A robot intelligent control method and system based on multi-source data fusion

Abstract

The present application relates to the field of building finish layer detection and robot intelligent control, and particularly relates to a robot intelligent control method and system based on multi-source data fusion.A robot intelligent control method based on multi-source data fusion comprises the following steps: S1: controlling a vibration exciter to apply a fixed-frequency sinusoidal vibration excitation to the surface of the finish layer to induce breathing effect of hidden cracks in the bonding interface, and synchronously collecting surface temperature sequence data, three-dimensional topography point cloud data, environmental temperature and humidity data, and vibration exciter reference signals.The present application uses vibration excitation to induce crack breathing effect as an endogenous thermal modulation signal, fuses multi-source data to invert thermal inertia of the bonding interface, predicts thermal and moisture risks, and drives the robot to perform graded intervention, realizes early perception, advanced warning and precise regulation closed loop under the condition of non-destructive finish layer, effectively blocks the evolution path of hollow and wet damage, and improves the accuracy and robustness of risk assessment.

Description

Technical Field

[0001] This invention relates to the field of inspection under building cladding layers and intelligent robot control, and particularly to an intelligent robot control method and system based on multi-source data fusion. Background Technology

[0002] In interior decoration projects, stone and ceramic tile finishes are wet-applied to the surface of light steel keel gypsum board partition walls using cement mortar, with self-tapping screws completely covered under the finish layer. Over time, factors such as temperature and humidity cycles and mortar shrinkage can induce hidden micro-debonding cracks at the bonding interface around the screws. These cracks continue to develop and, once they penetrate, will form large areas of hollowness, leading to the peeling of the finish layer and condensation and mold growth on the wall. Current detection methods mainly rely on the tapping and listening method, judging whether hollowness has formed by the dullness or crispness of the sound. However, the aforementioned micro-debonding cracks have not yet formed macroscopic hollows in the early stages, and the difference in acoustic response generated by tapping is very weak, making it difficult for the human ear and conventional acoustic testing equipment to effectively distinguish them from intact areas; at the same time, the constraint that the finishing layer itself cannot be damaged makes it difficult to directly explore the state of the bonding interface; the initiation and subcritical propagation of these hidden cracks have a significant lag characteristic of damage before characterization, and often when the scale of identifiable hollows is reached, irreversible structural differentiation has already occurred inside the bonding layer, greatly increasing the cost and difficulty of repair; In assessing the bonding quality under the finishing layer, current technologies generally employ passive acoustic detection and external thermally excited infrared detection. Passive acoustic methods have almost no resolution capability at the micro-damage stage, while external thermally excited methods are limited by the high reflectivity of the stone surface and the thermal bridging effect created by the self-tapping screws, making it difficult to establish a stable and analytical temperature modulation signal at the bonding interface. Furthermore, these methods can only judge the damage state at the current moment and cannot extrapolate future thermal-humidity coupling deterioration trends based on the current state. However, the process from microcrack propagation to blistering, and then to condensation and wet damage, is a dynamic evolution process jointly controlled by the material's thermal inertia distribution, environmental temperature and humidity boundary conditions, and moisture migration paths. Due to the lack of a technology that can penetrate the finishing layer and achieve non-destructive inversion of the thermal properties of the bonding interface at the micro-damage stage, and the lack of a complete framework for combining the inversion results with the thermal-humidity coupling model for advanced risk prediction and proactive intervention, current technologies cannot achieve closed-loop control of the entire process from early perception and trend prediction to precise intervention under the constraint that the finishing layer cannot be damaged. Summary of the Invention

[0003] To overcome the difficulties in early detection of micro-damage and prediction of thermal and humidity risks, this invention provides a robot intelligent control method and system based on multi-source data fusion.

[0004] The technical solution is as follows: A robot intelligent control method based on multi-source data fusion, comprising the following steps: S1: Control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal; S2: Use the three-dimensional topography point cloud data to perform geometric correction and periodic averaging on the surface temperature sequence data, and use the exciter reference signal as a reference to perform phase-locked analysis on the corrected temperature sequence to extract normalized amplitude attenuation and phase delay. S3: Based on the phase delay, the known design depth of the self-tapping screw, and the frequency of the sinusoidal vibration excitation, the thermal diffusivity is inverted pixel by pixel to construct a thermal inertia map. The consistency of the inversion results is verified by using the normalized amplitude attenuation. S4: Using the thermal inertia map and ambient temperature and humidity data as input, construct a thermal-humidity coupling prediction model, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion; S5: When there is a region in the comprehensive risk index map that exceeds the comprehensive risk judgment threshold, an intervention command is generated based on the risk type and location characteristics to control the robot end effector to perform differentiated active intervention actions.

[0005] Preferably, the control exciter applies sinusoidal vibration excitation of a fixed frequency to the surface of the finishing layer, inducing a breathing effect in the hidden cracks at the bonding interface, and simultaneously acquiring surface temperature sequence data, three-dimensional topographic point cloud data, ambient temperature and humidity data, and exciter reference signals, including: the exciter mounted on the robot end effector outputting a fixed frequency... The sinusoidal vibration excitation acts on the finishing layer-adhesive layer-base layer structure, causing the hidden debonding cracks at the adhesive interface to periodically open and close under excitation, thereby modulating the surface temperature field. A high frame rate infrared thermal imager fixed to the same platform as the exciter continuously collects the surface temperature distribution of the finishing layer to obtain the original surface temperature sequence data. A laser profilometer fixed to the same platform as the high frame rate infrared thermal imager synchronously acquires the three-dimensional topographic point cloud data of the wall surface. A temperature and humidity sensor mounted on the robot records the indoor air temperature in real time. and relative humidity The ambient temperature and humidity data are recorded; at the same time, the sinusoidal signal output by the exciter is recorded as the exciter reference signal.

[0006] Preferably, the step of using the three-dimensional topography point cloud data to perform geometric correction and periodic averaging on the surface temperature sequence data, and performing phase-locked analysis on the corrected temperature sequence based on the exciter reference signal to extract normalized amplitude attenuation and phase delay includes: calculating the surface unit normal, distance, and viewpoint factor corresponding to each pixel using the three-dimensional topography point cloud data; performing emissivity compensation and distance attenuation correction on the surface temperature sequence data to eliminate artifacts introduced by geometric factors; and aligning, stacking, and averaging the corrected surface temperature sequence data according to the excitation cycle to obtain a single-cycle representative temperature. The sequence involves: extracting local flatness information of the wall surface based on the three-dimensional topographic point cloud data, selecting areas with good tiling quality as reference areas; using the exciter reference signal as a benchmark, performing digital phase-locked loop operation on the single-cycle representative temperature sequence of each pixel to extract in-phase and quadrature components, and then calculating the amplitude and phase; selecting reference points within the reference area to obtain reference amplitude and reference phase, and calculating the normalized amplitude attenuation and phase delay of each pixel, wherein the normalized amplitude attenuation is the ratio of the amplitude of each pixel to the reference amplitude, and the phase delay is the difference between the phase of each pixel and the reference phase.

[0007] Preferably, the step of inverting the thermal diffusivity pixel-by-pixel based on the phase delay, the known design embedment depth of the self-tapping screw, and the frequency of the sinusoidal vibration excitation to construct a thermal inertia map, and using the normalized amplitude attenuation to verify the consistency of the inversion results, includes: the inversion of the thermal diffusivity based on the formula... Performed pixel by pixel, where pixel position Thermal diffusivity at that location The frequency of the sinusoidal vibration excitation is... Design the embedment depth for the self-tapping screw. The phase delay is defined as follows: when the absolute value of the phase delay of a pixel is less than a preset minimum threshold, the pixel is recorded as an invalid pixel; the consistency of the inverted thermal diffusivity is verified by using the theoretical relationship between the normalized amplitude attenuation and thermal diffusivity; the theoretical value of the amplitude obtained from the phase inversion is compared with the measured normalized amplitude attenuation; pixels exceeding the preset error threshold are judged as unreliable pixels and are filled by interpolation of the neighboring valid pixels; pixels with differences not exceeding the preset error threshold are retained to obtain a reliable thermal diffusivity distribution map, which constitutes the thermal inertia map.

