Device and Method for Intelligent Processing of Layered Interfaces and In-situ Performance Evaluation of Fillers

Abstract

This invention provides a device and method for intelligent processing and in-situ performance evaluation of the interlayer interface of filling materials. The device includes: a specimen state sensing unit for monitoring the solidification state of the lower layer filling material; an intelligent processing robotic arm for performing interlayer interface processing operations; an interlayer reinforcement unit for sequentially performing laser etching, ultrasonic cleaning, and interface agent spraying on the surface of the lower layer filling material; a non-destructive testing unit for performing acoustic-electrical combined non-destructive testing on the upper layer filling material after casting; a central processing unit connected to the specimen state sensing unit and the intelligent processing robotic arm for determining the solidification state of the lower layer filling material and controlling the start of the intelligent processing robotic arm; and a connection to an imaging inversion workstation for generating an interlayer bonding quality assessment report. The beneficial effects of this invention are: achieving intelligent and precise processing, non-destructive automated processing, and in-situ non-destructive evaluation of the interlayer bonding quality of filling materials, realizing closed-loop quality control throughout the entire preparation process.

Description

Technical Field

[0001] This invention belongs to the field of laboratory infill material layered casting technology, and in particular relates to a device and method for intelligent treatment of in-situ performance evaluation of infill material layered interface. Background Technology

[0002] In practical engineering projects such as backfill mining, large-scale backfill bodies commonly employ layered and segmented casting processes due to operational continuity, production capacity limitations, and heat dissipation requirements. However, this process inevitably introduces a fatal weakness in the engineering field: weak interfaces between layers (commonly known as "cold joints"). These hidden defects existing within the backfill body significantly weaken the overall structural strength and stability of the backfill, potentially leading to major safety hazards such as mine collapse due to ineffective support of the surrounding rock, and also causing direct economic risks such as support failure and decreased ore recovery rates. Due to the complex on-site environment, uncontrollable variables, and extremely high monitoring costs, it is difficult to effectively control and accurately assess the quality of interlayer bonding. Therefore, conducting systematic research on layered backfill bodies in the laboratory has become a core prerequisite and an essential path to solving this engineering challenge. In the laboratory, by constructing precise and controllable physical models, the aim is to reproduce and deeply analyze the microscopic mechanisms of interlayer formation, thereby optimizing material ratios, accurately determining pouring sequence, and developing efficient new interface treatment processes and equipment (such as intelligent roughening and uniform coating). The ultimate goal is to completely eliminate the cold joint effect and ensure that each layer of filling material can be firmly bonded into a unified whole. This will provide fully validated, reliable, and economical optimization solutions for on-site construction, fundamentally ensuring the safety and economy of underground engineering, and promoting the intelligent, standardized, and high-quality development of mining engineering.

[0003] However, in the laboratory preparation of layered specimens (i.e., filling bodies, including upper and lower layers with an interlayer interface), numerous technical bottlenecks severely restrict the interlayer bonding quality and the reliability of experimental data. First, relying on manual experience to judge the initial setting time of the lower layer is significantly subjective and lag-dependent, easily missing the optimal casting window, leading to poor interlayer interface bonding or even the formation of cold cracks that are difficult to heal. Second, the application of interface agents is mostly done manually, resulting in uneven coating thickness and large fluctuations in coverage, leading to high dispersion in interlayer bond strength and poor experimental repeatability. Finally, existing technologies lack effective in-situ, non-destructive interlayer performance testing methods, necessitating destructive sampling for post-assessment, which is not only inefficient but also fails to achieve real-time quality feedback and optimization during the preparation process. These systemic defects collectively make the interlayer interface a weak link in the filling body's performance; therefore, an intelligent treatment system for the layered casting interface of filling bodies and an in-situ performance evaluation method are urgently needed. Summary of the Invention

[0004] In view of this, the present invention aims to overcome the shortcomings of the above-mentioned problems in the prior art and proposes a device and method for intelligent processing of the layered interface of filling body and in-situ performance evaluation.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows: The first aspect of this invention provides a device for intelligent processing of the layered interface of filling materials and in-situ performance evaluation, comprising: The specimen state sensing unit includes a micro piezoelectric sensor array, an infrared surface moisture monitor, and an environmental monitoring module, which is used to monitor the solidification state of the lower filling material poured into the specimen mold in real time. Intelligent robotic arms are used to perform processing tasks at interlayer interfaces; The interlayer reinforcement unit includes a laser etching head, an ultrasonic negative pressure cleaning head, and an ultrasonic atomizing nozzle, which are connected to an intelligent processing robotic arm to sequentially perform laser etching, ultrasonic cleaning, and interface agent spraying on the surface of the lower filling material. The non-destructive testing unit includes an ultrasonic tomography array, a resistance tomography electrode array, a multi-channel data acquisition instrument, and an imaging inversion workstation. The ultrasonic tomography array and the resistance tomography electrode array are connected to the input end of the multi-channel data acquisition instrument, and the imaging inversion workstation is connected to the output end of the multi-channel data acquisition instrument. It is used to perform acoustic and electrical combined non-destructive testing on the filling body after the upper layer is poured. The central processing unit (CPU) has a built-in prediction model based on a long short-term memory network and is connected to the specimen state sensing unit and the intelligent processing robotic arm. It is used to receive signals collected by the specimen state sensing unit, determine the solidification state of the lower filling material, and generate trigger commands to control the intelligent processing robotic arm to start. It is also connected to the imaging inversion workstation to generate an interlayer bonding quality assessment report.

[0006] Furthermore, the micro piezoelectric sensor array is embedded in the side wall of the specimen mold, and the detection end face of each micro piezoelectric sensor in the micro piezoelectric sensor array is flush with the inner wall of the specimen mold.

