A multimodal data acquisition robot dog inspection method and system

Abstract

The present application relates to the technical field of intelligent inspection and blockchain, and more particularly to a multi-modal data acquisition robot dog inspection method and system, the method comprising: collecting equipment state parameters and inspection multi-source data of a target boiler house; generating a basic inspection scheme according to the equipment state parameters; inspecting the target boiler house according to the basic inspection scheme, and correcting the basic inspection scheme according to environmental risk factor data during the inspection process, and optimizing the correction process according to equipment operation voiceprint data, the optimization including attenuating a path gain coefficient according to an equipment failure risk index, and redirecting the corrected inspection path according to an abnormal equipment azimuth; obtaining segmented inspection data during the inspection process, and generating a blockchain storage scheme according to the segmented inspection data; and uploading and storing the segmented inspection data according to the blockchain storage scheme.

Description

Technical Field

[0001] This invention relates to the fields of intelligent inspection and blockchain technology, and in particular to a multimodal data acquisition robot dog inspection method and system. Background Technology

[0002] As a core infrastructure for urban heating, chemical production, and energy supply, the safe operation of industrial boiler rooms is directly related to public safety and energy security. Traditional boiler room inspections rely on regular manual checks, which have problems such as long inspection intervals, high risk of exposure to hazardous environments, easy tampering of data records, and difficulty in tracing equipment status. With the development of embodied intelligence and blockchain technology, robot dog inspection has become an important direction to replace manual inspection. However, the existing technology has the following defects: First, the inspection path of robot dogs is mostly a pre-programmed fixed trajectory, which cannot be dynamically adjusted according to the real-time environmental hazard factors in the boiler room. When harmful gas leaks or local overheating occurs, the fixed path may cause the robot dog to enter a high-risk area or miss key monitoring points. Second, the existing environmental hazard monitoring mostly relies on the threshold alarm of a single gas sensor, lacking multi-dimensional risk fusion and quantification, and cannot distinguish between normal combustion fluctuations and real leakage risks, resulting in a high false alarm rate or missed detection. Third, the path calibration and equipment failure warning during the inspection process are independent of each other. Local path deviation may cause the robot dog to move away from equipment that has already shown abnormal bearing noise or unstable combustion, resulting in a disconnect between failure warning and on-site verification. Fourth, the inspection data storage is disconnected from the actual status and ownership of the equipment. It fails to dynamically adjust the blockchain consensus strength according to the aging of the equipment, and does not consider the difference in the credibility of the storage of equipment with frequent changes in ownership. This results in a mismatch between the anti-tampering level of on-chain data and the risk level of the equipment, insufficient guarantee of the authenticity of platform reports, and a lack of identity management of embodied intelligent agents.

[0003] Chinese patent CN110125953A discloses a cable inspection robot and a cable inspection method. However, this solution still has the problem that the sensing component only performs single data collection and uploading, and does not establish a multi-dimensional risk fusion and quantification model based on the rate of exceedance of harmful gas concentration, surface temperature rise rate, hot spot gradient ratio, and dangerous area coverage. Furthermore, it does not incorporate the dynamic index of inspection risk into the path calibration decision-making process, resulting in the fixed marker navigation path being unable to be dynamically adjusted according to real-time environmental hazard factors, leading to a high rate of false alarms and false negatives in environmental risks. Chinese patent with publication number CN111682446A discloses a cable inspection robot dog and a cable inspection method. However, this solution still focuses only on the lighting angle adjustment of the mechanical structure, does not introduce a device operation soundprint analysis mechanism based on bearing abnormality and combustion instability, and does not incorporate the device failure risk index into the optimization closed loop of the path calibration process. This results in the problem that local path deviation and device failure warning are independent of each other, and abnormal devices are out of the monitoring range. Chinese patent CN121291627A discloses a railway communication cable inspection robot dog. However, this solution still has the following problems: the perception module and the decision module are only designed for general inspection needs. It does not establish a blockchain evidence generation mechanism based on the cumulative value of the inspection trajectory, the highest risk level, and the response delay time. Furthermore, it does not include the equipment aging index and the equipment ownership coefficient in the dynamic adjustment and correction closed loop of the consensus node number. This results in a mismatch between the anti-tampering level of on-chain data and the equipment risk level, insufficient authenticity of platform reports, and a lack of identity management for the embodied intelligent agent. Summary of the Invention

[0004] To address these issues, the present invention provides a multimodal data acquisition robot dog inspection method and system, which overcomes the problems in the prior art where fixed paths fail due to real-time changes in environmental hazard factors during boiler room inspections, insufficient multi-dimensional risk fusion and quantification, disconnect between path calibration and equipment failure early warning, and disconnect between inspection data storage and equipment aging and ownership status.

[0005] To achieve the above objectives, in one aspect, the present invention provides a multimodal data acquisition robot dog inspection method, comprising: Step S1: Collect equipment status parameters and multi-source inspection data for the target boiler room; Step S2: Generate a basic inspection plan based on equipment status parameters; Step S3: Conduct an inspection of the target boiler room according to the basic inspection plan, and execute steps S4 and S5 during the inspection process; Step S4: Verify the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data; Step S5: Optimize the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data. The optimization includes attenuating the path gain coefficient based on the equipment failure risk index and redirecting the calibrated inspection path based on the azimuth angle of abnormal equipment. Step S6: During the inspection process, segmented inspection data is acquired, and a blockchain evidence storage scheme is generated based on the segmented inspection data. Step S7: Obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index; Step S8: Correct the adjustment process of the blockchain evidence storage scheme according to the device ownership coefficient in the device status parameters. The correction includes reducing the node adjustment coefficient when the device ownership coefficient is less than or equal to the ownership benchmark value, and increasing the node adjustment coefficient when the device ownership coefficient is greater than the ownership benchmark value. Step S9: The segmented inspection data is stored on the blockchain according to the blockchain evidence storage scheme.

[0006] Further, step S4 includes: Step S41: Perform time-series alignment and outlier removal on the environmental hazard factor data to obtain standardized environmental hazard factor data; Step S42: Perform multi-dimensional risk quantification on the standardized environmental hazard factor data to obtain an environmental risk vector; the multi-dimensional risk quantification includes extracting concentration features, temperature rise features and spatial features, the concentration features include the rate of exceedance of harmful gas concentrations, the temperature rise features include the surface temperature rise rate and hot spot gradient ratio, and the spatial features include the coverage rate of hazardous areas. Step S43: Calculate the dynamic index of inspection risk based on the environmental risk vector, and calibrate the basic inspection plan based on the dynamic index of inspection risk to obtain the calibrated inspection plan.

[0007] Furthermore, in step S41, when performing time-series alignment and outlier removal on the environmental hazard factor data, a sliding window midpoint filtering algorithm is used to remove outliers from the environmental hazard factor data, resulting in outlier-removed environmental hazard factor data. The window width of the sliding window midpoint filtering algorithm is 5 sampling periods. A timestamp synchronization algorithm is then used to perform time-series alignment on the outlier-removed environmental hazard factor data, resulting in standardized environmental hazard factor data. The synchronization accuracy of the timestamp synchronization algorithm is 10 milliseconds.

[0008] Further, in step S42, when performing multi-dimensional risk quantification on the standardized environmental hazard factor data, the exceedance rate R of the hazardous gas concentration is calculated using the formula R=∑(Ci-C0i) / C0i / Nc, where Ci is the measured concentration of the hazardous gas at the i-th sampling point, C0i is the standard safe concentration threshold for the corresponding hazardous gas, and Nc is the total number of sampling points; the surface temperature rise rate Vr is calculated using the formula Vr=(Tt-Tt-Δt) / Δt, where Tt is the current surface temperature, Tt-Δt is the surface temperature before time Δt, and Δt is the sampling... The interval; the hotspot gradient ratio G is obtained by collecting the boiler surface temperature field through an infrared thermal imager array, extracting the highest temperature point Thot and the average temperature point Tavg, and then calculating it using the formula G=(Thot-Tavg) / Tavg; the dangerous area coverage A is obtained by performing three-dimensional reconstruction of the boiler room space through lidar point cloud data, identifying obstacle boundaries and dangerous restricted areas, and then calculating it using the formula A=Sdanger / Stotal, where Sdanger is the projected area of ​​the dangerous restricted area on the inspection path, and Stotal is the total coverage area of ​​the inspection path.