[0008] Preferably, the step of using the theoretical relationship between the normalized amplitude attenuation and thermal diffusivity to perform consistency verification on the inverted thermal diffusivity, comparing the theoretical amplitude value obtained from phase inversion with the measured normalized amplitude attenuation, identifying pixels exceeding a preset error threshold as unreliable pixels and using interpolation with neighboring valid pixels to fill in the gaps, and retaining pixels whose differences do not exceed the preset error threshold, includes: the theoretical relationship between the normalized amplitude attenuation and thermal diffusivity is as follows... ,in pixel position Theoretical value of amplitude at that location, Design the embedment depth for the self-tapping screw. The frequency of the sinusoidal vibration excitation is... The thermal diffusivity is defined as follows: For each valid pixel, the relative deviation between the theoretical amplitude value and the measured normalized amplitude attenuation is calculated; the relative deviation is compared with the preset error threshold; when the relative deviation is greater than the preset error threshold, the pixel is determined to be an unreliable pixel; with the unreliable pixel as the center, valid pixels in the neighboring region that have not been determined to be unreliable pixels are searched; and the average thermal diffusivity of the valid pixels is taken as the interpolation padding value of the unreliable pixel.

[0009] Preferably, the step of constructing a thermal-humidity coupling prediction model using the thermal inertial map and ambient temperature and humidity data as input, calculating the heat accumulation risk index, condensation risk index, and cumulative condensation amount in the future time period, and weightedly fusing them to generate a comprehensive risk index map includes: obtaining the thermal conductivity distribution by combining the thermal inertial map with the material volumetric heat capacity, and establishing a two-dimensional transient heat conduction finite element model of the wall surface; using the indoor air temperature extrapolated from the ambient temperature and humidity data as the indoor side boundary condition to solve for the future temperature field; calculating the product of the peak-to-peak temperature of each pixel and the time-averaged temperature gradient modulus within a temperature cycle as the heat accumulation risk index; calculating the dew point temperature using the indoor air temperature and relative humidity from the ambient temperature and humidity data, and accumulating it in the future temperature field. The duration for which the surface temperature of each pixel is below the dew point temperature is used as a condensation risk indicator. A regression model between thermal diffusivity and water vapor permeability coefficient is established through prefabricated specimen experiments. The water vapor permeability coefficient distribution is obtained by mapping the thermal inertia map, and a steady-state diffusion model of water vapor is established. The water vapor partial pressure field is solved, and the water vapor flux vector is calculated based on the gradient of the water vapor partial pressure field, and a humidified flow line is drawn. The local condensation rate of the cold bridge condensation zone is calculated and integrated within the predicted time range to obtain the cumulative condensation amount. The thermal accumulation risk indicator, condensation risk indicator, and cumulative condensation amount are normalized to preset intervals and then weighted and summed with preset weights to generate a comprehensive risk index map, wherein the preset weights are determined based on the contribution of each indicator to the actual damage results in historical samples.

[0010] Preferably, the step of calculating the dew point temperature using the indoor air temperature and relative humidity from the environmental temperature and humidity data, and accumulating the duration for which the surface temperature of each pixel is lower than the dew point temperature in the future temperature field as a condensation risk indicator, includes: the dew point temperature. According to the formula Calculation, where , and It is a constant. The indoor air temperature is... The relative humidity; the future temperature field obtained by solving the two-dimensional transient heat conduction finite element model on the wall surface. In the middle, regarding the prediction time range Every moment within Determine the surface temperature of each pixel Is it below the stated dew point temperature? The aforementioned condensation risk indicators Defined as ,in For indicator functions, when When the value is 1, The value is 0. To predict the duration of time.

[0011] Preferably, when there are regions in the comprehensive risk index map that exceed the comprehensive risk judgment threshold, an intervention command is generated based on the risk type and location characteristics to control the robot end effector to perform differentiated active intervention actions, including: determining the regions where the pixels in the comprehensive risk index map have a comprehensive risk value greater than or equal to the comprehensive risk judgment threshold as high-risk areas requiring intervention; when a high-risk area requiring intervention simultaneously meets multiple intervention conditions, the intervention actions are executed in the priority order of microporous grouting, heat insulation and anti-condensation coating spraying, and waterproof vapor interface agent spraying, and the corresponding risk index decreases to the point where the comprehensive risk is resolved after the previous intervention action is completed. When the threshold is below, no further intervention actions are performed; for areas located at the center of self-tapping screws where the heat accumulation risk index and cumulative condensation amount are co-located, a micro-hole grouting command is generated, controlling the end drilling device to drill micro-holes in the finishing layer directly above the screw head to the debonding interface, and injecting repair grout with a thermal conductivity matching that of the substrate under low pressure to fill the debonding cracks and restore thermal contact; for screw head areas with extremely high condensation risk index but no hollow areas, a heat insulation and anti-condensation coating spraying command is generated, controlling the end spraying device to apply the heat insulation and anti-condensation coating; for board seam areas where humid airflow lines converge densely and there are no self-tapping screws, a waterproof vapor interface agent spraying command is generated, controlling the end spraying device to spray the waterproof vapor interface agent.

[0012] Preferably, the step of generating a micro-hole grouting command for the area located at the center of the self-tapping screw and where the heat accumulation risk index and cumulative condensation amount coexist, controlling the end drilling device to drill micro-holes in the finishing layer directly above the screw head to the debonding interface, and injecting a repair grout with a thermal conductivity matching that of the substrate under low pressure to fill the debonding crack and restore thermal contact includes: using the three-dimensional topographic point cloud data to locate the position of the self-tapping screw head, and determining the precise coordinates of each screw-tapping center by combining the self-tapping screw position coordinates provided by the construction drawings; and selecting the area in the comprehensive risk index map that simultaneously meets the conditions of heat accumulation risk index exceeding a preset heat accumulation threshold and cumulative condensation amount exceeding a certain threshold. The area with a preset condensation threshold is determined as the area where the heat accumulation risk index and the cumulative condensation amount coexist. The precise coordinates of the high-risk area requiring intervention and the center of the self-tapping screw, as well as the coexisting area, are spatially superimposed and analyzed to screen out the target intervention point located at the center of the self-tapping screw and coexisting with it. A micro-hole grouting command is generated for the target intervention point. The micro-hole grouting command controls the end drilling device to drill a micro-hole in the finishing layer directly above the screw head. The drilling depth reaches the designed embedment depth of the self-tapping screw. The grouting device is controlled to inject repair grout that matches the thermal conductivity of the substrate at a preset low pressure to fill the debonding crack and restore thermal contact.

[0013] Preferably, a robot intelligent control system based on multi-source data fusion further includes: The excitation and multi-source sensing module is used to control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal. The data preprocessing and phase-locked feature extraction module is used to perform geometric correction and periodic averaging on the surface temperature sequence data using the three-dimensional topography point cloud data, and to perform phase-locked analysis on the corrected temperature sequence based on the exciter reference signal to extract normalized amplitude attenuation and phase delay. The thermal inertial inversion and map construction module is used to invert the thermal diffusivity pixel by pixel based on the phase delay, the known self-tapping screw design burial depth and the frequency of the sinusoidal vibration excitation, construct the thermal inertial map, and use the normalized amplitude attenuation to verify the consistency of the inversion results. The thermal-humidity coupling prediction and risk fusion module is used to construct a thermal-humidity coupling prediction model with the thermal inertia map and environmental temperature and humidity data as input, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion. The integrated decision-making and active intervention control module is used to generate intervention commands based on risk type and location characteristics when there are areas in the comprehensive risk index map that exceed the comprehensive risk judgment threshold, and control the robot end effector to perform differentiated active intervention actions. The robot end effector module is used to receive intervention commands generated by the fusion decision and active intervention control module, and control the drilling device, grouting device and spraying device to perform corresponding active intervention actions according to the intervention commands, including micro-hole grouting, heat insulation and anti-condensation coating spraying and waterproof vapor interface agent spraying.