[0007] Furthermore, in the non-destructive testing unit, the ultrasonic tomography array is used to construct the acoustic wave transmission field and includes multiple high-frequency ultrasonic transducers distributed in a rectangular array. The high-frequency ultrasonic transducers are embedded in the opposite side walls of the specimen mold and are flush with the inner wall of the specimen mold. The resistivity tomography electrode array is used to construct the current excitation field and the potential measurement field to capture dielectric defects at the interlayer interface. The resistivity tomography electrode array includes multiple electrodes that are uniformly distributed along the circumferential direction of the inner wall of the specimen mold. A multi-channel data acquisition instrument is used to acquire acoustic signals from an ultrasonic tomography array and electrical signals from an electrical resistance tomography electrode array. The imaging inversion workstation is used for data processing and imaging. It is equipped with an acoustic-electric joint inversion algorithm, receives acoustic and electrical signals from a multi-channel data acquisition instrument, and generates a three-dimensional fused image of internal defects of the filling body through synchronous iterative inversion. It outputs resistivity and sound velocity data, providing core data support for the central processing unit to calculate the interface bonding coefficient and determine the interlayer bonding quality.

[0008] A second aspect of the present invention provides a method for intelligent processing and in-situ performance evaluation of the layered interface of filling materials, implemented by the aforementioned intelligent processing and in-situ performance evaluation device for the layered interface of filling materials, comprising the following steps: S1. Multi-dimensional state perception and initial setting prediction: The shear wave velocity change is monitored in real time by a micro piezoelectric sensor array, the surface humidity is scanned in real time by an infrared surface moisture monitor, and the multi-modal data is input into a prediction model based on a long short-term memory network to determine when the lower filling material reaches the initial setting state. If the lower filling material is determined to have reached the initial setting state, a trigger command is generated by the central processing unit. S2. Automated inter-layer interface processing: After receiving a trigger command, the intelligent processing robotic arm executes the following steps sequentially. S21. Laser micro-etching process: The surface slurry is removed by a laser etching head to form a micro-rough interface; S22. Ultrasonic negative pressure coupling cleaning: The ultrasonic negative pressure cleaning head loosens the dust embedded in the pores of the interlayer interface and simultaneously captures and removes it through the negative pressure airflow field. S23. Ultrasonic atomization spraying: the interface agent is atomized into micron-sized droplets through an ultrasonic atomization nozzle and uniformly deposited on the interlayer interface after step S22 to form a modified film with controllable thickness. S3. Time-dependent upper layer casting: Based on the curing characteristic curve of the interface agent and the ambient temperature and humidity, the optimal bonding activity window period is obtained. Within the optimal bonding activity window period, when the interface agent is in a touch-dry state, the upper layer filling is cast. The optimal bonding activity window period is the theoretical time range within which the interface agent can form an effective chemical bond with the upper layer filling. The touch-dry state is the state in which the interface agent forms a film on the surface but the molecular chains are still active during the process of changing from liquid to solid. The curing characteristic curve is the relationship curve between environmental parameters and the corresponding touch-dry time threshold. S4. Multiphysics field non-destructive in-situ evaluation: After the upper layer is poured, the acoustic signal is scanned by an ultrasonic tomography array, the electrical signal is scanned by a resistance tomography electrode array, and the acoustic and electrical signals are acquired by a multi-channel data acquisition instrument. The interlayer bonding quality assessment report is generated by data fusion and cross-gradient constraint inversion through an imaging inversion workstation, thus realizing closed-loop quality control of the preparation process.

[0009] Furthermore, in step S1, after the multimodal data is standardized, the Bayesian optimization algorithm is used to optimize hyperparameters and extract key features. The multimodal data includes shear wave velocity, surface humidity data, ambient temperature and humidity data, and timestamp data. The extracted key features are input into a prediction model based on a long short-term memory network, which outputs the predicted value of the current filling's settling degree and compares it with a preset optimal interface processing threshold. If the error between the predicted value of the current filling's settling degree and the preset optimal interface processing threshold is less than the preset error value, it is determined that the initial settling state has been reached. The central processing unit generates a trigger command in milliseconds and sends it to the control system of the intelligent processing robotic arm to start the subsequent interlayer interface processing process.

[0010] Furthermore, the process parameters for laser micro-etching in step S21 include power, pulse frequency, spot diameter, and scanning speed, and the robotic arm automatically calibrates the working height according to the preset path.

[0011] Furthermore, in step S3, the upper filling body is poured by oblique injection, controlling the material drop point to deviate from the center of the interlayer interface, and the interface contact is completed by flowing and covering.

[0012] Further, the specific process of step S4 is as follows: the ultrasonic tomography array acquires acoustic signals, the electrical resistance tomography electrode array acquires electrical signals, and the multi-channel data acquisition instrument simultaneously acquires acoustic and electrical signals and transmits them to the imaging inversion workstation; the imaging inversion workstation reconstructs the sound velocity distribution through the acoustic signals and the resistivity distribution through the electrical signals, and synchronously iterates and updates the sound velocity distribution and resistivity distribution based on the acoustic-electric joint inversion algorithm with cross-gradient constraints, generating a three-dimensional sound velocity matrix and a three-dimensional resistivity matrix and transmitting them to the central processing unit; the central processing unit takes the sound velocity value and resistivity value of each voxel unit under the same spatial coordinates as input, calculates the comprehensive quality index of each voxel unit, and simultaneously constructs a two-parameter fusion feature vector input to the support vector machine classifier to output physical attribute labels; the central processing unit calculates the interface bonding coefficient based on the comprehensive quality index and generates an interlayer bonding quality assessment report; if the interface bonding coefficient is lower than a preset threshold, the specimen is determined to be unqualified, and the defect coordinates are marked.