[0009] Further, step S43 calculates the dynamic index of inspection risk based on the environmental risk vector, and verifies the basic inspection plan based on the dynamic index of inspection risk, specifically including: The inspection risk dynamic index Qr is calculated based on the harmful gas concentration exceedance rate R, surface temperature rise rate Vr, hot spot gradient ratio G, and hazardous area coverage A. Qr is set as Qr = w1 × (R / R0) + w2 × (Vr / V0) + w3 × (G / G0) + w4 × (A / A0), where R0 is the baseline exceedance rate, V0 is the baseline temperature rise rate, G0 is the baseline hot spot gradient, A0 is the baseline coverage, w1 is the first risk weight, w2 is the second risk weight, w3 is the third risk weight, and w4 is the fourth risk weight. The inspection risk dynamic index Qr is compared with the preset risk value Qr0. Based on the comparison result, the inspection risk status is judged, and the basic inspection plan is checked based on the judgment result. When Qr≤Qr0, the inspection risk status is determined to be controllable, and the basic inspection plan is not checked. When Qr > Qr0, the inspection risk status is determined to be out of limit, and the basic inspection plan is checked. The check includes: calculating the path offset ΔP = Rg × (Qr - Qr0) to obtain the checked inspection path L1 = L0 + ΔP, where L0 is the basic inspection path, Rg is the path gain coefficient, and the basic inspection path L0 in the basic inspection plan is replaced with the checked inspection path L1.

[0010] Furthermore, step S5 optimizes the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data, specifically including: Based on the equipment operation acoustic data, the bearing anomaly degree Eb and combustion instability Eu are extracted. Then, based on the bearing anomaly degree Eb, combustion instability Eu, first acoustic weight φ1, and second acoustic weight φ2, the equipment failure risk index Ds is calculated, set as Ds = φ1 × (Eb / Eb0) + φ2 × (Eu / Eu0), where Eb0 is the baseline bearing anomaly degree and Eu0 is the baseline combustion instability degree. The equipment failure risk index Ds is compared with the preset failure value Ds0. Based on the comparison result, the equipment failure risk is judged, and based on the judgment result, the calibration process of the basic inspection plan is optimized. When Ds≤Ds0, the risk of equipment failure is determined to be controllable, and the calibration process of the basic inspection plan is not optimized. When Ds > Ds0, the equipment failure risk is determined to be an over-limit risk. The calibration process of the basic inspection plan is optimized. The optimization includes: adjusting the path gain coefficient Rg to Rg', setting Rg' = Rg / (1 + Rd × (Ds - Ds0)), where Rd is the attenuation coefficient; calculating the redirection angle correction Δθ = Ka × (Ds - Ds0) × sign(θtarget - θcurrent), to obtain the redirected inspection path L2, where L2 keeps the radial distance of L1 unchanged, the azimuth angle is corrected to θcurrent + Δθ, θtarget is the azimuth angle of the abnormal equipment, θcurrent is the current azimuth angle of the robot dog, Ka is the redirection gain coefficient, and sign is the sign function.

[0011] Furthermore, in step S6, during the inspection process, segmented inspection data is acquired, and a blockchain evidence storage scheme is generated based on the segmented inspection data, specifically including: The cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr in the segmented inspection data are obtained. Based on the cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr, the basic evidence storage level Cl and the basic consensus node number Cv are calculated. Cl = round(Cbase + m1 × (W / Wbase) + m2 × (Rmax / Rbase)) and Cv = ceil(Cvbase × [1 + n1 × (tr / tbase)]), where Cbase is the baseline evidence storage level, Wbase is the baseline inspection trajectory, Rbase is the baseline risk level, Cvbase is the baseline consensus node number, tbase is the baseline response time, m1 is the correlation coefficient of the first blockchain, m2 is the correlation coefficient of the second blockchain, n1 is the correlation coefficient of the third blockchain, round is the rounding function, and ceil is the rounding function. This yields a blockchain evidence storage scheme that includes the basic evidence storage level Cl and the basic consensus node number Cv.

[0012] Further, step S7 involves obtaining the device aging index and adjusting the blockchain evidence storage scheme based on the device aging index, specifically including: The aging index of the equipment is obtained by using the vibration acceleration sensor mounted on the robot dog, and the aging deviation value Ag is obtained. The aging deviation value Ag is compared with the preset aging value Ag0. Based on the comparison result, the aging status of the equipment is judged, and the blockchain evidence storage scheme is adjusted according to the judgment result, wherein: When Ag≤Ag0, the device aging status is determined to be normal, and no adjustment is made to the blockchain evidence storage scheme; When Ag > Ag0, the device is determined to be in a deteriorated state, and the blockchain evidence storage scheme is adjusted. The adjustment includes: calculating the consensus node increment ΔC = Jc × (Ag - Ag0) to obtain the adjusted number of consensus nodes C1 = Cv + ΔC, where Cv is the basic consensus node number, Jc is the node adjustment coefficient, and the basic consensus node number Cv in the blockchain evidence storage scheme is replaced with the adjusted number of consensus nodes C1.

[0013] Further, step S8 modifies the adjustment process of step S7 based on the equipment ownership coefficient in the equipment status parameters, specifically including: The device ownership coefficient Ow is calculated based on the device ownership chain length D and ownership change frequency f in the device status parameters, and is set as Ow=(D / D0)×(f / f0), where D0 is the baseline ownership chain length and f0 is the baseline change frequency. The device ownership coefficient Ow is compared with the preset ownership value Ow0, and the device ownership status is judged based on the comparison result. The adjustment process of the blockchain evidence storage scheme is then corrected based on the judgment result. When Ow≤Ow0, the ownership status of the device is determined to be clear, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc'=Jc×(1-Kf×(Ow0-Ow) / Ow0), where Kf is the ownership correction coefficient. When Ow > Ow0, the device ownership status is determined to be complex, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc' = Jc × (1 + Kf × (Ow - Ow0) / Ow0), where Kf is the ownership correction coefficient.

[0014] On the other hand, the present invention also provides a system for a multimodal data acquisition robot dog inspection method, comprising: The multi-source acquisition module is used to collect equipment status parameters and inspection data from multiple sources in the target boiler room. The inspection plan generation module is used to generate basic inspection plans based on equipment status parameters; The plan push module is used to push basic inspection plans; The inspection feedback module is used to calibrate the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data, optimize the calibration process of the basic inspection plan based on the equipment operation soundprint data in the multi-source inspection data, acquire segmented inspection data during the inspection process, and generate a blockchain evidence storage plan based on the segmented inspection data. The evidence storage correction module is used to obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index. It is also used to correct the adjustment process of the blockchain evidence storage scheme according to the equipment ownership coefficient in the equipment status parameters. The evidence storage solution push module is used to push blockchain evidence storage solutions.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the method collects equipment status parameters and multi-source inspection data of the target boiler room in step S1, providing a data foundation for subsequent control; the method also generates a basic inspection plan based on the equipment status parameters in step S2, solving the static setting deviation problems of inspection path dependence on pre-programmed fixed trajectories and failure to consider the actual layout and thermophysical characteristics of equipment in the prior art, making the initial inspection path adaptable to operating conditions; the method also executes steps S4 and S5 in real time during the inspection process in step S3, solving the problems of open-loop control and inability to respond in real time to changes in hazardous environmental conditions in the prior art. To address the issue of risk fusion and achieve dynamic closed-loop control of the inspection process, the method further resolves, through step S4, the technical problems in existing technologies where false alarms and missed alarms are caused by threshold judgments from a single sensor, and where it is difficult to quantify and perceive hazardous environmental conditions in real time. By integrating multi-dimensional risks, the method improves the accuracy and reliability of inspection risk monitoring, thereby enhancing the safety of the inspection path. Furthermore, through step S5, the method optimizes the calibration process in step S4 based on equipment operation acoustic data, resolving the technical problems in existing technologies where local path calibration and equipment failure early warning are independent, and path deviation causes abnormal equipment to fall outside the monitoring range. The method adjusts the path calibration intensity through equipment failure risk feedback and, through... The method utilizes equipment azimuth angle retargeting to achieve coordinated optimization of risk avoidance and proximity, enhancing equipment failure early warning capabilities. Furthermore, in step S6, a blockchain evidence storage scheme is generated based on segmented inspection data during the inspection process, addressing the technical problem in existing technologies where the blockchain evidence storage scheme is disconnected from the actual risk level of the inspection, thus achieving real-time correlation between the inspection process and the evidence storage process. In step S7, the method further adjusts the blockchain evidence storage scheme based on the equipment aging index, resolving the technical problem in existing technologies where the evidence storage consensus strength is fixed and cannot be specifically strengthened according to the actual aging state of the equipment. Finally, in step S8, the method adjusts step S7 based on the equipment ownership coefficient. The adjustment process is modified to address the technical problems in existing technologies, such as failing to consider the differences in the credibility of evidence storage for devices with different ownership change frequencies and applying a one-size-fits-all approach to evidence storage adjustments. By reducing the node coefficient to lower consensus costs when ownership is clear and increasing the node coefficient to enhance anti-tampering capabilities when ownership is complex, the method achieves precise evidence storage control for devices with different ownership complexities. The method also performs on-chain evidence storage according to the blockchain evidence storage scheme in step S9, which solves the technical problems in existing technologies, such as the ease with which evidence storage data can be tampered with and the insufficient authenticity of platform reports. Ultimately, a boiler room inspection evidence storage system with authentic and reliable data and controllable intelligent agent identity is obtained, realizing a closed-loop collaboration between inspection quality and blockchain evidence storage. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the multimodal data acquisition robot dog inspection method of this embodiment; Figure 2 This is a flowchart illustrating step S4 of this embodiment; Figure 3 This is a schematic diagram of the system structure of the multimodal data acquisition robot dog inspection method in this embodiment. Detailed Implementation