[0014] Beneficial effects: This invention overcomes the limitations of the finishing layer not being able to be damaged and external thermal excitation failure by controlling the exciter to apply sinusoidal vibration and utilizing the crack breathing effect to obtain the intrinsic thermal modulation signal, thus achieving early detection of hidden micro-debonding. By simultaneously acquiring surface temperature sequences, three-dimensional morphology point clouds, and environmental temperature and humidity data, geometric correction and phase-locked analysis are performed to extract normalized amplitude attenuation and phase delay, eliminating geometric artifacts and improving the signal-to-noise ratio of the modulation signal. The thermal diffusivity of the bonding interface is inverted using the phase delay and known self-tapping screw embedment depth, and consistency is verified by amplitude attenuation, constructing a reliable spectrum characterizing the interface thermal inertia, thus quantifying the thermal properties of invisible damage. Based on the thermal inertia spectrum, combined with environmental temperature and humidity boundary conditions... A thermal-humidity coupling prediction model is established to calculate the risks of thermal accumulation, cold bridging condensation, and cumulative condensation due to moisture diffusion. A comprehensive risk index is generated through weighted fusion, enabling proactive prediction of future risks from the current state. Based on the comprehensive risk index and self-tapping screw positioning information, a robot is driven to perform graded intervention actions such as micro-hole grouting, thermal insulation and anti-condensation coating spraying, and waterproof vapor interface agent spraying. The effects of previous interventions determine subsequent steps, forming a complete closed loop from perception, modeling, prediction to proactive intervention, blocking the evolution path of voids and moisture damage. By deeply fusing multi-source heterogeneous data at the feature layer and decision layer, the accuracy and robustness of the assessment are significantly improved, giving the ability to provide early warning and precise control under the constraint that the finishing layer cannot be damaged. Attached Figure Description

[0015] Figure 1 This is a flowchart of a robot intelligent control method based on multi-source data fusion according to the present invention; Figure 2 This is a schematic diagram of the structure of a robot intelligent control system based on multi-source data fusion according to the present invention. Detailed Implementation

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: A robot intelligent control method based on multi-source data fusion, such as Figure 1 As shown, it includes the following steps: S1: Control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal; S2: Use the three-dimensional topography point cloud data to perform geometric correction and periodic averaging on the surface temperature sequence data, and use the exciter reference signal as a reference to perform phase-locked analysis on the corrected temperature sequence to extract normalized amplitude attenuation and phase delay. S3: Based on the phase delay, the known design depth of the self-tapping screw, and the frequency of the sinusoidal vibration excitation, the thermal diffusivity is inverted pixel by pixel to construct a thermal inertia map. The consistency of the inversion results is verified by using the normalized amplitude attenuation. S4: Using the thermal inertia map and ambient temperature and humidity data as input, construct a thermal-humidity coupling prediction model, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion; S5: When there is a region in the comprehensive risk index map that exceeds the comprehensive risk judgment threshold, an intervention command is generated based on the risk type and location characteristics to control the robot end effector to perform differentiated active intervention actions.

[0018] A vibrator mounted on the end effector of the robot outputs a fixed frequency. The sinusoidal vibration excitation acts on the finishing layer-adhesive layer-base layer structure, causing the hidden debonding cracks at the adhesive interface to periodically open and close under excitation, thereby modulating the surface temperature field. A high frame rate infrared thermal imager fixed to the same platform as the exciter continuously collects the surface temperature distribution of the finishing layer to obtain the original surface temperature sequence data. A laser profilometer fixed to the same platform as the high frame rate infrared thermal imager synchronously acquires the three-dimensional topographic point cloud data of the wall surface. A temperature and humidity sensor mounted on the robot records the indoor air temperature in real time. and relative humidity The ambient temperature and humidity data are recorded; at the same time, the sinusoidal signal output by the exciter is recorded as the exciter reference signal.

[0019] It should be explained that the vibrator is installed on the robot's end effector and electrically connected to the robot's motion control unit, and the sampling frequency for continuously collecting the surface temperature distribution of the finishing layer... The fixed frequency The exciter control module generates and outputs the results to the exciter drive circuit based on the preset detection task. Before performing the detection, the robot moves to the target area according to the pre-established wall detection path, ensuring that the exciter maintains a preset contact state with the surface of the finishing layer. After confirming through the contact state detection unit that the excitation can be effectively transmitted to the finishing layer-adhesive layer-base structure, the vibration excitation is initiated. The hidden debonding cracks repeatedly open and close under the periodic vibration, creating a periodic heat transfer difference between the cracked area and the surrounding intact area. The high frame rate infrared thermal imager and the exciter are rigidly connected and fixed to the same mounting platform, with a sampling frequency of [missing information]. Based on the excitation frequency The required spectral resolution for subsequent phase-locked loop analysis was determined, and the original surface temperature sequence data was continuously acquired by maintaining time synchronization with the exciter through a unified clock synchronization module. ,in for According to time The acquired sequence data; the laser profilometer and the high frame rate infrared thermal imager share a coordinate reference and synchronously acquire point cloud data of the three-dimensional shape of the wall surface. Furthermore, point cloud integrity verification is performed during the data acquisition process, and missing points are filled in using a neighborhood reconstruction method; the indoor air temperature... and the relative humidity The ambient temperature and humidity data are obtained in real time by temperature and humidity sensors mounted on the robot and stored as environmental temperature and humidity data after the data validity is verified; simultaneously, the sinusoidal signal output by the exciter in real time is also stored. The data is recorded as a reference signal for the exciter and is associated with the original surface temperature sequence data, the three-dimensional topographic point cloud data, and the environmental temperature and humidity data according to a unified timestamp.

[0020] Using the three-dimensional topographic point cloud data, the surface unit normal, distance, and viewpoint factor corresponding to each pixel are calculated. Emissivity compensation and distance attenuation correction are applied to the surface temperature sequence data to eliminate artifacts introduced by geometric factors. The corrected surface temperature sequence data is aligned, superimposed, and averaged according to the excitation cycle to obtain a single-cycle representative temperature sequence. Based on the three-dimensional topographic point cloud data, local wall flatness information is extracted, and areas with good tiling quality are selected as reference areas. Using the exciter reference signal as a reference, digital phase-locked loop (PLL) is performed on the single-cycle representative temperature sequence of each pixel to extract in-phase and quadrature components, and then the amplitude and phase are calculated. A reference point is selected within the reference area to obtain the reference amplitude and reference phase. The normalized amplitude attenuation and phase delay of each pixel are calculated, where the normalized amplitude attenuation is the ratio of the pixel amplitude to the reference amplitude, and the phase delay is the difference between the pixel phase and the reference phase.