[0013] Compared with the prior art, the present invention has the following advantages: The intelligent processing and in-situ performance evaluation device for interlayer interfaces of filling materials described in this invention achieves intelligent and precise processing of the interlayer interfaces of filling materials. It monitors the setting state of the lower layer of filling material in real time through a specimen state sensing unit, and determines the optimal timing for interface processing based on a prediction model using a long short-term memory network. This solves the problems of traditional manual judgment being highly subjective, lagging, and prone to missing the optimal pouring window, thus realizing a shift from experience-based judgment to data-driven intelligent decision-making. It also achieves non-destructive automated processing of the interlayer interfaces. An intelligent processing robotic arm equipped with an interlayer reinforcement unit sequentially completes laser micro-etching, ultrasonic cleaning, and interface agent spraying, replacing traditional mechanical roughening and avoiding micro-cracks in the specimens during the initial setting period. This method effectively prevents damage to the filling material while ensuring interface cleanliness and uniformity of the interface agent coating. It enables in-situ non-destructive evaluation of interlayer bonding quality by utilizing the ultrasonic tomography array and electrical resistance tomography electrode array of the non-destructive testing unit, along with the acoustic-electrical joint inversion algorithm of the imaging inversion workstation. This generates a three-dimensional fused image of internal defects in the filling material, achieving real-time quality feedback and closed-loop control during the preparation process. This changes the traditional post-evaluation method that relies on destructive sampling. It achieves closed-loop quality control throughout the entire preparation process, with each unit working collaboratively to form a complete closed loop of perception, processing, and evaluation. This ensures that the interlayer bonding quality of the filling material is controllable and repeatable, providing reliable data support for laboratory research. Attached Figure Description

[0014] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the structure of the intelligent processing and in-situ performance evaluation device for the layered interface of the filling body described in Embodiment 1 of the present invention after the lower layer of filling body has been poured and is waiting for initial setting. Figure 2 This is a schematic diagram of the distribution structure of the non-destructive testing unit according to Embodiment 1 of the present invention; Figure 3 This is a flowchart of the intelligent processing and in-situ performance evaluation method for the layered interface of the filling material according to Embodiment 2 of the present invention; Figure 4 This is a flowchart illustrating the intelligent processing and in-situ performance evaluation method for the layered interface of the filling material as described in Embodiment 2 of the present invention.

[0015] Explanation of reference numerals in the attached figures: 1. Specimen mold; 2. Lower filling material; 3. Shear wave signal; 4. Miniature piezoelectric sensor array; 5. Environmental monitoring module; 6. Central processing unit; 7. Intelligent processing robotic arm; 8. Infrared surface moisture monitor; 9. Ultrasonic tomography array; 10. Resistance tomography electrode array; 11. Internal defects of the filling material; 12. Acoustic wave transmission field; 13. Current excitation field. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0017] In the description of this invention, it should be understood that these descriptions are merely exemplary and not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0018] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0019] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0020] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] Example 1 like Figure 1 and Figure 2 As shown, the intelligent processing and in-situ performance evaluation device for the layered interface of the filling material includes: The specimen state sensing unit includes a miniature piezoelectric sensor array 4, an infrared surface moisture monitor 8, and an environmental monitoring module 5. It is used to monitor the solidification state of the lower filling material 2 poured into the specimen mold 1 in real time. The miniature piezoelectric sensor array is used to transmit and receive high-frequency shear wave signals 3, characterizing the evolution of the material's internal rheological properties by monitoring changes in shear wave velocity. The infrared surface moisture monitor is positioned above the filling material to scan and quantify the water evaporation rate and surface humidity data on the filling material surface in real time. The environmental monitoring module collects ambient temperature and humidity data as correction factors. The intelligent processing robotic arm 7 includes a six-degree-of-freedom robotic arm body and an adapter at the end of the robotic arm body, used to perform interlayer interface processing operations. The adapter enables the switching between a laser etching head, an ultrasonic negative pressure cleaning head, and an ultrasonic atomizing nozzle. The interlayer reinforcement unit includes a laser etching head, an ultrasonic negative pressure cleaning head, and an ultrasonic atomizing nozzle, which are respectively connected to the intelligent processing robotic arm and connected to the robotic arm body through an adapter. The lower filling body surface is sequentially treated with laser etching, ultrasonic cleaning, and interface agent spraying. The non-destructive testing unit includes an ultrasonic tomography array 9, a resistance tomography electrode array 10, a multi-channel data acquisition instrument, and an imaging inversion workstation. The ultrasonic tomography array and the resistance tomography electrode array are connected to the input end of the multi-channel data acquisition instrument, and the imaging inversion workstation is connected to the output end of the multi-channel data acquisition instrument. It is used to perform acoustic and electrical combined non-destructive testing on the filling body (i.e., the entire filling body) after the upper layer is poured. The central processing unit 6 has a built-in prediction model based on long short-term memory network and is connected to the specimen state sensing unit and the intelligent processing robotic arm. It is used to receive signals collected by the specimen state sensing unit, determine the solidification state of the lower filling material, and generate trigger commands to control the intelligent processing robotic arm to start. It is also connected to the imaging inversion workstation to generate an interlayer bonding quality assessment report.

[0022] A miniature piezoelectric sensor array is embedded within the sidewall of the specimen mold, with the detection end face of each miniature piezoelectric sensor flush with the inner wall of the mold to ensure contact with the specimen. A sealing and fixation between the miniature piezoelectric sensors and the mold is achieved through an acoustic coupling agent layer and a flexible sealing ring. In this embodiment, the miniature piezoelectric sensor array comprises multiple miniature piezoelectric sensors arranged in a rectangular array, distributed on opposite sidewalls of the specimen mold.