[0017] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0018] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0019] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0020] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] Please see Figure 1 As shown, this is a flowchart illustrating the multimodal data acquisition robot dog inspection method of this embodiment. The method includes: Step S1: Collect equipment status parameters and multi-source inspection data for the target boiler room; Step S2: Generate a basic inspection plan based on equipment status parameters; Step S3: Conduct an inspection of the target boiler room according to the basic inspection plan, and execute steps S4 and S5 during the inspection process; Step S4: Verify the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data; Step S5: Optimize the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data. The optimization includes attenuating the path gain coefficient based on the equipment failure risk index and redirecting the calibrated inspection path based on the azimuth angle of abnormal equipment. Step S6: During the inspection process, segmented inspection data is acquired, and a blockchain evidence storage scheme is generated based on the segmented inspection data. Step S7: Obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index; Step S8: Correct the adjustment process of the blockchain evidence storage scheme according to the device ownership coefficient in the device status parameters. The correction includes reducing the node adjustment coefficient when the device ownership coefficient is less than or equal to the ownership benchmark value, and increasing the node adjustment coefficient when the device ownership coefficient is greater than the ownership benchmark value. Step S9: The segmented inspection data is stored on the blockchain according to the blockchain evidence storage scheme.

[0022] Specifically, the multimodal data acquisition robot dog inspection method is applied to intelligent safety inspection scenarios in urban centralized heating boiler rooms, chemical industrial park steam boiler rooms, or power plant auxiliary boiler rooms. Through multi-dimensional dynamic coupling analysis of equipment status parameters, environmental hazard factor data, equipment operation acoustic data, and segmented inspection data, it achieves collaborative optimization of real-time control and blockchain-based evidence storage during the inspection process. The method collects equipment status parameters and multi-source inspection data from the target boiler room in step S1, providing a data foundation for subsequent control. The method also generates a basic inspection plan based on the equipment status parameters in step S2, addressing the static nature of existing technologies that rely on pre-programmed fixed trajectories and fail to consider the actual layout and thermophysical characteristics of the equipment. By setting deviation issues to make the initial inspection path adaptable to operating conditions, the method further addresses the problems of open-loop control and inability to respond in real time to changes in environmental hazards in existing technologies through step S3, thereby achieving dynamic closed-loop control of the inspection process. Step S4 also solves the technical problems of false alarms and missed alarms caused by single sensor threshold judgments and the difficulty in real-time quantification of environmental hazards in existing technologies. Multi-dimensional risk fusion improves the accuracy and reliability of inspection risk monitoring, thereby enhancing the safety of the inspection path. Step S5 further optimizes the calibration process of step S4 based on equipment operating acoustic data, resolving the issues of local path calibration and equipment... The method addresses the technical issues of independent failure warnings and path deviations causing abnormal equipment to fall outside the monitoring range. It adjusts the path calibration strength through equipment failure risk feedback and achieves coordinated optimization of risk avoidance and proximity through abnormal equipment azimuth angle redirection, thereby enhancing the equipment failure warning capability. Furthermore, in step S6, a blockchain evidence storage scheme is generated based on segmented inspection data during the inspection process, resolving the technical problem of existing blockchain evidence storage schemes being disconnected from the actual risk level of the inspection. This achieves real-time correlation between the inspection process and the evidence storage process. Finally, in step S7, the method adjusts the blockchain evidence storage scheme based on the equipment aging index, addressing the technical problem in existing technologies where the evidence storage consensus strength is fixed and cannot be specifically strengthened according to the actual aging state of the equipment. To address the problem, the method further modifies the adjustment process in step S7 based on the equipment ownership coefficient in step S8, resolving the technical issues of existing technologies that fail to consider the differences in the credibility of evidence storage for devices with different ownership change frequencies and that evidence storage adjustments are applied in a one-size-fits-all manner. By reducing the node coefficient when ownership is clear to lower consensus costs and increasing the node coefficient when ownership is complex to enhance anti-tampering capabilities, the method achieves precise evidence storage control for devices with different ownership complexities. Furthermore, the method performs on-chain evidence storage based on a blockchain evidence storage scheme in step S9, resolving the technical issues of easily tampered evidence storage data and insufficient authenticity of platform reports in existing technologies. Ultimately, the method obtains a boiler room inspection evidence storage system with authentic and reliable data and controllable intelligent agent identity, achieving closed-loop collaboration between inspection quality and blockchain evidence storage.

[0023] Specifically, the equipment status parameters include boiler model, rated power P, years in operation Y, thermal efficiency η, and equipment ownership chain length D. The boiler model refers to the national standard model identifier for industrial boilers, such as WNS4-1.25-Q. The rated power P refers to the boiler's rated heat output power, in megawatts (MW). The years in operation Y refers to the cumulative number of years the boiler has been in operation from its initial commissioning to the current inspection date, in years. The thermal efficiency η refers to the ratio of the boiler's effective heat utilization to its input heat, in percentage. The equipment ownership chain length D refers to the number of ownership transfers recorded on the blockchain since the boiler left the factory, representing the ownership complexity. Step S1 collects the boiler model, rated power, years in operation, and thermal efficiency by reviewing the boiler quality certificate and operating records. Step S1 also queries the boiler through a blockchain explorer. The NFT ownership token transfer record is used to collect the equipment ownership chain length D. The multi-source inspection data includes environmental hazard factor data and equipment operation soundprint data. Step S1 collects environmental hazard factor data through a multi-gas sensor array mounted on the robot dog. The multi-gas sensor array includes an electrochemical carbon monoxide sensor, a catalytic combustion combustible gas sensor, and an infrared absorption carbon dioxide sensor arranged at the front of the robot dog. The sampling frequency is 1 Hz, and the data is aggregated through the CAN bus interface of the robot dog. Step S1 collects equipment operation soundprint data through a microphone array arranged on both sides of the robot dog. The microphone array includes 4 MEMS digital microphones arranged linearly and equally spaced at a distance of 5 cm. The sampling frequency is 48 kHz, and the soundprints of the boiler body and auxiliary equipment are collected in a directional manner through a beamforming algorithm.

[0024] Specifically, step S2 generates a basic inspection plan based on equipment status parameters, including: The basic inspection path L0, basic inspection frequency F0, and basic dwell time T0 are calculated based on the boiler model, rated power P, service life Y, and thermal efficiency η in the equipment status parameters. The basic inspection path L0 is calculated using the formula L0=a1×√(P / P0)×S, where a1 is the path calculation coefficient, P0 is the reference power, S is the boiler room building area, and √ is the square root. The basic inspection frequency F0 is calculated using the formula F0=a2×(1+Y / Y0)×(η0 / η), where a2 is the frequency calculation coefficient, Y0 is the reference year, and η0 is the reference thermal efficiency. The basic stay duration T0 is calculated using the formula T0=a3 / F0, where a3 is the duration coefficient; The basic inspection scheme is obtained, which includes the basic inspection path L0, the basic inspection frequency F0, and the basic dwell time T0.

[0025] Specifically, the basic inspection path refers to the initial inspection trajectory setting value determined based on the boiler power and boiler room area, in meters. The path calculation coefficient is an engineering experience proportional coefficient that converts power and area into path length. The path calculation coefficient is set to 15. The reason for this value is that, according to regression analysis of the industrial boiler room inspection process database, when the rated power is 4 MW and the boiler room area is 200 square meters, this coefficient can ensure that the basic inspection path length is approximately 150 meters, covering the key monitoring points of the boiler body, water supply system, combustion system, and flue gas system. This avoids monitoring omissions due to excessively short paths and prevents inspection cycles from exceeding limits due to excessively long paths. The reference power refers to the calculated... The reference power value for power normalization is 4. The reason for this reference power value is that it corresponds to the commonly used rated power specifications of small and medium-sized industrial boiler rooms, serving as a dimensionless benchmark to facilitate standardized conversion of inspection paths for boiler rooms of different power ratings. The boiler room building area refers to the net floor area inside the boiler room available for the robot dog to inspect, in square meters. The basic inspection frequency refers to the initial set value of the number of times the robot dog performs a complete inspection per unit time, in times per day. The frequency calculation coefficient is a proportional coefficient used to calculate the inspection frequency based on the direct proportionality between the years of operation and the inverse proportionality between thermal efficiency. The frequency calculation coefficient is 1. The reason for this frequency calculation coefficient is that it corresponds to the commonly used rated power specifications of small and medium-sized industrial boiler rooms, serving as a dimensionless benchmark to facilitate standardized conversion of inspection paths for boiler rooms of different power ratings. Under the conditions of a 10-year service life and a 90% baseline thermal efficiency, this coefficient ensures a basic inspection frequency of approximately twice daily. This ensures that the inspection interval balances response to changes in equipment status with the robot's battery life constraints, preventing excessively high frequency from depleting the battery or excessively low frequency from causing missed inspections. The baseline service life refers to the reference operational years value used when normalizing the calculated service life. The baseline service life is set to 10 years, based on the median service life of industrial boilers. This establishes a direct proportional relationship between operational years and inspection frequency; a longer service life results in a higher frequency to enhance monitoring of aging equipment. The baseline thermal efficiency refers to the reference thermal efficiency value used when correcting the calculated frequency, set to 90%. The following is a summary of the key points: The nominal design thermal efficiency of the new gas-fired boiler is used as a benchmark to adjust the inspection frequency of boilers with different thermal efficiencies. Inefficient boilers require increased inspection frequency to prevent combustion abnormalities. The basic dwell time refers to the initial time setting for the robot dog to collect data at each monitoring point, measured in seconds. The dwell time coefficient is a proportional coefficient of the required dwell time per inspection frequency, measured in seconds per inspection. The dwell time coefficient is set to 30. The reason for this value is that, based on the robot dog's multi-sensor data acquisition sequence, this coefficient ensures a dwell time of approximately 15 seconds per point at a daily inspection frequency of twice, matching the sensor sampling cycle and maintaining stable data acquisition quality and sufficient signal accumulation length.

