing under the same conditions has reached the predetermined threshold. When all logical conditions are met, the system sends a grasping permission instruction to the robot control system and records the current flow action, operator information, and timestamp in the blockchain sidechain. The data-driven process permission mechanism eliminates detection errors caused by human intervention.

8. The method for full-process management of testing laboratory equipment according to claim 7, characterized in that, In step 4, configuring the execution logic includes: using the visual logic modeling interface provided by the graphical process chain orchestrator, constructing a detection flowchart that includes a loading module, a pressure holding module, a photo evidence collection module, and an unloading module by dragging and dropping functional components; The orchestrator automatically converts the detection flowchart into an executable script that conforms to the linked data format. The executable script defines the execution parameters, logical decision branches, and exception handling mechanisms for each step. The edge computing node parses the executable script and establishes bidirectional communication with the specific testing equipment through industrial automation standard protocols, converting the pressurization step size and holding time parameters in the script into process data object mapping tables for controlling solenoid valves and servo motors.

9. The method for full-process management of testing laboratory equipment according to claim 8, characterized in that, In step 5, implementing closed-loop control and result backtracking management includes: in the steel bar tensile test, when the edge computing node identifies abnormal signs of fracture shape through real-time image analysis, the cloud decision model calculates the corrected displacement rate and sends the corrected parameters to the device controller through a high-speed communication link; The controller adjusts the output pulse frequency of the motor driver in real time according to the correction parameters to achieve real-time dynamic compensation for the physical execution process; After the test is completed, a hash algorithm is used to calculate the digest of the original pressure curve, displacement sampling data and key frame video of the test process. The generated hash digest and the distributed storage path of the original data are written into the subsequent blocks of the blockchain. The digest in the report is compared and verified by the public key to ensure that the data throughout the process has not been tampered with.

10. The method for full-process management of testing laboratory equipment according to claim 9, characterized in that, The method is applied to intelligent testing scenarios for concrete impermeability, and includes the following refined features: This is achieved by configuring an intelligent anti-seepage cabinet, which is equipped with multiple independent sealing units inside. Each sealing unit is a test specimen barrel formed by one-time die casting. The inner wall of the specimen barrel is embedded with a composite material airbag made of reinforced nylon layer and natural rubber laminate; A flexible seal is achieved on the side of the specimen by filling the airbag with compressed air at a preset pressure. The pressure during the sealing process is monitored in real time by a high-precision pressure sensor with a preset resolution. The intelligent anti-seepage cabinet is equipped with a lifting mechanism driven by a servo motor and a ball screw, and its repeatability positioning accuracy reaches the preset micron level standard, so as to cooperate with the pneumatic gripper with pressure and displacement dual sensors integrated at the end of the composite robot for loading and unloading operations. The system uses the slope change criterion of the leakage pressure versus time curve. When the pressure value collected by the independent pressure sensor decreases by more than a predetermined threshold within a preset time step, and the corresponding slope change rate satisfies the preset logical decreasing relationship, it is determined that the specimen has leaked and triggers the industrial camera to capture the water seepage image at the bottom of the specimen. The image recognition algorithm is used to confirm the location distribution of the water seepage point and update it in the specimen digital passport.

Images