[0023] The laser etching head includes a high-frequency pulsed laser generator and an optical focusing lens, used for non-contact removal of surface slurry through photothermal effects; the ultrasonic negative pressure cleaning head includes a retractable ultrasonic amplitude transformer and a negative pressure dust extraction interface, used for integrated cleaning of loosening and dust extraction; the ultrasonic atomizing nozzle includes a miniature precision metering pump and an ultrasonic atomizing nozzle, used for quantitative atomization deposition of the interface agent. In this embodiment, the laser etching head, ultrasonic negative pressure cleaning head, and ultrasonic atomizing nozzle are all conventional equipment in the art.

[0024] In the non-destructive testing unit, the ultrasonic tomography array is used to construct the acoustic wave transmission field 12, and includes multiple high-frequency ultrasonic transducers distributed in a rectangular array. The high-frequency ultrasonic transducers are embedded in the opposite side walls of the specimen mold and are flush with the inner wall of the specimen mold to ensure contact with the specimen. After the filling body is poured, the inner end face of the high-frequency ultrasonic transducer will directly contact the filling body to ensure that the acoustic wave signal can effectively penetrate the filling body and accurately capture the acoustic characteristics of the interlayer interface. The resistivity tomography electrode array is used to construct a current excitation field 13 and a potential measurement field to capture dielectric defects at the interlayer interface, providing key electrical data for evaluating the quality of interlayer bonding. The resistivity tomography electrode array includes multiple electrodes uniformly distributed circumferentially along the inner wall of the specimen mold. Specifically, the resistivity tomography electrode array injects alternating current into the specimen to form a current excitation field. When there are defects inside the specimen, there will be differences in conductivity. By capturing the voltage distribution generated on each electrode and calculating the voltage difference, the location of high resistance can be deduced, that is, the dielectric defects at the interlayer interface can be captured. A multi-channel data acquisition instrument is used to acquire acoustic signals from an ultrasonic tomography array and electrical signals from an electrical resistance tomography electrode array. The imaging inversion workstation is used for data processing and imaging. It is equipped with an acoustic-electric joint inversion algorithm, receives acoustic and electrical signals from a multi-channel data acquisition instrument, and generates a three-dimensional fused image of the internal defect 11 of the filling body through synchronous iterative inversion. It outputs resistivity and sound velocity data, providing core data support for the central processor to calculate the interface bonding coefficient and determine the interlayer bonding quality.

[0025] In this embodiment, the miniature piezoelectric sensor is a PZT-5A type piezoelectric ceramic sheet, the multi-channel data acquisition instrument is a PXIe-5170 data acquisition module, and the imaging inversion workstation is a GD-S01HF ultrasonic CT imaging inversion workstation.

[0026] Example 2 like Figure 3 and Figure 4 As shown, Figure 4 for Figure 3 The specific operation of the Chinese method process, the intelligent processing and in-situ performance evaluation method for the layered interface of the filling material, is implemented through the intelligent processing and in-situ performance evaluation device for the layered interface of the filling material described in Example 1, and includes the following steps: S1. Multi-dimensional state perception and initial setting prediction: The shear wave velocity change is monitored in real time by a micro piezoelectric sensor array, the surface humidity is scanned in real time by an infrared surface moisture monitor, and the multi-modal data is input into a prediction model based on a long short-term memory network to determine when the lower filling material reaches the initial setting state. If the lower filling material is determined to have reached the initial setting state, a trigger command is generated by the central processing unit. S2. Automated inter-layer interface processing: After receiving a trigger command, the intelligent processing robotic arm executes the following steps sequentially. S21. Laser micro-etching process: The surface slurry is removed non-contactly by a high-frequency pulsed laser beam generated by a laser etching head, forming a micro-rough interface; where high frequency refers to the pulse repetition frequency of the laser is 1kHz-100kHz; the etching depth of the laser etching head on the interlayer interface is 0.1mm-1mm to form a micro-rough interface, 0.1mm is to penetrate the weak slurry layer, and 1mm is to prevent excessive damage and cracks. S22. Ultrasonic negative pressure coupling cleaning: The ultrasonic negative pressure cleaning head generates an ultrasonic vibration field and a negative pressure airflow field to loosen the dust embedded in the pores of the interlayer interface (i.e., the micro-rough interface) and simultaneously capture and remove it. S23. Ultrasonic atomization spraying: The interface agent is atomized into micron-sized droplets through an ultrasonic atomizing nozzle and uniformly deposited on the interlayer interface after step S22 to form a modified film with controllable thickness. The thickness of the modified film is 10μm-50μm. If it is less than 10μm, a continuous physical coating may not be formed. If it is greater than 50μm, the problem of cracking due to stress shrinkage needs to be considered. S3. Time-sensitive upper layer pouring: Based on the curing characteristic curve of the interface agent and the ambient temperature and humidity, the optimal bonding activity window period is obtained. Within the optimal bonding activity window period, when the interface agent is in a touch-dry state, the upper layer filling is poured. The optimal bonding activity window period is the theoretical time range within which the interface agent can form an effective chemical bond with the upper layer filling. It is equivalent to a pre-constraint, which aims to eliminate the interference of false alarms from the infrared surface moisture monitor and the identification of individual outstanding dryness but overall incompleteness, thus saving time for advance scheduling of subsequent preparations. The touch-dry state is the state in which the interface agent forms a film on the surface but the molecular chains still have high activity during the process of changing from liquid to solid. It is the only optimal pouring time within the optimal bonding activity window period. The curing characteristic curve is a preset curve accumulated from experiments. Each interface agent corresponds to a specific curing characteristic curve, which is the relationship curve between ambient temperature and humidity and the corresponding touch-dry time threshold. S4. Multiphysics field non-destructive in-situ evaluation: After the upper layer is poured, the acoustic signal is scanned by an ultrasonic tomography array, the electrical signal is scanned by a resistance tomography electrode array, and the acoustic and electrical signals are acquired by a multi-channel data acquisition instrument. The interlayer bonding quality assessment report is generated by data fusion and cross-gradient constraint inversion through an imaging inversion workstation, thus realizing closed-loop quality control of the preparation process.