[0026] Specifically, step S3 involves inspecting the target boiler room according to the basic inspection plan, which includes: deploying the robot dog at the entrance of the boiler room and establishing a wireless communication link with the edge computing gateway; setting initial motion parameters through the robot dog's motion control unit and setting the inspection trajectory through a path planning algorithm according to the basic inspection path, basic inspection frequency, and basic dwell time in the basic inspection plan; starting the inspection task, and the robot dog autonomously navigating and inspecting the boiler room along the inspection trajectory.

[0027] Please see Figure 2 As shown, this is a flowchart illustrating step S4 of this embodiment, which includes: Step S41: Perform time-series alignment and outlier removal on the environmental hazard factor data to obtain standardized environmental hazard factor data; Step S42: Perform multi-dimensional risk quantification on the standardized environmental hazard factor data to obtain an environmental risk vector; the multi-dimensional risk quantification includes extracting concentration features, temperature rise features and spatial features, the concentration features include the rate of exceedance of harmful gas concentrations, the temperature rise features include the surface temperature rise rate and hot spot gradient ratio, and the spatial features include the coverage rate of hazardous areas. Step S43: Calculate the dynamic index of inspection risk based on the environmental risk vector, and calibrate the basic inspection plan based on the dynamic index of inspection risk to obtain the calibrated inspection plan.

[0028] Specifically, in step S41, when performing time-series alignment and outlier removal on the environmental hazard factor data, a sliding window midpoint filtering algorithm is used to remove outliers from the environmental hazard factor data, resulting in outlier-removed environmental hazard factor data. The window width of the sliding window midpoint filtering algorithm is 5 sampling periods. A timestamp synchronization algorithm is then used to perform time-series alignment on the outlier-removed environmental hazard factor data, resulting in standardized environmental hazard factor data. The synchronization accuracy of the timestamp synchronization algorithm is 10 milliseconds.

[0029] Specifically, the sliding window median filtering algorithm refers to a nonlinear filtering method that replaces the center point value with the median of the data within a fixed-length sliding window to suppress pulse-type abnormal interference. This method is used to remove abnormal readings caused by occasional electromagnetic interference spikes and communication errors from multiple gas sensors. The timestamp synchronization algorithm refers to a data fusion preprocessing method that uses the machine dog's main control clock as a reference and aligns the sampling times of each sensor through interpolation. This embodiment does not limit the specific implementation of the timestamp synchronization algorithm for time-series alignment of environmental hazard factor data after outlier removal. For example, a timestamp synchronization algorithm using linear interpolation alignment or nearest neighbor interpolation alignment can be used for time-series alignment of environmental hazard factor data after outlier removal.

[0030] Specifically, in step S42, when performing multi-dimensional risk quantification on the standardized environmental hazard factor data, the exceedance rate R of the hazardous gas concentration is calculated using the formula R=∑(Ci-C0i) / C0i / Nc, where Ci is the measured concentration of the hazardous gas at the i-th sampling point, C0i is the standard safe concentration threshold of the corresponding hazardous gas, and Nc is the total number of sampling points; the surface temperature rise rate Vr is calculated using the formula Vr=(Tt-Tt-Δt) / Δt, where Tt is the current surface temperature, Tt-Δt is the surface temperature before time Δt, and Δt is the sampling... The interval; the hotspot gradient ratio G is obtained by collecting the boiler surface temperature field through an infrared thermal imager array, extracting the highest temperature point Thot and the average temperature point Tavg, and then calculating it using the formula G=(Thot-Tavg) / Tavg; the dangerous area coverage A is obtained by performing three-dimensional reconstruction of the boiler room space through lidar point cloud data, identifying obstacle boundaries and dangerous restricted areas, and then calculating it using the formula A=Sdanger / Stotal, where Sdanger is the projected area of ​​the dangerous restricted area on the inspection path, and Stotal is the total coverage area of ​​the inspection path.

[0031] Specifically, the measured concentration of harmful gases refers to the volumetric concentration value of harmful gases collected in real time by the multi-gas sensor array carried by the robot dog during the inspection process. This concentration is obtained by converting the chemical reaction of the gas into an electrical signal using an electrochemical sensor, followed by sampling via an analog-to-digital converter, and is expressed in ppm or mg / m³. The standard safe concentration threshold refers to the 8-hour time-weighted average permissible concentration as specified in GBZ2.1-2019 "Occupational Exposure Limits for Hazardous Factors in the Workplace". The total number of sampling points refers to the total number of locations where the robot dog samples the concentration of harmful gases within one inspection cycle. The current surface temperature refers to the highest temperature of the boiler surface collected by the infrared thermal imager array at the current sampling time. The values ​​are in degrees Celsius. The sampling interval refers to the time interval between two adjacent temperature samples, in seconds. The highest temperature point refers to the coordinates and temperature value of the extreme temperature point identified by the infrared thermal imager array in the boiler surface temperature field. The average temperature point refers to the arithmetic mean of the full field of view temperature calculated by the infrared thermal imager array in the boiler surface temperature field. The dangerous restricted area refers to the high-risk area in the boiler room space model reconstructed by lidar, which is designated as a prohibited area for robot dogs according to boiler safety regulations, such as the burner flame area and the high-pressure steam pipe rupture risk area. The total coverage area of ​​the inspection path refers to the bounding box area of ​​the robot dog's basic inspection path on the horizontal plane, in square meters.

[0032] Specifically, step S43 calculates the dynamic index of inspection risk based on the environmental risk vector, and calibrates the basic inspection plan based on the dynamic index of inspection risk, specifically including: The inspection risk dynamic index Qr is calculated based on the harmful gas concentration exceedance rate R, surface temperature rise rate Vr, hot spot gradient ratio G, and hazardous area coverage A. Qr is set as Qr = w1 × (R / R0) + w2 × (Vr / V0) + w3 × (G / G0) + w4 × (A / A0), where R0 is the baseline exceedance rate, V0 is the baseline temperature rise rate, G0 is the baseline hot spot gradient, A0 is the baseline coverage, w1 is the first risk weight, w2 is the second risk weight, w3 is the third risk weight, and w4 is the fourth risk weight. The inspection risk dynamic index Qr is compared with the preset risk value Qr0. Based on the comparison result, the inspection risk status is judged, and the basic inspection plan is checked based on the judgment result. When Qr≤Qr0, the inspection risk status is determined to be controllable, and the basic inspection plan is not checked. When Qr > Qr0, the inspection risk status is determined to be out of limit, and the basic inspection plan is checked. The check includes: calculating the path offset ΔP = Rg × (Qr - Qr0) to obtain the checked inspection path L1 = L0 + ΔP, where L0 is the basic inspection path, Rg is the path gain coefficient, and the basic inspection path L0 in the basic inspection plan is replaced with the checked inspection path L1.