[0021] It should be explained that the aforementioned three-dimensional topographic point cloud data After being acquired by a laser profilometer, coordinate unification and noise filtering are performed. Then, a local plane is fitted based on the neighborhood spatial coordinates of each point to obtain the surface unit normal vector of the corresponding pixel position. The distance is the spatial distance from the optical center of the infrared thermal imager to the corresponding surface unit. The point cloud coordinates and thermal imager extrinsic parameter calibration results are used to calculate the viewing angle factor, which is determined by the angle between the surface unit normal vector and the thermal imager observation direction. Based on the surface unit normal vector, the distance, and the viewing angle factor, the original surface temperature sequence data is processed. Perform emissivity compensation and range attenuation correction to obtain the corrected temperature sequence. Emissivity compensation is used to correct the radiant energy deviation caused by local wall tilt, and distance attenuation correction is used to eliminate temperature measurement errors caused by changes in observation distance; after correction, the exciter reference signal is used. Corresponding incentive cycle As a basis for segmentation, temperature data from multiple continuously acquired excitation cycles are periodically aligned, and the temperature values ​​at corresponding times are superimposed and averaged to obtain a representative temperature sequence for a single cycle. ,in The time variable is defined as a single-cycle variable; the superposition averaging process enhances the temperature response component synchronized with the excitation frequency and suppresses random noise; the flatness index of the local area of ​​the wall is calculated using the three-dimensional topography point cloud data, and the area with no self-tapping screws, no obvious geometric changes, and good tiling quality is selected as the reference area by statistically analyzing the deviation of the point cloud in the area from the fitting plane; during the phase-locked loop analysis, the exciter reference signal is used as the synchronization benchmark, and the single-cycle representative temperature sequence corresponding to each pixel is calculated. Perform digital phase-locked loop (PLL) operation to calculate the in-phase components separately. and orthogonal components , , Among them, the in-phase component The temperature response component, orthogonal component, is in phase with the exciter reference signal. The temperature response component, which indicates a phase shift relative to the exciter reference signal, and the two together characterize the thermal wave response state; then, based on... Calculate amplitude based on Calculate the phase; select a reference point within the reference region. To obtain the reference amplitude and reference phase The reference point is determined by a location with optimal flatness and stable temperature response; subsequently, the normalized amplitude attenuation is calculated. and phase delay The normalized amplitude attenuation is used to characterize the degree of thermal wave energy attenuation at each location, and the phase delay is used to characterize the time lag characteristics during thermal wave propagation, ultimately forming the basic parameter dataset required for subsequent thermal diffusivity inversion.

[0022] The inversion of the thermal diffusivity is based on the formula Performed pixel by pixel, where pixel position Thermal diffusivity at that location The frequency of the sinusoidal vibration excitation is... Design the embedment depth for the self-tapping screw. The phase delay is defined as follows: when the absolute value of the phase delay of a pixel is less than a preset minimum threshold, the pixel is recorded as an invalid pixel; the consistency of the inverted thermal diffusivity is verified by using the theoretical relationship between the normalized amplitude attenuation and thermal diffusivity; the theoretical value of the amplitude obtained from the phase inversion is compared with the measured normalized amplitude attenuation; pixels exceeding the preset error threshold are judged as unreliable pixels and are filled by interpolation of the neighboring valid pixels; pixels with differences not exceeding the preset error threshold are retained to obtain a reliable thermal diffusivity distribution map, which constitutes the thermal inertia map.

[0023] It should be explained that the thermal diffusivity The phase delay is used to characterize the ability of heat to diffuse and propagate within the material at the corresponding pixel location. The results obtained directly from the phase-locked analysis are used; the sinusoidal vibration excitation frequency is used. The exciter reference signal parameters are derived from the actual output and synchronous recording of the exciter control module; the self-tapping screw design embedment depth Design data derived from construction drawings or building information models, combined with the three-dimensional topographic point cloud data, is used to spatially register the surface position of the finishing layer to determine the corresponding burial depth parameters. Before performing thermal diffusivity inversion, the phase delay is first tested for effectiveness, and phase delay fluctuation data within the reference area is statistically analyzed to establish a phase noise distribution. The boundary value corresponding to the upper limit of phase noise is determined as the preset minimum threshold. When the pixel thermal response is below the preset minimum threshold, it is considered that the pixel thermal response is dominated by noise and is recorded as an invalid pixel; subsequently, based on... The thermal diffusivity is calculated pixel by pixel, where the formula is... This indicates the scale of heat wave propagation, which is jointly determined by the excitation frequency and the depth of the heat propagation path. The phase lag intensity generated during heat wave propagation is represented, and the two are combined to invert the internal thermal diffusivity of the material. After the inversion is completed, the correspondence between the phase inversion thermal diffusivity and the amplitude attenuation response is established. The theoretical amplitude value is calculated for each effective pixel and compared with the measured normalized amplitude attenuation. The preset error threshold is obtained by collecting historical normal wall sample data. The deviation distribution between the theoretical amplitude value and the measured normalized amplitude attenuation is calculated respectively, and the error value corresponding to the boundary of the deviation distribution is determined as the preset error threshold. When the comparison result exceeds the preset error threshold, the corresponding pixel is marked as an unreliable pixel. After searching for effective pixels in the neighborhood, the thermal diffusivity data is supplemented by spatial interpolation. When the comparison result does not exceed the preset error threshold, the original inversion result is retained, and finally a continuous and reliable thermal diffusivity distribution map is obtained as the thermal inertial map.

[0024] The theoretical relationship between the normalized amplitude decay and the thermal diffusivity is as follows: ,in pixel position Theoretical value of amplitude at that location, Design the embedment depth for the self-tapping screw. The frequency of the sinusoidal vibration excitation is... The thermal diffusivity is defined as follows: For each valid pixel, the relative deviation between the theoretical amplitude value and the measured normalized amplitude attenuation is calculated; the relative deviation is compared with the preset error threshold; when the relative deviation is greater than the preset error threshold, the pixel is determined to be an unreliable pixel; with the unreliable pixel as the center, valid pixels in the neighboring region that have not been determined to be unreliable pixels are searched; and the average thermal diffusivity of the valid pixels is taken as the interpolation padding value of the unreliable pixel.

[0025] It should be explained that the theoretical value of the amplitude... The self-tapping screw design embedment depth is calculated based on a theoretical model of heat wave propagation and attenuation within the material. The design burial depth parameters recorded in the construction drawings or building information model are used, and spatial location matching is completed by combining the three-dimensional topographic point cloud data before being retrieved; the sinusoidal vibration excitation frequency... The thermal diffusivity is derived from the recorded reference signal of the exciter. Thermal diffusivity distribution data obtained from the preceding thermal diffusivity inversion step; based on Calculate the theoretical amplitude value corresponding to each effective pixel, where the negative exponent term in the exponential function represents the energy attenuation effect generated during the propagation of the heat wave. This indicates the design embedment depth of the self-tapping screw as determined by the construction drawings, and is used as a parameter for heat wave propagation distance in the calculation of the theoretical amplitude value. The attenuation rate is determined by both the excitation frequency and the material's thermal diffusivity. These two parts are multiplied together to reflect the amplitude loss of the heat wave propagating from the hidden interface to the surface. After calculating the theoretical amplitude value, the measured normalized amplitude attenuation obtained from the preceding phase-locked analysis is called. Calculate the relative deviation ;in This represents the absolute difference between theoretical predictions and actual results. The measured normalized amplitude attenuation is used to eliminate the dimensional influence caused by different amplitude levels. The preset error threshold is determined using the statistical results of the deviation of historical normal samples. The theoretical amplitude value and the measured normalized amplitude attenuation data are extracted from the historical normal wall samples, and the corresponding relative deviation set is calculated. The preset error threshold is determined based on the deviation distribution boundary. When the relative deviation is greater than the preset error threshold, the corresponding pixel is determined to be an unreliable pixel. Then, a neighborhood search is performed with the unreliable pixel as the center to filter the effective pixels in the neighboring area that have not been determined to be unreliable pixels. The arithmetic mean of the thermal diffusivity of the effective pixels is calculated, and the mean is used as the interpolation result. After all unreliable pixels are corrected, thermal diffusivity distribution data that meets the continuity requirement is formed.