[0027] In this embodiment, after the layered casting process is initiated, the specimen state sensing unit immediately enters the working state. First, active acoustic detection is activated, and a micro piezoelectric sensor array emits a weak shear wave signal to the newly cast filling material at a preset high sampling frequency. At this time, by measuring the propagation time (ToF) of the shear wave between the transmitting and receiving ends, and combining this with the known spacing (L) between the micro piezoelectric sensors, the real-time shear wave velocity (Vs) is first calculated using the formula Vs = L / ToF. The shear wave velocity has the following relationship with the shear modulus (G) of the medium and the density (ρ) of the filling material. , The propagation speed of shear waves varies at different stages of the solidification of the filling material, so the propagation time parameter can directly reflect the internal physical and mechanical state. Simultaneously, surface condition scanning is performed. The infrared surface moisture monitor analyzes the absorption spectrum of a specific infrared band to accurately measure the surface moisture evaporation rate and surface humidity gradient, reflecting the surface hardening trend.

[0028] In step S1, after standardization, the multimodal data is optimized using a Bayesian algorithm for hyperparameter optimization and key feature extraction. Key features include abrupt changes in wave velocity growth rate and inflection points in surface water loss, representing the most sensitive combination of key features to condensation states. The multimodal data includes shear wave velocity, surface humidity data, ambient temperature and humidity data, and timestamp data. In this embodiment, temperature and humidity data are collected using an environmental monitoring probe. All raw multimodal data are filtered and denoised, then mapped to the same dimension range to eliminate model training bias caused by different physical magnitudes. The central processing unit calculates the wave velocity growth rate by taking the first derivative of the shear wave velocity; the extreme values ​​of the wave velocity growth rate correspond to abrupt changes in the wave velocity growth rate. The surface humidity data is calculated by taking the first derivative with respect to time to obtain the water loss rate, and the second derivative is used to obtain the acceleration of the water loss rate. When the second derivative is 0, it indicates a sharp slowdown in the water loss rate, indicating the inflection point of surface water loss has been reached.

[0029] In step S1, the extracted key features are input into the prediction model based on the long short-term memory network, and the predicted value of the current filling material's coagulation degree is output and compared with the preset optimal interface processing threshold. If the error between the predicted value of the current filling material's coagulation degree and the preset optimal interface processing threshold is less than the preset error value, it is determined that the initial coagulation state has been reached. The central processing unit generates a trigger command in milliseconds and sends it directly to the control system of the intelligent processing robotic arm to start the subsequent interlayer interface processing process.

[0030] The process parameters for laser micro-etching in step S21 include power, pulse frequency, spot diameter, and scanning speed. The robotic arm automatically calibrates the working height according to a preset path to ensure that the focal length is within the optimal etching range. In this embodiment, after receiving a trigger command from the central processing unit, the intelligent processing robotic arm immediately wakes up from the standby position and calls the pre-stored 3D CAD data of the specimen mold. Combined with the current process parameters, it automatically generates the optimal working path covering the surface of the lower filling material, which is usually a "bow" shaped path with a high overlap rate or a grating scanning path. The robotic arm body carries the laser etching head to directly above the starting point of the operation and automatically calibrates the working height using a laser rangefinder to ensure that the focal length is always within the optimal etching range. According to the preset hardness and target roughness of the filling material, the corresponding laser process parameter package (power P, pulse frequency f, spot diameter D, scanning speed v) is retrieved. The laser etching head is activated to emit a high-frequency pulsed laser beam. The laser beam energy is highly concentrated, and the laser angle can be adjusted by a mirror to irradiate the surface of the filling material. The cement paste enriched on the surface of the filling body (weak interface layer) absorbs laser energy, causing a rapid temperature increase, instantaneous vaporization, and a sharp expansion and peeling of its volume; while the underlying coarse aggregate is retained due to differences in thermophysical properties. The entire process is free of mechanical contact and impact vibration, avoiding the microcrack damage to the internal structure of the specimen during the initial setting period caused by traditional mechanical roughening, while simultaneously constructing an ideal bonding surface with uniform depth and controllable micro-roughness.

[0031] In step S22, the robotic arm unloads the laser etching head via an adapter and attaches an ultrasonic negative pressure cleaning head. To prevent damage to the sample from the ultrasonic amplitude transformer, the ultrasonic negative pressure cleaning head adopts a non-contact, suspended cleaning design, with the ultrasonic amplitude transformer retracted 5mm into the cleaning head shell, ensuring the sample remains undamaged even in case of operational errors. As the cleaning head traverses the work surface, the internal piezoelectric vibrator generates an ultrasonic vibration field of tens of thousands of hertz. This vibrational energy is transmitted to the micropores at the interlayer interface, overcoming the van der Waals forces and electrostatic adsorption forces between fine dust and the substrate, causing deep-seated dust to suspend and loosen. Simultaneously, a high-power micro vacuum pump is activated, creating a strong negative pressure airflow field at the suction port, instantly drawing the debris and dust loosened by the sound waves into the dust collection box. Through the dual mechanism of ultrasonic desorption and negative pressure capture, the effect of removing dust embedded in micropores is achieved, ensuring a clean interface.

[0032] In step S23, if the experimental design requires interface enhancement, the robotic arm switches to an ultrasonic atomizing nozzle via an adapter and connects to a constant-temperature feeding pipeline. The micro-precision pump and ultrasonic generator are activated. As the liquid interface agent flows through the ultrasonic atomizing nozzle, it is dispersed into micron-sized droplets with extremely narrow diameters by high-frequency oscillation. The robotic arm scans at a constant speed and height. The atomized droplets settle uniformly on the rough interface, forming a continuous, dense, and controllable-thickness micron-sized modified film. This avoids the localized liquid accumulation and missed coating phenomena caused by manual brushing.