[0033] Specifically, the benchmark exceedance rate refers to the benchmark value of the exceedance rate of harmful gas concentration measured under standard safety conditions through boiler room safety assessment tests. The benchmark exceedance rate is set at 0.1. The reason for setting the benchmark exceedance rate at 0.1 is that it corresponds to the average deviation of the harmful gas concentration collected by the multi-gas sensor array in a standard boiler room environment from the safety threshold under normal ventilation conditions. This deviation is used as a normalization benchmark, ensuring that the R / R0 term in Qr is significantly greater than 1 when the exceedance is severe, and close to 1 under normal conditions. The benchmark temperature rise rate refers to the benchmark value of the surface temperature rise rate measured under standard safety conditions through boiler room safety assessment tests. The benchmark temperature rise rate is set at 0.5. The reason for setting the quasi-temperature rise rate to 0.5 is as follows: it corresponds to the stable rate of change of boiler surface temperature measured by an infrared thermal imager under rated load, in degrees Celsius per minute. This serves as the normalization benchmark, ensuring that the Vr / V0 term in Qr is significantly greater than 1 during temperature abrupt changes. The benchmark hot spot gradient refers to the ratio of the hot spot gradient measured under standard safety conditions through boiler room safety assessment tests to the benchmark value. The benchmark hot spot gradient is set to 0.05. The reason for setting the benchmark hot spot gradient to 0.05 is that it corresponds to the relative non-uniformity of surface temperature distribution in a standard boiler under uniform combustion conditions. This serves as the normalization benchmark, ensuring that the G / G0 term in Qr is significantly greater than 1 during localized overheating. Coverage rate refers to the baseline value of hazardous area coverage measured under standard safety conditions through boiler room safety assessment tests. The baseline coverage rate is set at 0.1. The reason for setting the baseline coverage rate at 0.1 is that it corresponds to the normal proportion of the hazardous restricted area area to the projected area of ​​the inspection path in a standard boiler room layout. Using this as a normalization benchmark ensures that the A / A0 term in Qr is significantly greater than 1 when spatial conflicts intensify. The first risk weight refers to the weighting coefficient assigned to the rate of exceedance of harmful gas concentration when calculating the dynamic index of inspection risk. The value of the first risk weight is 0.3. The reason for setting the first risk weight at 0.3 is that the rate of exceedance of harmful gas concentration directly reflects the risk of incomplete combustion or leakage. When the concentration exceeds the standard, there is a risk of explosion and... The risk of poisoning is considered, but this indicator is greatly affected by ventilation conditions, and relying solely on it may lead to excessive bias. Therefore, it is given a moderate weight. The second risk weight refers to the weighting coefficient assigned to the surface temperature rise rate value when calculating the dynamic index of inspection risk, with a value of 0.3. The reason for assigning a value of 0.3 to the second risk weight is that the surface temperature rise rate reflects sudden changes in boiler heat load and the degree of scaling. Excessive temperature rise indicates water circulation failure or burner abnormality. This characteristic is strongly correlated with equipment failure and complements the concentration characteristic, jointly characterizing environmental risk. Therefore, it is assigned a value of 0.3. The third risk weight refers to the weighting coefficient assigned to the hot spot gradient ratio when calculating the dynamic index of inspection risk, with a value of 0.2. The third risk weight is set to 0.The reasons for 2 are as follows: The hotspot gradient ratio reflects the risk of local overheating and material creep in the boiler, and is an auxiliary diagnostic indicator for structural safety. Although it can reflect abnormal conditions, its stability is lower than that of concentration and temperature rise characteristics due to the limitation of the sensor's field of view. Therefore, it is given a lower weight. The fourth risk weight refers to the weight coefficient assigned to the coverage rate of dangerous areas when calculating the dynamic index of inspection risk, and its value is 0.2. The reason for the fourth risk weight being 0.2 is that the coverage rate of dangerous areas reflects the degree of spatial conflict between the inspection path and the safety restricted area, and is a geometric constraint indicator for path safety. Although it can avoid the risk of physical collision, it belongs to static spatial information, and its dynamic risk response priority is lower than that of gas and temperature indicators. Therefore, it is given a lower weight. The preset risk value refers to the preset value for judging the inspection risk status, and its value is 1.0. The reason for the preset risk threshold being 1.0 is that, based on the statistical test of the safety inspection process window in industrial boiler rooms, when the dynamic index of inspection risk Qr exceeds 1.0, the actual environmental parameters deviate from the safety value by more than 20% of the comprehensive threshold, or the instrument detectable... The presence of harmful gas leaks and signs of localized overheating has exceeded the safety limits, necessitating a calibration action. The inspection risk status refers to the controllability of the boiler room environmental risk as determined by the inspection risk dynamic index and the preset risk value. The inspection risk status includes controllable and exceeding limits. The path gain coefficient is the proportional conversion coefficient between the inspection risk dynamic index deviation and the path deviation when calculating the inspection path offset. The path gain coefficient Rg is determined based on the rated power P, and its value ranges from 5 to 15 meters per unit deviation of Qr. The rationale for this path gain coefficient is as follows: Based on boiler room spatial layout experience, in small and medium-sized boiler rooms, an offset of 5 to 15 meters can avoid high-risk areas and cover backup monitoring points. A 0.1 unit change in the inspection risk dynamic index Qr corresponds to approximately a 10% change in safety margin. To ensure that the path offset ΔP does not exceed 5 meters when Qr deviates by 0.3 to avoid excessive deviation, while maintaining response sensitivity, Rg is set to 5 to 15 meters, with lower values ​​for small-power boiler rooms and higher values ​​for large-power boiler rooms.

[0034] Specifically, step S5 optimizes the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data, including: Based on the equipment operation acoustic data, the bearing anomaly degree Eb and combustion instability Eu are extracted. Then, based on the bearing anomaly degree Eb, combustion instability Eu, first acoustic weight φ1, and second acoustic weight φ2, the equipment failure risk index Ds is calculated, set as Ds = φ1 × (Eb / Eb0) + φ2 × (Eu / Eu0), where Eb0 is the baseline bearing anomaly degree and Eu0 is the baseline combustion instability degree. The equipment failure risk index Ds is compared with the preset failure value Ds0. Based on the comparison result, the equipment failure risk is judged, and based on the judgment result, the calibration process of the basic inspection plan is optimized. When Ds≤Ds0, the risk of equipment failure is determined to be controllable, and the calibration process of the basic inspection plan is not optimized. When Ds > Ds0, the equipment failure risk is determined to be an over-limit risk. The calibration process of the basic inspection plan is optimized. The optimization includes: adjusting the path gain coefficient Rg to Rg', setting Rg' = Rg / (1 + Rd × (Ds - Ds0)), where Rd is the attenuation coefficient; calculating the redirection angle correction Δθ = Ka × (Ds - Ds0) × sign(θtarget - θcurrent), to obtain the redirected inspection path L2, where L2 keeps the radial distance of L1 unchanged, the azimuth angle is corrected to θcurrent + Δθ, θtarget is the azimuth angle of the abnormal equipment, θcurrent is the current azimuth angle of the robot dog, Ka is the redirection gain coefficient, and sign is the sign function.

[0035] Specifically, this embodiment does not limit the method of extracting bearing anomaly degree Eb and combustion instability degree Eu based on the device operation acoustic data. Those skilled in the art can set it according to the actual situation. For example, by using microphone arrays arranged on both sides of the robot dog to directionally collect the acoustic data of boiler feedwater pump bearing and burner flame pulsation, the bearing anomaly degree can be obtained by extracting the energy ratio of the bearing characteristic frequency band through fast Fourier transform, and the combustion instability degree can be obtained by analyzing the combustion noise spectrum characteristics through Mel frequency cepstral coefficient analysis. The bearing anomaly degree refers to the ratio of the energy of the characteristic fault frequency band to the energy of the entire frequency band in the vibration acoustic data of the feedwater pump bearing, which characterizes the bearing wear degree and is dimensionless. The combustion instability refers to the ratio of low-frequency pulsation energy to steady-state combustion noise energy in the burner flame acoustic signature, characterizing the stability of combustion conditions. It is dimensionless. The first acoustic signature weight refers to the weight coefficient assigned to bearing anomaly in the calculation of the equipment failure risk index. The first acoustic signature weight φ1 is set to 0.6. The reason for setting the first acoustic signature weight to 0.6 is that bearing anomaly is the dominant factor leading to boiler feedwater pump shutdown and water shortage accidents, and the mechanical failures caused by it account for the highest proportion of boiler safety accidents. According to industrial boiler failure statistics, boiler shutdown accidents caused by bearing failure are about 1.5 times that caused by combustion failure. Therefore, a higher weight is assigned to prioritize the control of mechanical failure risk. The second voiceprint weight refers to the weighting coefficient assigned to combustion instability in the calculation of the equipment failure risk index. The second voiceprint weight φ2 is set to 0.4. The reason for setting the second voiceprint weight to 0.4 is that combustion instability mainly leads to a decrease in combustion efficiency and excessive pollutant emissions. Although it may cause the risk of deflagration, its immediate threat to the overall safe operation of the boiler room is less than that of bearing failure. In normal operation, combustion fluctuations can usually be automatically adjusted by the control system, so a lower weight is assigned. The preset failure value refers to the critical threshold for determining whether the equipment failure risk exceeds the limit. The preset failure value Ds0 is set to 1.0. The reason for setting the preset failure value to 1.0 is: based on industrial... The boiler auxiliary equipment acoustic signature diagnostic database indicates that when the equipment failure risk index Ds = 1.0, the corresponding bearing or combustion system is at the critical point between a sub-healthy state and a faulty state. Exceeding this value indicates entering a risk zone requiring immediate maintenance, which may lead to unplanned boiler shutdowns. The attenuation coefficient refers to the adjustment parameter of the attenuation magnitude of the control path gain coefficient during the optimization process. The attenuation coefficient Rd is set to 0.3. The reason for setting the attenuation coefficient to 0.3 is as follows: According to the gain scheduling strategy in control theory, the denominator-increasing attenuation formula Rg' = Rg / (1 + Rd × (Ds - Ds0)) is adopted. When Ds reaches its maximum value Dsmax, i.e., 2.0, Rg' = Rg / 1.3 ≈ 0.77Rg ensures that the path gain coefficient is always positive and not lower than 77% of the original value. This effectively prevents abnormal equipment from falling out of the monitoring range due to excessive path deviation, while maintaining sufficient response sensitivity to avoid path stagnation. The redirection gain coefficient refers to the proportional conversion coefficient between the equipment failure risk index deviation and the angle correction amount when calculating the redirection angle correction. The redirection gain coefficient Ka is set to 15 degrees per unit Ds deviation. The reason for setting the redirection gain coefficient to 15 degrees is that, based on boiler room equipment layout experience, every 15-degree correction allows the robot dog to maintain a safe distance while aiming the monitoring angle at the location of the abnormal equipment. When Ds deviates by 0.3, the angle correction Δθ does not exceed 4.5 degrees. This avoids excessive turning that could lead to loss of control while ensuring that abnormal devices enter the effective monitoring field of the sensors. The azimuth angle of the abnormal device refers to the azimuth angle of the abnormal sound source relative to the robot dog's current position, determined by the acoustic signature orientation algorithm and calculated from the time delay estimation result of the microphone array beamforming. The current azimuth angle of the robot dog refers to the current heading angle measured by the robot dog's inertial measurement unit. The sign function is a mathematical function that returns the sign of the independent variable; it returns 1 when the independent variable is positive, -1 when it is negative, and 0 when it is zero, to ensure that the redirection direction always points to the abnormal device.