[0026] The thermal conductivity distribution is obtained by combining the thermal inertia map with the material's volumetric heat capacity, and a two-dimensional transient heat conduction finite element model of the wall surface is established. The future temperature field is obtained by extrapolating the future temperature curve from the indoor air temperature in the environmental temperature and humidity data as the indoor side boundary condition. The product of the peak-to-peak temperature of each pixel and the time-averaged temperature gradient modulus within a temperature cycle is calculated as a heat accumulation risk indicator. The dew point temperature is calculated using the indoor air temperature and relative humidity in the environmental temperature and humidity data, and the duration for which the surface temperature of each pixel is below the dew point temperature in the future temperature field is accumulated as a condensation risk indicator. The thermal diffusivity is established through prefabricated specimen experiments. A regression model between the water vapor permeability coefficient and the thermal inertia map is used to map the water vapor permeability coefficient distribution, establish a steady-state diffusion model for water vapor, solve for the water vapor partial pressure field, calculate the water vapor flux vector based on the gradient of the water vapor partial pressure field and draw the humid flow line, calculate the local condensation rate of the cold bridge condensation zone and integrate it within the predicted time range to obtain the cumulative condensation amount; after normalizing the thermal accumulation risk index, condensation risk index and cumulative condensation amount to preset intervals, they are weighted and summed with preset weights to generate a comprehensive risk index map, wherein the preset weights are determined according to the contribution of each index to the actual damage result in historical samples.

[0027] It should be explained that the thermal inertia map is the thermal diffusivity distribution. Before constructing the thermal-humidity coupling prediction model, the spatial range of the finishing layer, adhesive layer and base layer is determined according to the construction drawings. The volumetric heat capacity of the materials in each region is extracted from the material parameter database corresponding to the engineering material models. The material parameter database is established based on material production parameters and engineering acceptance data, and a mapping relationship is established between material type and regional location; subsequently, based on the thermal conductivity formula... Calculate the thermal conductivity distribution, where Indicates the material's internal heat dissipation capacity. This represents the material's ability to store heat per unit volume, and the product of these two values ​​characterizes the thermal conductivity at the corresponding location. After obtaining the thermal conductivity distribution, a two-dimensional transient thermal conduction finite element model of the wall surface is established. ,in This indicates the change in the amount of heat stored inside a material per unit time. This represents the amount of heat conducted and diffused within a spatial area, and the balance between the two describes the change in wall temperature over time; the indoor air temperature... Data is derived from continuous temperature and humidity sensors. Before making predictions, outlier removal and time series integrity checks are performed. Then, a time series prediction model is used to extrapolate future temperature trends to obtain the future temperature curve. The indoor boundary conditions of the finite element model are used to solve for the future temperature field. After obtaining the future temperature field, the temperature change process of each pixel within the predicted time range is analyzed, and the peak-to-peak temperature in the temperature cycle is calculated. and time-averaged temperature gradient modulus ,in accordance with Calculate the thermal accumulation risk index, among which Indicates the intensity of temperature fluctuations. This indicates the degree of variation in spatial temperature distribution; both factors together reflect the risk of accumulated thermal stress. The indoor air temperature is used... and relative humidity The dew point temperature is calculated, and it is compared with the surface temperature of each pixel in the future temperature field at each time step. The cumulative duration of the temperature being below the dew point temperature is used as a condensation risk indicator. For the moisture migration analysis, a regression model between thermal diffusivity and water vapor permeability is pre-established and stored using pre-prepared specimens. The regression model uses pre-prepared specimens with different degrees of damage and different moisture content as sample sources. The thermal diffusivity and water vapor permeability of each specimen are measured to construct a sample correspondence between thermal diffusivity and water vapor permeability. Regression analysis is performed based on the sample correspondence to obtain the mapping function between thermal diffusivity and water vapor permeability. The regression error is verified by independent validation samples. After successful verification, the mapping function is stored as a pre-trained regression model for subsequent online calculation of the water vapor permeability distribution. The water vapor permeability distribution is then obtained. Then, a corresponding two-dimensional calculation area is established based on the wall structure, and water vapor partial pressure boundary conditions are set according to the indoor and base layer environmental conditions; the water vapor permeability coefficient distribution is then calculated. Assign values ​​to the corresponding computational domain cells to construct the steady-state diffusion control equations. Subsequently, the water vapor partial pressure at each node was discretized using the finite element method to obtain the water vapor partial pressure field. ;according to Calculate the water vapor flux vector, where Indicates the medium's ability to transport water vapor. The gradient of water vapor partial pressure variation is represented, and the product of the two is used to characterize the direction and intensity of moisture migration; the local condensation rate is calculated based on the local temperature field and partial pressure field of the cold bridge condensation area, and the cumulative condensation is obtained by integrating over the predicted time range. After calculating the heat accumulation risk index, condensation risk index, and cumulative condensation, normalization was performed based on the statistical distribution of the indexes in the historical sample database. For the heat accumulation risk index... ,according to Normalization is performed; for condensation risk indicators ,according to Normalize; for cumulative condensation ,according to Normalization is performed; the maximum and minimum values ​​are derived from the statistical results of the corresponding indicators in the historical sample database; the preset weights are calculated using the correlation data between each indicator in the historical samples and the actual damage results; let the total number of historical samples be... For the first A historical sample was used, and its thermal accumulation risk index was recorded as follows: The condensation risk indicator is The cumulative condensation amount is The actual damage result is A linear regression model was established using all historical samples. ;in, For the intercept term, , and These represent the regression contribution coefficients of the heat accumulation risk index, condensation risk index, and cumulative condensation amount to the actual damage results, respectively. After completing the regression solution, the absolute value of each regression contribution coefficient is taken as the contribution degree of the corresponding index, and normalized according to the following formula: , , ;in, , and These are the preset weights corresponding to the heat accumulation risk index, condensation risk index, and cumulative condensation amount, respectively; normalization is applied to make the sum of the three preset weights equal to 1; finally, based on... Weighted fusion is performed to generate a comprehensive risk index map.

[0028] The dew point temperature According to the formula Calculation, where , and It is a constant. The indoor air temperature is... The relative humidity; the future temperature field obtained by solving the two-dimensional transient heat conduction finite element model on the wall surface. In the middle, regarding the prediction time range Every moment within Determine the surface temperature of each pixel Is it below the dew point temperature? The aforementioned condensation risk indicators Defined as ,in For indicator functions, when When the value is 1, The value is 0. To predict the duration of time.

[0029] It should be explained that the indoor air temperature... and the relative humidity The data, derived from continuous data collection by the robot's onboard temperature and humidity sensors, undergoes time synchronization, outlier removal, and missing data completion before being used in dew point temperature calculation. Environmental temperature and humidity data corresponding to the predicted future temperature field are used as the calculation input. The dew point temperature... According to public Calculation, where The formula uses the Magnus dew point calculation model, where... For temperature and humidity coupling intermediate variables, in the formula The second item is used to characterize the degree of influence of air temperature on the state of saturated water vapor. This constant is used to characterize the degree of deviation of the actual humidity level of the air from the saturation state. The sum of the two parts reflects the critical condition under which the current air reaches a saturated condensation state. and The parameters are derived from the Magnus model library and remain constant throughout the prediction process; the dew point temperature is obtained. Then, the future temperature field was obtained by solving the two-dimensional transient heat conduction finite element model of the wall. Within the predicted time range The system iterates through each pixel position time-by-time to measure the surface temperature. With dew point temperature Comparison and judgment; the predicted time length The prediction time length is determined based on the risk assessment cycle corresponding to the wall inspection task, and the preset assessment cycle parameter in the inspection task configuration file is used. When the inspection task does not have a separate assessment cycle set, the standard prediction cycle used for thermal and humidity risk assessment in the historical engineering sample database is used as the prediction time length. ;when Below When this occurs, it indicates that the corresponding location meets the conditions for condensation formation, and the indicator function... The value is a valid state; when Greater than or equal to At that time, the indicator function takes an invalid value; subsequently, according to Calculate the condensation risk index, where the integral process is used to accumulate the duration of condensation conditions within the predicted time range. The larger the value, the longer the corresponding pixel location has been in a condensation environment, and the higher the risk of subsequent cold bridging condensation and wet damage.