[0033] Then, the intelligent processing robotic arm controls the robotic arm body to automatically return to the safe stopping position and sends a signal to the central controller indicating that the processing is complete, preparing to enter the next pouring stage.

[0034] In step S3, the upper filling material is poured using an oblique grouting method, controlling the material's landing point to deviate from the center of the interlayer interface. Interface contact is achieved through flowing coverage, and low-frequency micro-vibration is used to prevent material impact from damaging the interface layer. In this embodiment, the oblique grouting method is widely used in fields such as mine filling and concrete engineering, and is a conventional technique. No improvements have been made to it here, so its specific operation process will not be described in detail.

[0035] In step S3, the curing characteristic curve corresponding to the interface agent type sprayed in step S2 is retrieved from the database and stored. Using the curing characteristic curve and considering the current ambient temperature and humidity, the optimal bonding activity window for the interface agent is determined. The interlayer interface state is continuously monitored, and a pouring permission signal is issued when the interface agent reaches a touch-dry state. The upper filling material is poured immediately within the optimal bonding activity window.

[0036] The specific process of multiphysics non-destructive in-situ evaluation in step S4 is as follows: The high-frequency ultrasonic transducers of the ultrasonic tomography array operate in turn according to a preset sequence. A multi-channel data acquisition instrument acquires acoustic signals passing through the interface, extracting acoustic duration, amplitude attenuation coefficient, and dominant frequency drift as acoustic signals to capture mechanical defects. Prolonged acoustic duration or a sharp drop in amplitude indicates the presence of air gaps, voids, or cracks at the interface. The resistive tomography electrode array adopts an adjacent excitation-adjacent measurement mode, injecting a microampere-level safe AC current into the specimen. The multi-channel data acquisition instrument synchronously measures the boundary voltage distribution as an electrical signal to capture dielectric defects. In this embodiment, the preset sequence includes sector scanning or cross scanning. The imaging inversion workstation receives acoustic and electrical signals. It uses a travel-time imaging algorithm to reconstruct the sound velocity distribution model using acoustic signals and a sensitivity matrix and back-projection algorithm to reconstruct the resistivity distribution cloud map using electrical signals. The acoustic-electric joint inversion algorithm based on cross-gradient constraints introduces a cross-gradient function as a regularization term in the objective function, and the gradient direction of the sound velocity distribution model and the gradient direction of the resistivity distribution cloud map are consistent in space. By minimizing the objective function, which includes data residual terms, model smoothing terms, and cross-gradient coupling terms, the synchronous iterative update of the sound velocity distribution and resistivity distribution is achieved, generating a three-dimensional sound velocity matrix and a three-dimensional resistivity matrix. The three-dimensional sound velocity matrix and the three-dimensional resistivity matrix are then fused at the pixel level in the same spatial coordinate system to generate a three-dimensional fused image of the internal defect 11 of the filling body. The imaging inversion workstation transmits the obtained three-dimensional sound velocity matrix and three-dimensional resistivity matrix to the central processing unit (CPU). The CPU takes the sound velocity and resistivity values ​​of each voxel unit in the same spatial coordinate system as input, and calculates a comprehensive quality index using preset weights. This index serves as the basis for calculating the interface bonding coefficient. Simultaneously, the sound velocity and resistivity values ​​are normalized to construct a two-parameter fusion feature vector, which is input into a support vector machine classifier to automatically output physical attribute labels, comprehensively determining the nature of the defects. The CPU calculates the interface bonding coefficient based on the comprehensive quality index and generates an interlayer bonding quality assessment report. If the interface bonding coefficient is lower than a preset threshold, the specimen is deemed unqualified, and the defect coordinates are marked. In this embodiment, the pixel-level image fusion algorithm fuses the three-dimensional sound velocity matrix and the three-dimensional resistivity matrix in the same spatial coordinate system, ensuring spatial consistency of the dual-modal inversion results based on cross-gradient constraints. Physical attribute labels include voids, water-rich, and dense. Defect coordinates are marked in three-dimensional coordinate form. The imaging inversion workstation feeds back the quality assessment results to the central processor, enabling closed-loop quality control of the preparation process.

[0037] The process by which the central processing unit (CPU) calculates the interface combination coefficients based on the composite quality index is as follows: Under the same spatial coordinate system, the CPU takes the normalized sound velocity and resistivity values ​​as inputs and calculates the composite quality index Q of a single voxel unit i using preset weights. i , , in V i Let be the normalized speed of sound value of the i-th voxel. ρ i Let be the normalized resistivity value of the i-th voxel; and These are preset weighting coefficients for sound velocity and resistivity, respectively, and satisfy... ; After obtaining the comprehensive quality index of all voxel elements within the interlayer interface region, the system calculates the overall statistical mean of the region to derive the interfacial bonding coefficient, which reflects the macroscopic bonding strength of the specimen. C , , in N This represents the total number of voxel units extracted from the interlayer interface region.

[0038] The following is a specific example to illustrate Embodiment 1 and Embodiment 2.

[0039] The miniature piezoelectric sensor array uses PZT-5A type piezoelectric ceramic sheets with a diameter of 5mm and a center frequency of 500kHz. These sheets are embedded at a specific depth (100mm × 300mm) into two opposing inner walls of the specimen mold using a 20mm × 20mm grid, ensuring the sensing surface is flush with the inner wall of the mold to avoid stress concentration points. The first row of miniature piezoelectric sensors is positioned 50mm above the bottom of the mold, with a total of 3 rows and 5 columns, comprising 15 miniature piezoelectric sensors forming a sensor network. A multi-channel data acquisition instrument (sampling rate ≥10MS / s) synchronously controls the excitation and reception of all miniature piezoelectric sensors to ensure data timing consistency. The infrared surface moisture meter uses a near-infrared (NIR) spectrometer based on an InGaAs (indium gallium arsenide) detector, measuring in the 1450-1550nm wavelength range, which is highly sensitive to moisture absorption. It is connected to an adapter on the robotic arm and moves with the arm. During each measurement, the robotic arm drives the infrared surface moisture meter to pause at a height of 100mm above the surface of the filling material for precise point-to-point measurement, avoiding ambient light interference, and the diameter of the measurement spot is ≤3mm.