[0036] Specifically, step S6 involves acquiring segmented inspection data during the inspection process and generating a blockchain evidence storage scheme based on the segmented inspection data, specifically including: The cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr in the segmented inspection data are obtained. Based on the cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr, the basic evidence storage level Cl and the basic consensus node number Cv are calculated. Cl = round(Cbase + m1 × (W / Wbase) + m2 × (Rmax / Rbase)) and Cv = ceil(Cvbase × [1 + n1 × (tr / tbase)]), where Cbase is the baseline evidence storage level, Wbase is the baseline inspection trajectory, Rbase is the baseline risk level, Cvbase is the baseline consensus node number, tbase is the baseline response time, m1 is the correlation coefficient of the first blockchain, m2 is the correlation coefficient of the second blockchain, n1 is the correlation coefficient of the third blockchain, round is the rounding function, and ceil is the rounding function. This yields a blockchain evidence storage scheme that includes the basic evidence storage level Cl and the basic consensus node number Cv.

[0037] Specifically, the cumulative inspection trajectory value refers to the total length of the path actually traveled by the robot dog within a complete inspection cycle. In this embodiment, the cumulative inspection trajectory value is obtained by fusing and integrating the pulse count of the robot dog's wheel encoder and the dead reckoning data from the inertial measurement unit. Specifically, it is calculated using the formula W equals the encoder mileage plus the inertial compensation mileage and the time integral of the trailing edge, with the unit being meters. The highest risk level refers to the risk level classification corresponding to the maximum value of the inspection risk dynamic index Qr within this inspection cycle. In this embodiment, the highest risk level is obtained by recording the peak value of the inspection risk dynamic index Qr calculated in real time during the inspection process and mapping the Qr value range to three risk levels: low, medium, and high. The response delay... Time refers to the time interval from when the robot dog detects an abnormal signal to when the edge computing gateway generates an early warning command. In this embodiment, high-precision real-time clocks are deployed on the robot dog's main control chip and the edge gateway respectively to record the trigger time of the abnormal signal and the issuance time of the early warning command, and the time difference between the two is calculated to obtain the response delay time in milliseconds. The benchmark evidence level refers to the blockchain evidence level benchmark value required to obtain the credibility of basic data for a standard boiler room under the conditions of standardized inspection trajectory and risk level. The benchmark evidence level is set to 1. The reason for setting the benchmark evidence level to 1 is that it corresponds to the basic evidence strength under low-risk routine inspection scenarios. Using this as an integer level benchmark, the round function ensures that Cl is always constant. The integer value satisfies the discrete management requirements of blockchain evidence storage levels. The benchmark inspection trajectory refers to the nominal inspection trajectory length used for a specific boiler power in the standard boiler room inspection process evaluation, serving as a benchmark for comparing the deviation of the cumulative inspection trajectory value. The benchmark inspection trajectory value is 150. The reason for using a benchmark inspection trajectory value of 150 is that it corresponds to the standard inspection path length of a 4 MW boiler room, in meters. Using this as a normalization benchmark ensures that the W / Wbase term increases linearly as the path length increases. The benchmark risk level refers to the expected risk level benchmark value under normal operating conditions in the standard safety inspection process. The benchmark risk level value is 1. The reason for using a benchmark risk level value of 1 is that it corresponds to a low-risk condition. The risk level is encoded under a certain state, and this is used as a normalization benchmark to ensure that the Rmax / Rbase term increases linearly as the risk escalates. The benchmark consensus node number refers to the baseline number of consensus nodes required to ensure basic tamper-proof capabilities in a standard blockchain evidence storage scheme. It is usually calculated by rounding up half the number of participants in the consortium blockchain. The benchmark consensus node number is set to 4. The reason for setting the benchmark consensus node number to 4 is that, for a four-party consortium blockchain composed of boiler room operators, equipment manufacturers, maintenance units, and regulatory agencies, taking half the number of participants and rounding up satisfies the basic requirement of the PBFT algorithm for the number of consensus nodes N≥3f+1. This is used as the integer node benchmark, and the ceil function ensures that Cv is always a positive integer.The baseline response time refers to the target response time baseline value from anomaly detection to early warning generation. The baseline response time is set to 500. The reason for setting the baseline response time to 500 is that it corresponds to the standard latency of the robot dog communicating with the edge gateway via the 5G network, in milliseconds. Using this as a normalization baseline ensures that the tr / tbase term increases linearly as latency increases. The first blockchain correlation coefficient is a correction coefficient characterizing the degree of influence of the cumulative deviation of the inspection trajectory on the adjustment of the evidence storage level. The first blockchain correlation coefficient is set to 0.5. The reason for setting the first blockchain correlation coefficient to 0.5 is that, according to the regression analysis of blockchain evidence storage credibility and inspection coverage integrity, for every 10% increase in the inspection trajectory, due to the increase in monitoring data volume, the evidence storage level needs to be increased by approximately 0.5 levels to enhance hash verification density. The second blockchain correlation coefficient is a correction coefficient characterizing the deviation of the highest risk level. The correction coefficient for the impact of the deviation on the evidence storage level adjustment is as follows: The second blockchain correlation coefficient is set to 1.0. The reason for setting the second blockchain correlation coefficient to 1.0 is that when the highest risk level exceeds the standard value, it indicates that the current inspection involves a security anomaly, and the evidence storage level needs to be increased by 1 level on the basis of the baseline evidence storage level to ensure that the immutability of the abnormal data on the chain meets the security audit requirements. The third blockchain correlation coefficient is a correction coefficient that characterizes the impact of the response delay time deviation on the adjustment of the number of consensus nodes. The third blockchain correlation coefficient is set to 0.1. The reason for setting the third blockchain correlation coefficient to 0.1 is that when the response delay time is extended, it indicates that the edge computing gateway load is increased or the network is congested, and the number of consensus nodes needs to be increased accordingly to distribute the verification pressure. For every 10% extension of the response time, the number of consensus nodes needs to be increased by about 1%. n1 is set to 0.1 to achieve this proportional relationship.

[0038] Specifically, step S7 involves obtaining the device aging index and adjusting the blockchain evidence storage scheme based on the device aging index, specifically including: The aging index of the equipment is obtained by using the vibration acceleration sensor mounted on the robot dog, and the aging deviation value Ag is obtained. The aging deviation value Ag is compared with the preset aging value Ag0. Based on the comparison result, the aging status of the equipment is judged, and the blockchain evidence storage scheme is adjusted according to the judgment result, wherein: When Ag≤Ag0, the device aging status is determined to be normal, and no adjustment is made to the blockchain evidence storage scheme; When Ag > Ag0, the device is determined to be in a deteriorated state, and the blockchain evidence storage scheme is adjusted. The adjustment includes: calculating the consensus node increment ΔC = Jc × (Ag - Ag0) to obtain the adjusted number of consensus nodes C1 = Cv + ΔC, where Cv is the basic consensus node number, Jc is the node adjustment coefficient, and the basic consensus node number Cv in the blockchain evidence storage scheme is replaced with the adjusted number of consensus nodes C1.