[0030] The areas where the comprehensive risk value in the comprehensive risk index map is greater than or equal to the comprehensive risk judgment threshold are identified as high-risk areas requiring intervention. When a high-risk area requiring intervention meets multiple intervention conditions, the intervention actions are executed in the order of priority: micro-hole grouting, thermal insulation and anti-condensation coating spraying, and waterproof vapor interface agent spraying. After the previous intervention action is completed, the corresponding risk index drops below the comprehensive risk relief threshold, and the subsequent intervention action is no longer executed. For areas located at the center of self-tapping screws where the heat accumulation risk index and the accumulated condensation amount are co-located, a micro-hole grouting command is generated, controlling the end drilling device to drill micro-holes in the finishing layer directly above the screw head to the debonding interface, and injecting repair grout with a thermal conductivity matching that of the substrate under low pressure to fill the debonding cracks and restore thermal contact. For screw head areas with extremely high condensation risk index but no hollow areas, a thermal insulation and anti-condensation coating spraying command is generated, controlling the end spraying device to apply the thermal insulation and anti-condensation coating. For board seam areas where humid airflow lines converge and there are no self-tapping screws, a waterproof vapor interface agent spraying command is generated, controlling the end spraying device to spray the waterproof vapor interface agent.

[0031] It should be explained that after generating the comprehensive risk index map, region clustering analysis is performed on the risk values ​​of each pixel in the comprehensive risk index map to divide spatially continuous pixels with similar risk characteristics into candidate risk regions; the comprehensive risk judgment threshold is determined based on a historical engineering sample database, collecting historical sample data that have completed detection and have subsequent damage verification results, statistically analyzing the correspondence between the comprehensive risk index and the actual damage occurrence, constructing a risk index-damage probability correlation model, and determining the risk index boundary value corresponding to the preset damage judgment condition as the comprehensive risk judgment threshold; when the candidate region When the comprehensive risk value of pixels within the domain is greater than or equal to the comprehensive risk judgment threshold, the region is determined as a high-risk area requiring intervention, and corresponding risk area coordinate information is generated. Subsequently, the risk type of the high-risk area requiring intervention is identified, and the spatial distribution characteristics of heat accumulation risk indicators, condensation risk indicators, and accumulated condensation are analyzed to determine whether the corresponding intervention conditions are met. When the same region meets multiple intervention conditions simultaneously, they are executed in the order of priority: microporous grouting, thermal insulation and anti-condensation coating spraying, and waterproof vapor interface agent spraying. This is because microporous grouting can directly restore the thermal contact state of the debonded area and prevent subsequent thermal and wet deterioration. The path therefore has the highest priority; the comprehensive risk relief threshold and the comprehensive risk judgment threshold are determined independently, the comprehensive risk judgment threshold is used to trigger intervention decisions, and the comprehensive risk relief threshold is used to evaluate the intervention effect. The comprehensive risk relief threshold is determined through the statistical results of risk indicators in the stable operating area after intervention in historical intervention cases; after each intervention action is completed, the local area data collection, risk calculation, and risk assessment process is re-executed. When the corresponding risk indicator drops below the comprehensive risk relief threshold, it is determined that the current risk has been effectively controlled, and subsequent intervention actions are terminated; for those located in the self In areas where the center of the nail tapping point and the risk index of heat accumulation and the cumulative condensation volume are spatially co-located, a micro-hole grouting command is generated to control the end drilling device to reach the target position and form a micro-hole leading to the debonding interface in the finishing layer directly above the nail head. Repair grout matching the thermal conductivity of the substrate is injected through the grouting device. For nail head areas with extremely high condensation risk indicators but no hollow characteristics detected, a heat insulation and anti-condensation coating spraying command is generated to reduce the thermal bridging effect of the nail head by spraying a heat insulation and anti-condensation coating. For board seam areas where moisture airflow lines converge densely and there are no self-tapping screws, a waterproof vapor interface agent spraying command is generated to inhibit the continuous migration of moisture along the board seam by forming a moisture barrier layer.

[0032] The three-dimensional topographic point cloud data is used to locate the self-tapping screw head position, and the precise coordinates of each self-tapping screw center are determined by combining the self-tapping screw position coordinates provided by the construction drawings. The area in the comprehensive risk index map that simultaneously meets the conditions of thermal accumulation risk index exceeding the preset thermal accumulation threshold and cumulative condensation exceeding the preset condensation threshold is identified as the area where thermal accumulation risk index and cumulative condensation coexist. The precise coordinates of the high-risk area requiring intervention, the center of the self-tapping screw, and the coexisting area are spatially superimposed and analyzed to screen out the target intervention point located at the center of the self-tapping screw and coexisting with it. A micro-hole grouting command is generated for the target intervention point. The micro-hole grouting command controls the end drilling device to drill a micro-hole in the finishing layer directly above the screw head. The drilling depth reaches the designed embedment depth of the self-tapping screw. The grouting device is controlled to inject repair grout that matches the thermal conductivity of the substrate at a preset low pressure to fill the debonding crack and restore thermal contact.

[0033] It should be explained that the three-dimensional topographic point cloud data first undergoes coordinate registration processing to establish a unified wall spatial coordinate system. Then, based on the geometric features of the point cloud surface, local micro-protrusions, depressions, and regular circular structural features of the wall are identified to extract suspected self-tapping screw head areas. Simultaneously, the coordinate data of self-tapping screw positions stored in the construction drawings or building information model are retrieved, and spatial transformation between the construction drawing coordinate system and the point cloud coordinate system is completed through wall boundary feature points. Nearest neighbor matching and error verification are used to match the suspected self-tapping screw head areas with the theoretical screw position coordinates, and the coordinates where the matching error meets the positioning requirements are determined as the precise coordinates of their respective screw-tapping centers. The preset heat accumulation threshold is obtained through statistical analysis of the correlation between heat accumulation risk indicators and subsequent hollow expansion results in historical engineering samples. The preset condensation threshold is obtained through analysis of the correspondence between accumulated condensation and actual condensation damage results in historical samples. Subsequently, each risk area in the comprehensive risk index map is traversed. When the heat accumulation risk indicator in the same area exceeds the preset heat accumulation threshold and the accumulated condensation... When the condensation amount exceeds the preset condensation threshold, the area is identified as a region where the heat accumulation risk index and the accumulated condensation amount coexist. Based on this, the high-risk area requiring intervention, the precise coordinates of the self-tapping screw center, and the coexisting region are loaded into a unified spatial coordinate system for spatial overlay analysis. By calculating the intersection relationship of the regions, the location that simultaneously satisfies the conditions of being located at the center of the self-tapping screw, belonging to the high-risk area requiring intervention, and being within the coexisting region is selected to determine the target intervention point. A micro-hole grouting command is generated for the target intervention point, where the drilling depth parameter is derived from the self-tapping screw design embedment depth data recorded in the construction drawings, and the depth is corrected by combining the current surface position of the finishing layer measured by the point cloud. The thermal conductivity parameter of the repair grout is obtained through the material performance test report, and a grout formula that meets the matching requirements with the thermal conductivity deviation of the corresponding substrate is selected. The preset low pressure is determined by a pre-established grouting process database, which is established based on the process test results corresponding to different finishing materials, crack morphologies, and grout rheological properties. Finally, the heat transfer continuity and interface contact state of the crack area are restored through grouting.