[0040] The experimental object is a mold with a size of 100mm×100mm×300mm, with pre-embedded components for the specimen status sensing unit and non-destructive testing unit inside; the filling material is a tailings cemented filling material with a mass ratio of 1:4 of ash and sand, and a mass concentration of 70%.

[0041] The prepared filling slurry was poured into the specimen mold to a height of 150 mm and then slightly vibrated. The central processing unit was started and loaded with the prediction model based on the long short-term memory network, which was pre-trained for the material formulation.

[0042] After pouring, the specimen status sensing unit immediately begins operation. A miniature piezoelectric sensor array emits shear waves at a frequency of 100Hz and receives reflected signals, calculating the wave velocity in real time. An infrared surface moisture meter mounted on the robotic arm scans the surface of the filling material once per minute and measures its humidity value. All data is transmitted to the central processing unit in real time. 115 minutes after pouring, the central processing unit combines the rising trend of wave velocity with the decreasing humidity curve to determine that the current state has reached the initial setting point with a probability of 96.7%. The system then sends a trigger command to the intelligent processing robotic arm.

[0043] The intelligent processing robotic arm first moves to a position 50mm above the starting point at the upper left corner of the mold. The end effector switches to a pulsed fiber laser etching head, setting the laser parameters to 20W power, 5kHz pulse frequency, and 0.1mm spot diameter. Simultaneously, the robotic arm scans and ablates the entire surface of the lower filling material at a speed of 10mm / s in a "bow" shaped path (1mm row spacing), controlling the etching depth to 0.2mm. Next, ultrasonic negative pressure coupling cleaning is performed. The robotic arm switches to an ultrasonic negative pressure cleaning head, setting the ultrasonic frequency to 40kHz and the vacuum suction to -80kPa. The robotic arm then traverses the surface of the lower filling material again at a speed of 20mm / s, completing vibration and adsorption cleaning. After the debris is adsorbed, an interface agent is sprayed. The robotic arm switches to an ultrasonic atomizing nozzle and connects to a cement paste interface agent storage tank with a water-cement ratio of 0.4, setting the spray flow rate to... The robotic arm scans at a speed of 30 mm / s and a height of 100 mm above the surface, ultrasonically atomizing and spraying the interface agent to form a uniform thin layer.

[0044] Within 5 minutes of the interface agent spraying (the optimal bonding activity window), the upper filling material is poured up to the top surface of the specimen mold (300mm height). After pouring, it is allowed to stand for 10 minutes before the non-destructive testing unit is activated. Ultrasonic testing is performed; a pair of probes (center frequency 500kHz) measures the acoustic time penetrating the specimen as 25.6μs with an amplitude attenuation of -18dB. This is compared with pre-stored homogeneous baseline data (acoustic time 25.1μs, attenuation -16dB), and the difference is within the allowable range (error <5%). Resistance testing is also performed; the resistivity distribution image reconstructed by the 16-electrode array shows that the interface layer resistivity is uniform, with no abnormally high resistivity (voids) or low resistivity (water accumulation) areas. Finally, the central processing unit generates a report, determining the interlayer bonding quality to be "excellent," with a uniformity index >95%, indicating that subsequent standard curing can be performed.

[0045] 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A device for intelligent processing of the layered interface of filling materials and in-situ performance evaluation, characterized in that, include: The specimen state sensing unit includes a micro piezoelectric sensor array, an infrared surface moisture monitor, and an environmental monitoring module, which is used to monitor the solidification state of the lower filling material poured into the specimen mold in real time. Intelligent robotic arms are used to perform processing tasks at interlayer interfaces; The interlayer reinforcement unit includes a laser etching head, an ultrasonic negative pressure cleaning head, and an ultrasonic atomizing nozzle, which are connected to an intelligent processing robotic arm to sequentially perform laser etching, ultrasonic cleaning, and interface agent spraying on the surface of the lower filling material. The non-destructive testing unit includes an ultrasonic tomography array, a resistance tomography electrode array, a multi-channel data acquisition instrument, and an imaging inversion workstation. The ultrasonic tomography array and the resistance tomography electrode array are connected to the input end of the multi-channel data acquisition instrument, and the imaging inversion workstation is connected to the output end of the multi-channel data acquisition instrument. It is used to perform acoustic and electrical combined non-destructive testing on the filling body after the upper layer is poured. The central processing unit (CPU) has a built-in prediction model based on a long short-term memory network and is connected to the specimen state sensing unit and the intelligent processing robotic arm. It is used to receive signals collected by the specimen state sensing unit, determine the solidification state of the lower filling material, and generate trigger commands to control the intelligent processing robotic arm to start. It is also connected to the imaging inversion workstation to generate an interlayer bonding quality assessment report.

2. The intelligent processing and in-situ performance evaluation device for the layered interface of the filling body according to claim 1, characterized in that: The miniature piezoelectric sensor array is embedded in the side wall of the specimen mold, and the detection end face of each miniature piezoelectric sensor in the array is flush with the inner wall of the specimen mold.