[0039] Specifically, the aging deviation value refers to the relative deviation of the effective value of vibration acceleration of the boiler body and auxiliary equipment from the factory reference vibration value during the inspection process. Ag is set as (a_rms - a_base) / a_base, where a_rms is the current effective value of vibration acceleration, and a_base is the factory reference effective value of vibration acceleration, dimensionless, representing the degree of mechanical aging of the equipment. The preset aging value refers to the permissible deviation threshold of vibration intensity specified in the boiler safety technical regulations. The preset aging value is set to 0.2. The reason for setting the preset aging value to 0.2 is as follows: According to the vibration limits for medium-sized machines in GB / T6075.3-2011 "Mechanical Vibration - Measurement and Evaluation of Machine Vibration on Non-Rotating Components," combined with long-term operating experience of industrial boiler auxiliary equipment, when the effective value of vibration acceleration deviates from the reference value by more than 20%, the bearing clearance expansion and rotor imbalance have entered a detectable deterioration stage, requiring strengthened evidence to trace the aging trend. Simultaneously, the preset aging value also considers regional... The economic efficiency of blockchain consensus costs is demonstrated by the fact that when the aging deviation value is less than 20%, the standard number of consensus nodes is sufficient to meet the credibility of evidence storage without additional node overhead, thus avoiding excessive load on the blockchain network. The equipment aging status refers to the classification and judgment result of the mechanical health status of boiler equipment based on the comparison result of the aging deviation value and the preset aging value. The equipment aging status includes two types: normal and deteriorated. The node adjustment coefficient refers to the increase in consensus nodes corresponding to a unit aging deviation when calculating the increase in consensus nodes. The node adjustment coefficient is set to 2. The reason for setting the node adjustment coefficient to 2 is as follows: According to the security analysis of the PBFT consensus algorithm of the consortium blockchain, for every 2 additional consensus nodes, the Byzantine fault tolerance capability increases by 1 node, which can improve the ability of evidence storage data to resist collusive tampering by about 15%. For aging equipment, this value can provide sufficient consensus strength to ensure the credibility of data traceability, while avoiding consensus latency exceeding the business tolerance due to too many nodes. It has been verified by blockchain performance testing as the optimal compromise value.

[0040] Specifically, step S8 modifies the adjustment process of step S7 based on the equipment ownership coefficient in the equipment status parameters, specifically including: The device ownership coefficient Ow is calculated based on the device ownership chain length D and ownership change frequency f in the device status parameters, and is set as Ow=(D / D0)×(f / f0), where D0 is the baseline ownership chain length and f0 is the baseline change frequency. The device ownership coefficient Ow is compared with the preset ownership value Ow0, and the device ownership status is judged based on the comparison result. The adjustment process of the blockchain evidence storage scheme is then corrected based on the judgment result. When Ow≤Ow0, the ownership status of the device is determined to be clear, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc'=Jc×(1-Kf×(Ow0-Ow) / Ow0), where Kf is the ownership correction coefficient. When Ow > Ow0, the device ownership status is determined to be complex, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc' = Jc × (1 + Kf × (Ow - Ow0) / Ow0), where Kf is the ownership correction coefficient.

[0041] Specifically, the equipment ownership chain length refers to the number of ownership transfers recorded on the blockchain since the boiler equipment left the factory, determined by the number of NFT ownership token transactions. The ownership change frequency refers to the average number of ownership transfers of the boiler equipment per unit time, measured in times per year. The benchmark ownership chain length refers to the standard ownership transfer frequency value used as a comparison benchmark. The benchmark ownership chain length is set to 3. The reason for setting the benchmark ownership chain length to 3 is that this value corresponds to the typical number of ownership transfers of an industrial boiler from factory to its first major overhaul, including three transfers: from manufacturer to distributor, from distributor to user, and from user to maintenance unit. Using this as a normalization benchmark makes the equipment ownership coefficient Ow a dimensionless parameter, which is convenient for... Comparing the differences in ownership complexity among different boilers, the benchmark change frequency refers to the standard ownership change frequency value used as a comparison benchmark. The benchmark change frequency is set to 0.5. The reason for setting the benchmark change frequency to 0.5 is that it corresponds to the average ownership change frequency of an industrial boiler over a 10-year service life, measured in times per year. This frequency indicates that the equipment is in a stable operating phase, and the ownership relationship is clear and traceable. Using this as a normalization benchmark, the ownership coefficient Ow of equipment with high-frequency changes is significantly greater than 1, triggering the evidence enhancement mechanism. The preset ownership value refers to the critical threshold for determining the sensitivity of the equipment ownership status to evidence adjustment responses. The preset ownership value is set to 1.0. The reason for setting the preset ownership value to 1.0 is that when the equipment ownership chain length D is 3, and... When the frequency of ownership changes equals the benchmark value f0, the equipment ownership coefficient Ow is exactly 1.0. At this time, the boiler is in a standard ownership transfer state, possessing both a certain depth of historical traceability and maintaining a stable current ownership relationship. Using this value as the dividing line between clear ownership and complex ownership states ensures strengthened evidence preservation and adjustment for equipment with frequent ownership changes, while reducing evidence preservation costs for equipment with stable ownership. Equipment with frequent ownership changes refers to equipment with an ownership change frequency greater than once per year. The equipment ownership status refers to the classification and judgment result of the comprehensive degree of complexity of equipment ownership relationship and evidence preservation credibility requirements based on the comparison result of the equipment ownership coefficient Ow and the preset ownership value Ow0. The equipment ownership status includes two types: clear ownership and complex ownership. The ownership correction coefficient refers to the gain parameter that controls the rate of change of the adjustment magnitude with the complexity of device ownership when adjusting the node adjustment coefficient. The ownership correction coefficient is set to 0.5. The reason for setting the ownership correction coefficient to 0.5 is as follows: According to the marginal effect analysis of ownership complexity on the credibility of evidence storage, when the device ownership coefficient Ow increases from the critical value of 1.0 to 2.0, the node adjustment coefficient Jc needs to be increased by about 50% to achieve the same increase in credibility of evidence storage. Taking Kf as 0.5 can make Jc' = Jc × 1.5 when Ow = 2.0, which avoids data credibility disputes caused by insufficient evidence storage strength for complex devices, and also prevents excessive increase in consensus nodes from causing blockchain network congestion, thus achieving a balance between the credibility of evidence storage and network performance. At the same time, when Ow increases from 1.0 to 2.0, the ownership correction coefficient is set to 0.5.When the value is reduced from 0 to 0.5, Jc' = Jc × 0.75, reducing the node adjustment coefficient by 25%, decreasing the consensus overhead of devices with clear ownership, and achieving differentiated evidence storage cost control.

[0042] Specifically, step S9 involves storing the segmented inspection data on the blockchain according to the blockchain evidence storage scheme. This includes: packaging the segmented inspection data into data blocks; setting the consensus parameters of the consortium blockchain based on the basic evidence storage level and the number of basic consensus nodes in the blockchain evidence storage scheme; digitally signing the data blocks using the robot dog's device identity certificate as the transaction initiator's private key; broadcasting the signed data blocks to the verification nodes corresponding to the number of basic consensus nodes and executing the PBFT consensus protocol; and after consensus is reached, writing the data blocks into the blockchain distributed ledger and pushing the evidence storage hash value to the regulatory platform.

[0043] Please see Figure 3 As shown, this is a schematic diagram of the system structure of the multimodal data acquisition robot dog inspection method of this embodiment. The system includes: The multi-source acquisition module is used to collect equipment status parameters and inspection data from multiple sources in the target boiler room. The inspection plan generation module is used to generate a basic inspection plan based on equipment status parameters. The inspection plan generation module is connected to the multi-source acquisition module. The plan push module is used to push basic inspection plans, and the plan push module is connected to the inspection plan generation module; The inspection feedback module is used to calibrate the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data, optimize the calibration process of the basic inspection plan based on the equipment operation soundprint data in the multi-source inspection data, acquire segmented inspection data during the inspection process, and generate a blockchain evidence storage plan based on the segmented inspection data. The inspection feedback module is connected to the plan push module. The evidence correction module is used to obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index. It is also used to correct the adjustment process of the blockchain evidence storage scheme according to the equipment ownership coefficient in the equipment status parameters. The evidence correction module is connected to the inspection feedback module. The evidence storage scheme push module is used to push blockchain evidence storage schemes, and the evidence storage scheme push module is connected to the evidence storage correction module.

[0044] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for multimodal data acquisition robot dog inspection, characterized in that, include: Step S1: Collect equipment status parameters and multi-source inspection data for the target boiler room; Step S2: Generate a basic inspection plan based on equipment status parameters; Step S3: Conduct an inspection of the target boiler room according to the basic inspection plan, and execute steps S4 and S5 during the inspection process; Step S4: Verify the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data; Step S5: Optimize the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data. The optimization includes attenuating the path gain coefficient based on the equipment failure risk index and redirecting the calibrated inspection path based on the azimuth angle of abnormal equipment. Step S6: During the inspection process, segmented inspection data is acquired, and a blockchain evidence storage scheme is generated based on the segmented inspection data. Step S7: Obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index; Step S8: Correct the adjustment process of the blockchain evidence storage scheme according to the device ownership coefficient in the device status parameters. The correction includes reducing the node adjustment coefficient when the device ownership coefficient is less than or equal to the ownership benchmark value, and increasing the node adjustment coefficient when the device ownership coefficient is greater than the ownership benchmark value. Step S9: The segmented inspection data is stored on the blockchain according to the blockchain evidence storage scheme.