[0034] Example 2: Based on Example 1, a robot intelligent control system based on multi-source data fusion, such as... Figure 2 As shown, it also includes: The excitation and multi-source sensing module is used to control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal. The data preprocessing and phase-locked feature extraction module is used to perform geometric correction and periodic averaging on the surface temperature sequence data using the three-dimensional topography point cloud data, and to perform phase-locked analysis on the corrected temperature sequence based on the exciter reference signal to extract normalized amplitude attenuation and phase delay. The thermal inertial inversion and map construction module is used to invert the thermal diffusivity pixel by pixel based on the phase delay, the known self-tapping screw design burial depth and the frequency of the sinusoidal vibration excitation, construct the thermal inertial map, and use the normalized amplitude attenuation to verify the consistency of the inversion results. The thermal-humidity coupling prediction and risk fusion module is used to construct a thermal-humidity coupling prediction model with the thermal inertia map and environmental temperature and humidity data as input, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion. The integrated decision-making and active intervention control module is used to generate intervention commands based on risk type and location characteristics when there are areas in the comprehensive risk index map that exceed the comprehensive risk judgment threshold, and control the robot end effector to perform differentiated active intervention actions. The robot end effector module is used to receive intervention commands generated by the fusion decision and active intervention control module, and control the drilling device, grouting device and spraying device to perform corresponding active intervention actions according to the intervention commands, including micro-hole grouting, heat insulation and anti-condensation coating spraying and waterproof vapor interface agent spraying.

[0035] Although this disclosure has been shown and described with reference to specific exemplary embodiments thereof, those skilled in the art will understand that various changes in form and detail may be made to this disclosure without departing from the spirit and scope of the disclosure as defined by the appended claims and their equivalents. Therefore, the scope of this disclosure should not be limited to the above embodiments, but should be defined not only by the appended claims, but also by their equivalents.

Claims

1. A robot intelligent control method based on multi-source data fusion, characterized in that, Includes the following steps: S1: Control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal; S2: Use the three-dimensional topography point cloud data to perform geometric correction and periodic averaging on the surface temperature sequence data, and use the exciter reference signal as a reference to perform phase-locked analysis on the corrected temperature sequence to extract normalized amplitude attenuation and phase delay. S3: Based on the phase delay, the known design depth of the self-tapping screw, and the frequency of the sinusoidal vibration excitation, the thermal diffusivity is inverted pixel by pixel to construct a thermal inertia map. The consistency of the inversion results is verified by using the normalized amplitude attenuation. S4: Using the thermal inertia map and ambient temperature and humidity data as input, construct a thermal-humidity coupling prediction model, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion; S5: When there is a region in the comprehensive risk index map that exceeds the comprehensive risk judgment threshold, an intervention command is generated based on the risk type and location characteristics to control the robot end effector to perform differentiated active intervention actions.

2. The robot intelligent control method based on multi-source data fusion according to claim 1, characterized in that, The controlled exciter applies sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, inducing a breathing effect in the hidden cracks at the bonding interface. Simultaneously, it collects surface temperature sequence data, three-dimensional topographic point cloud data, ambient temperature and humidity data, and exciter reference signals, including: a fixed frequency output from the exciter mounted on the robot's end effector. The sinusoidal vibration excitation acts on the finishing layer-adhesive layer-base layer structure, causing the hidden debonding cracks at the adhesive interface to periodically open and close under excitation, thereby modulating the surface temperature field. A high frame rate infrared thermal imager fixed to the same platform as the exciter continuously collects the surface temperature distribution of the finishing layer to obtain the original surface temperature sequence data. A laser profilometer fixed to the same platform as the high frame rate infrared thermal imager synchronously acquires the three-dimensional topographic point cloud data of the wall surface. A temperature and humidity sensor mounted on the robot records the indoor air temperature in real time. and relative humidity The ambient temperature and humidity data are recorded; at the same time, the sinusoidal signal output by the exciter is recorded as the exciter reference signal.

3. The robot intelligent control method based on multi-source data fusion according to claim 1, characterized in that, The process of using the three-dimensional topography point cloud data to perform geometric correction and periodic averaging on the surface temperature sequence data, and using the exciter reference signal as a reference to perform phase-locked analysis on the corrected temperature sequence to extract normalized amplitude attenuation and phase delay, includes: calculating the surface unit normal, distance, and viewpoint factor corresponding to each pixel using the three-dimensional topography point cloud data; performing emissivity compensation and distance attenuation correction on the surface temperature sequence data to eliminate artifacts introduced by geometric factors; and aligning, stacking, and averaging the corrected surface temperature sequence data according to the excitation period to obtain a single-period representative temperature sequence. Based on the three-dimensional topographic point cloud data, local flatness information of the wall surface is extracted, and areas with good tiling quality are selected as reference areas. Using the exciter reference signal as a reference, digital phase-locked loop operation is performed on the single-cycle representative temperature sequence of each pixel to extract in-phase and quadrature components, and then the amplitude and phase are calculated. Reference points are selected in the reference area to obtain reference amplitude and reference phase, and the normalized amplitude attenuation and phase delay of each pixel are calculated, wherein the normalized amplitude attenuation is the ratio of the amplitude of each pixel to the reference amplitude, and the phase delay is the difference between the phase of each pixel and the reference phase.

4. The robot intelligent control method based on multi-source data fusion according to claim 1, characterized in that, Based on the phase delay, the known design embedment depth of the self-tapping screw, and the frequency of the sinusoidal vibration excitation, the thermal diffusivity is inverted pixel by pixel to construct a thermal inertia map. The normalized amplitude attenuation is used to verify the consistency of the inversion results, including: the thermal diffusivity inversion is based on the formula... Performed pixel by pixel, where pixel position Thermal diffusivity at that location The frequency of the sinusoidal vibration excitation is... Design the embedment depth for the self-tapping screw. The phase delay is defined as follows: when the absolute value of the phase delay of a pixel is less than a preset minimum threshold, the pixel is recorded as an invalid pixel; the consistency of the inverted thermal diffusivity is verified by using the theoretical relationship between the normalized amplitude attenuation and thermal diffusivity; the theoretical value of the amplitude obtained from the phase inversion is compared with the measured normalized amplitude attenuation; pixels exceeding the preset error threshold are judged as unreliable pixels and are filled by interpolation of the neighboring valid pixels; pixels with differences not exceeding the preset error threshold are retained to obtain a reliable thermal diffusivity distribution map, which constitutes the thermal inertia map.

5. The robot intelligent control method based on multi-source data fusion according to claim 4, characterized in that, The process involves using the theoretical relationship between normalized amplitude attenuation and thermal diffusivity to perform consistency verification on the inverted thermal diffusivity. This includes comparing the theoretical amplitude value obtained from phase inversion with the measured normalized amplitude attenuation. Pixels exceeding a preset error threshold are identified as unreliable pixels and interpolated using neighboring valid pixels. Pixels with differences not exceeding the preset error threshold are retained. This includes: the theoretical relationship between normalized amplitude attenuation and thermal diffusivity is... ,in pixel position Theoretical value of amplitude at that location, Design the embedment depth for the self-tapping screw. The frequency of the sinusoidal vibration excitation is... The thermal diffusivity is defined as follows: For each valid pixel, the relative deviation between the theoretical amplitude value and the measured normalized amplitude attenuation is calculated; the relative deviation is compared with the preset error threshold; when the relative deviation is greater than the preset error threshold, the pixel is determined to be an unreliable pixel; with the unreliable pixel as the center, valid pixels in the neighboring region that have not been determined to be unreliable pixels are searched; and the average thermal diffusivity of the valid pixels is taken as the interpolation padding value of the unreliable pixel.