3. The intelligent processing and in-situ performance evaluation device for the layered interface of the filling body according to claim 1, characterized in that: In the non-destructive testing unit, the ultrasonic tomography array is used to construct the acoustic wave transmission field and includes multiple high-frequency ultrasonic transducers distributed in a rectangular array. The high-frequency ultrasonic transducers are embedded in the opposite side walls of the specimen mold and are flush with the inner wall of the specimen mold. The resistivity tomography electrode array is used to construct the current excitation field and the potential measurement field to capture dielectric defects at the interlayer interface. The resistivity tomography electrode array includes multiple electrodes that are uniformly distributed along the circumferential direction of the inner wall of the specimen mold. A multi-channel data acquisition instrument is used to acquire acoustic signals from an ultrasonic tomography array and electrical signals from an electrical resistance tomography electrode array. The imaging inversion workstation is used for data processing and imaging. It is equipped with an acoustic-electric joint inversion algorithm, receives acoustic and electrical signals from a multi-channel data acquisition instrument, and generates a three-dimensional fused image of internal defects of the filling body through synchronous iterative inversion. It outputs resistivity and sound velocity data, providing core data support for the central processor to calculate the interface bonding coefficient and determine the interlayer bonding quality.

4. The method based on the intelligent processing and in-situ performance evaluation device for the layered interface of the filling material according to any one of claims 1-3, characterized in that, Includes the following steps: S1. Multi-dimensional state perception and initial setting prediction: The shear wave velocity change is monitored in real time by a micro piezoelectric sensor array, the surface humidity is scanned in real time by an infrared surface moisture monitor, and the multi-modal data is input into a prediction model based on a long short-term memory network to determine when the lower filling material reaches the initial setting state. If the lower filling material is determined to have reached the initial setting state, a trigger command is generated by the central processing unit. S2. Automated inter-layer interface processing: After receiving a trigger command, the intelligent processing robotic arm executes the following steps sequentially. S21. Laser micro-etching process: The surface slurry is removed by a laser etching head to form a micro-rough interface; S22. Ultrasonic negative pressure coupling cleaning: The ultrasonic negative pressure cleaning head loosens the dust embedded in the pores of the interlayer interface and simultaneously captures and removes it through the negative pressure airflow field. S23. Ultrasonic atomization spraying: the interface agent is atomized into micron-sized droplets through an ultrasonic atomization nozzle and uniformly deposited on the interlayer interface after step S22 to form a modified film with controllable thickness. S3. Time-dependent upper layer casting: Based on the curing characteristic curve of the interface agent and the ambient temperature and humidity, the optimal bonding activity window period is obtained. Within the optimal bonding activity window period, when the interface agent is in a touch-dry state, the upper layer filling is cast. The optimal bonding activity window period is the theoretical time range within which the interface agent can form an effective chemical bond with the upper layer filling. The touch-dry state is the state in which the interface agent forms a film on the surface but the molecular chains are still active during the process of changing from liquid to solid. The curing characteristic curve is the relationship curve between environmental parameters and the corresponding touch-dry time threshold. S4. Multiphysics field non-destructive in-situ evaluation: After the upper layer is poured, the acoustic signal is scanned by an ultrasonic tomography array, the electrical signal is scanned by a resistance tomography electrode array, and the acoustic and electrical signals are acquired by a multi-channel data acquisition instrument. The interlayer bonding quality assessment report is generated by data fusion and cross-gradient constraint inversion through an imaging inversion workstation, thus realizing closed-loop quality control of the preparation process.

5. The method for intelligent processing of the layered interface and in-situ performance evaluation of filling materials according to claim 4, characterized in that: In step S1, after the multimodal data is standardized, the Bayesian optimization algorithm is used to optimize hyperparameters and extract key features. The multimodal data includes shear wave velocity, surface humidity data, ambient temperature and humidity data, and timestamp data. The extracted key features are input into a prediction model based on a long short-term memory network, which outputs the predicted value of the current filling's settling degree and compares it with a preset optimal interface processing threshold. If the error between the predicted value of the current filling's settling degree and the preset optimal interface processing threshold is less than the preset error value, it is determined that the initial settling state has been reached. The central processing unit generates a trigger command in milliseconds and sends it to the control system of the intelligent processing robotic arm to start the subsequent interlayer interface processing process.

6. The method for intelligent processing of the layered interface and in-situ performance evaluation of filling materials according to claim 4, characterized in that: The process parameters for laser micro-etching in step S21 include power, pulse frequency, spot diameter, and scanning speed. The robotic arm automatically calibrates the working height according to the preset path.

7. The method for intelligent processing of the layered interface and in-situ performance evaluation of the filling material according to claim 4, characterized in that: In step S3, the upper filling body is poured by oblique injection, and the material landing point is controlled to deviate from the center of the interlayer interface. The interface contact is completed by flowing and covering.

8. The method for intelligent processing of the layered interface and in-situ performance evaluation of the filling material according to claim 4, characterized in that, The specific process of step S4 is as follows: the ultrasonic tomography array acquires acoustic signals, the electrical resistance tomography electrode array acquires electrical signals, and the multi-channel data acquisition instrument simultaneously acquires acoustic and electrical signals and transmits them to the imaging inversion workstation; the imaging inversion workstation reconstructs the sound velocity distribution through the acoustic signals and the resistivity distribution through the electrical signals, and synchronously iterates and updates the sound velocity distribution and resistivity distribution based on the acoustic-electric joint inversion algorithm with cross-gradient constraints, generating a three-dimensional sound velocity matrix and a three-dimensional resistivity matrix and transmitting them to the central processing unit; the central processing unit takes the sound velocity value and resistivity value of each voxel unit under the same spatial coordinate as input, calculates the comprehensive quality index of each voxel unit, and simultaneously constructs a two-parameter fusion feature vector input to the support vector machine classifier to output physical attribute labels; The central processing unit calculates the interface bonding coefficient based on the comprehensive quality index and generates an interlayer bonding quality assessment report; if the interface bonding coefficient is lower than the preset threshold, the specimen is deemed unqualified and the defect coordinates are marked.

Images