6. The robot intelligent control method based on multi-source data fusion according to claim 1, characterized in that, The process involves using the thermal inertial map and ambient temperature and humidity data as input to construct a thermal-humidity coupling prediction model. This model calculates the heat accumulation risk index, condensation risk index, and cumulative condensation amount over future periods, and weighted fusion to generate a comprehensive risk index map. The process includes: combining the thermal inertial map with the material's volumetric heat capacity to obtain the thermal conductivity distribution, and establishing a two-dimensional transient heat conduction finite element model of the wall surface; using the indoor air temperature extrapolated from the ambient temperature and humidity data as the indoor side boundary condition to solve for the future temperature field; calculating the product of the peak-to-peak temperature of each pixel and the time-averaged temperature gradient modulus within a temperature cycle as the heat accumulation risk index; calculating the dew point temperature using the indoor air temperature and relative humidity from the ambient temperature and humidity data, and accumulating the dew point temperature in each pixel within the future temperature field. The duration for which the surface temperature is below the dew point temperature is used as a condensation risk indicator. A regression model between thermal diffusivity and water vapor permeability is established through prefabricated specimen experiments. The water vapor permeability distribution is obtained by mapping the thermal inertia map, and a steady-state diffusion model of water vapor is established. The water vapor partial pressure field is solved, and the water vapor flux vector is calculated based on the gradient of the water vapor partial pressure field, and the humidification flow line is drawn. The local condensation rate of the cold bridge condensation zone is calculated and integrated within the predicted time range to obtain the cumulative condensation amount. The thermal accumulation risk indicator, condensation risk indicator, and cumulative condensation amount are normalized to preset intervals and then weighted and summed with preset weights to generate a comprehensive risk index map, wherein the preset weights are determined based on the contribution of each indicator to the actual damage results in historical samples.

7. The robot intelligent control method based on multi-source data fusion according to claim 6, characterized in that, The calculation of dew point temperature using indoor air temperature and relative humidity data from the ambient temperature and humidity data, and the accumulation of the duration for which the surface temperature of each pixel is below the dew point temperature in the future temperature field as a condensation risk indicator, includes: the dew point temperature. According to the formula Calculation, where , and It is a constant. The indoor air temperature is... The relative humidity; the future temperature field obtained by solving the two-dimensional transient heat conduction finite element model on the wall surface. In the middle, regarding the prediction time range Every moment within Determine the surface temperature of each pixel Is it below the stated dew point temperature? The aforementioned condensation risk indicators Defined as ,in For indicator functions, when When the value is 1, The value is 0. To predict the duration of time.

8. The robot intelligent control method based on multi-source data fusion according to claim 1, characterized in that, When there are regions in the comprehensive risk index map that exceed the comprehensive risk judgment threshold, an intervention command is generated based on the risk type and location characteristics to control the robot's end effector to perform differentiated active intervention actions. These actions include: identifying regions in the comprehensive risk index map where the comprehensive risk value is greater than or equal to the comprehensive risk judgment threshold as high-risk areas requiring intervention; when a high-risk area simultaneously meets multiple intervention conditions, the intervention actions are executed in the priority order of microporous grouting, heat insulation and anti-condensation coating spraying, and waterproof vapor interface agent spraying, and the corresponding risk index decreases to the comprehensive risk relief threshold after the previous intervention action is completed. In the following situations, no further intervention actions will be performed; for areas located at the center of self-tapping screws where the heat accumulation risk index and cumulative condensation amount are co-located, a micro-hole grouting command will be generated, controlling the end drilling device to drill micro-holes in the finishing layer directly above the screw head to the debonding interface, and low-pressure injection of repair grout matching the thermal conductivity of the substrate to fill the debonding cracks and restore thermal contact; for screw head areas with extremely high condensation risk index but no hollow areas, a heat insulation and anti-condensation coating spraying command will be generated, controlling the end spraying device to apply the heat insulation and anti-condensation coating; for board seam areas where humid airflow lines converge densely and there are no self-tapping screws, a waterproof vapor interface agent spraying command will be generated, controlling the end spraying device to spray the waterproof vapor interface agent.

9. A robot intelligent control method based on multi-source data fusion according to claim 8, characterized in that, For the area located at the center of the self-tapping screw where the heat accumulation risk index and cumulative condensation amount coexist, a micro-hole grouting command is generated. The terminal drilling device is controlled to drill micro-holes in the finishing layer directly above the screw head to the debonding interface. A repair grout with a thermal conductivity matching the substrate is injected under low pressure to fill the debonding cracks and restore thermal contact. This includes: locating the self-tapping screw head position using the three-dimensional topographic point cloud data; determining the precise coordinates of each screw's center using the self-tapping screw position coordinates provided in the construction drawings; and identifying areas in the comprehensive risk index map that simultaneously meet the criteria of a heat accumulation risk index exceeding a preset heat accumulation threshold and a cumulative condensation amount exceeding a preset threshold. The condensation threshold area is determined as the area where the heat accumulation risk index and the cumulative condensation amount coexist. The precise coordinates of the high-risk area requiring intervention and the center of the self-tapping screw, along with the coexisting area, are spatially superimposed and analyzed to screen out the target intervention point located at the center of the self-tapping screw and coexisting with it. A micro-hole grouting command is generated for the target intervention point. The micro-hole grouting command controls the end drilling device to drill a micro-hole in the finishing layer directly above the screw head. The drilling depth reaches the designed embedment depth of the self-tapping screw. The grouting device is controlled to inject repair grout that matches the thermal conductivity of the substrate at a preset low pressure to fill the debonding crack and restore thermal contact.

10. A robot intelligent control system based on multi-source data fusion, used to implement the robot intelligent control method based on multi-source data fusion as described in any one of claims 1-9, characterized in that, Also includes: The excitation and multi-source sensing module is used to control the vibrator to apply sinusoidal vibration excitation at a fixed frequency to the surface of the finishing layer, induce the hidden cracks at the bonding interface to produce a breathing effect, and simultaneously collect surface temperature sequence data, three-dimensional topography point cloud data, ambient temperature and humidity data and vibrator reference signal. The data preprocessing and phase-locked feature extraction module is used to perform geometric correction and periodic averaging on the surface temperature sequence data using the three-dimensional topography point cloud data, and to perform phase-locked analysis on the corrected temperature sequence based on the exciter reference signal to extract normalized amplitude attenuation and phase delay. The thermal inertial inversion and map construction module is used to invert the thermal diffusivity pixel by pixel based on the phase delay, the known self-tapping screw design burial depth and the frequency of the sinusoidal vibration excitation, construct the thermal inertial map, and use the normalized amplitude attenuation to verify the consistency of the inversion results. The thermal-humidity coupling prediction and risk fusion module is used to construct a thermal-humidity coupling prediction model with the thermal inertia map and environmental temperature and humidity data as input, calculate the heat accumulation risk index, condensation risk index and cumulative condensation amount in the future period, and generate a comprehensive risk index map by weighted fusion. The integrated decision-making and active intervention control module is used to generate intervention commands based on risk type and location characteristics when there are areas in the comprehensive risk index map that exceed the comprehensive risk judgment threshold, and control the robot end effector to perform differentiated active intervention actions. The robot end effector module is used to receive intervention commands generated by the fusion decision and active intervention control module, and control the drilling device, grouting device and spraying device to perform corresponding active intervention actions according to the intervention commands, including micro-hole grouting, heat insulation and anti-condensation coating spraying and waterproof vapor interface agent spraying.

Images