2. The multimodal data acquisition robot dog inspection method and system according to claim 1, characterized in that, Step S4 includes: Step S41: Perform time-series alignment and outlier removal on the environmental hazard factor data to obtain standardized environmental hazard factor data; Step S42: Perform multi-dimensional risk quantification on the standardized environmental hazard factor data to obtain an environmental risk vector; the multi-dimensional risk quantification includes extracting concentration features, temperature rise features and spatial features, the concentration features include the rate of exceedance of harmful gas concentrations, the temperature rise features include the surface temperature rise rate and hot spot gradient ratio, and the spatial features include the coverage rate of hazardous areas. Step S43: Calculate the dynamic index of inspection risk based on the environmental risk vector, and calibrate the basic inspection plan based on the dynamic index of inspection risk to obtain the calibrated inspection plan.

3. The multimodal data acquisition robot dog inspection method and system according to claim 2, characterized in that, In step S41, when performing time-series alignment and outlier removal on the environmental hazard factor data, a sliding window midpoint filtering algorithm is used to remove outliers from the environmental hazard factor data, resulting in outlier-removed environmental hazard factor data. The window width of the sliding window midpoint filtering algorithm is 5 sampling periods. A timestamp synchronization algorithm is then used to perform time-series alignment on the outlier-removed environmental hazard factor data, resulting in standardized environmental hazard factor data. The synchronization accuracy of the timestamp synchronization algorithm is 10 milliseconds.

4. The multimodal data acquisition robot dog inspection method and system according to claim 2, characterized in that, In step S42, when performing multi-dimensional risk quantification on the standardized environmental hazard factor data, the exceedance rate R of the hazardous gas concentration is calculated using the formula R=∑(Ci-C0i) / C0i / Nc, where Ci is the measured concentration of the hazardous gas at the i-th sampling point, C0i is the standard safe concentration threshold of the corresponding hazardous gas, and Nc is the total number of sampling points; the surface temperature rise rate Vr is calculated using the formula Vr=(Tt-Tt-Δt) / Δt, where Tt is the current surface temperature, Tt-Δt is the surface temperature before time Δt, and Δt is the sampling interval. The hotspot gradient ratio G is obtained by collecting the boiler surface temperature field using an infrared thermal imager array, extracting the highest temperature point Thot and the average temperature point Tavg, and then calculating it using the formula G=(Thot-Tavg) / Tavg; the hazardous area coverage A is obtained by performing three-dimensional reconstruction of the boiler room space using lidar point cloud data, identifying obstacle boundaries and hazardous restricted areas, and then calculating it using the formula A=Sdanger / Stotal, where Sdanger is the projected area of ​​the hazardous restricted area on the inspection path, and Stotal is the total coverage area of ​​the inspection path.

5. The multimodal data acquisition robot dog inspection method and system according to claim 2, characterized in that, Step S43 calculates the dynamic index of inspection risk based on the environmental risk vector, and verifies the basic inspection plan based on the dynamic index of inspection risk, specifically including: The inspection risk dynamic index Qr is calculated based on the harmful gas concentration exceedance rate R, surface temperature rise rate Vr, hot spot gradient ratio G, and hazardous area coverage A. Qr is set as Qr = w1 × (R / R0) + w2 × (Vr / V0) + w3 × (G / G0) + w4 × (A / A0), where R0 is the baseline exceedance rate, V0 is the baseline temperature rise rate, G0 is the baseline hot spot gradient, A0 is the baseline coverage, w1 is the first risk weight, w2 is the second risk weight, w3 is the third risk weight, and w4 is the fourth risk weight. The inspection risk dynamic index Qr is compared with the preset risk value Qr0. Based on the comparison result, the inspection risk status is judged, and the basic inspection plan is checked based on the judgment result. When Qr≤Qr0, the inspection risk status is determined to be controllable, and the basic inspection plan is not checked. When Qr > Qr0, the inspection risk status is determined to be out of limit, and the basic inspection plan is checked. The check includes: calculating the path offset ΔP = Rg × (Qr - Qr0) to obtain the checked inspection path L1 = L0 + ΔP, where L0 is the basic inspection path, Rg is the path gain coefficient, and the basic inspection path L0 in the basic inspection plan is replaced with the checked inspection path L1.

6. The multimodal data acquisition robot dog inspection method and system according to claim 5, characterized in that, Step S5 optimizes the calibration process of the basic inspection plan based on the equipment operation acoustic data in the multi-source inspection data, specifically including: Based on the equipment operation acoustic data, the bearing anomaly degree Eb and combustion instability Eu are extracted. Then, based on the bearing anomaly degree Eb, combustion instability Eu, first acoustic weight φ1, and second acoustic weight φ2, the equipment failure risk index Ds is calculated, set as Ds = φ1 × (Eb / Eb0) + φ2 × (Eu / Eu0), where Eb0 is the baseline bearing anomaly degree and Eu0 is the baseline combustion instability degree. The equipment failure risk index Ds is compared with the preset failure value Ds0. Based on the comparison result, the equipment failure risk is judged, and based on the judgment result, the calibration process of the basic inspection plan is optimized. When Ds≤Ds0, the risk of equipment failure is determined to be controllable, and the calibration process of the basic inspection plan is not optimized. When Ds > Ds0, the equipment failure risk is determined to be an over-limit risk. The calibration process of the basic inspection plan is optimized. The optimization includes: adjusting the path gain coefficient Rg to Rg', setting Rg' = Rg / (1 + Rd × (Ds - Ds0)), where Rd is the attenuation coefficient; calculating the redirection angle correction Δθ = Ka × (Ds - Ds0) × sign(θtarget - θcurrent), to obtain the redirected inspection path L2, where L2 keeps the radial distance of L1 unchanged, the azimuth angle is corrected to θcurrent + Δθ, θtarget is the azimuth angle of the abnormal equipment, θcurrent is the current azimuth angle of the robot dog, Ka is the redirection gain coefficient, and sign is the sign function.

7. The multimodal data acquisition robot dog inspection method and system according to claim 1, characterized in that, Step S6 involves acquiring segmented inspection data during the inspection process and generating a blockchain evidence storage scheme based on the segmented inspection data. Specifically, this includes: The cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr in the segmented inspection data are obtained. Based on the cumulative value W of the inspection trajectory, the highest risk level Rmax, and the response delay time tr, the basic evidence storage level Cl and the basic consensus node number Cv are calculated. Cl = round(Cbase + m1 × (W / Wbase) + m2 × (Rmax / Rbase)) and Cv = ceil(Cvbase × [1 + n1 × (tr / tbase)]), where Cbase is the baseline evidence storage level, Wbase is the baseline inspection trajectory, Rbase is the baseline risk level, Cvbase is the baseline consensus node number, tbase is the baseline response time, m1 is the correlation coefficient of the first blockchain, m2 is the correlation coefficient of the second blockchain, n1 is the correlation coefficient of the third blockchain, round is the rounding function, and ceil is the rounding function. This yields a blockchain evidence storage scheme that includes the basic evidence storage level Cl and the basic consensus node number Cv.

8. The multimodal data acquisition robot dog inspection method and system according to claim 7, characterized in that, Step S7 involves obtaining the device aging index and adjusting the blockchain evidence storage scheme based on the device aging index, specifically including: The aging index of the equipment is obtained by using the vibration acceleration sensor mounted on the robot dog, and the aging deviation value Ag is obtained. The aging deviation value Ag is compared with the preset aging value Ag0. Based on the comparison result, the aging status of the equipment is judged, and the blockchain evidence storage scheme is adjusted according to the judgment result, wherein: When Ag≤Ag0, the device aging status is determined to be normal, and no adjustment is made to the blockchain evidence storage scheme; When Ag > Ag0, the device is determined to be in a deteriorated state, and the blockchain evidence storage scheme is adjusted. The adjustment includes: calculating the consensus node increment ΔC = Jc × (Ag - Ag0) to obtain the adjusted number of consensus nodes C1 = Cv + ΔC, where Cv is the basic consensus node number, Jc is the node adjustment coefficient, and the basic consensus node number Cv in the blockchain evidence storage scheme is replaced with the adjusted number of consensus nodes C1.

9. The multimodal data acquisition robot dog inspection method and system according to claim 8, characterized in that, Step S8 modifies the adjustment process of step S7 based on the equipment ownership coefficient in the equipment status parameters, specifically including: The device ownership coefficient Ow is calculated based on the device ownership chain length D and ownership change frequency f in the device status parameters, and is set as Ow=(D / D0)×(f / f0), where D0 is the baseline ownership chain length and f0 is the baseline change frequency. The device ownership coefficient Ow is compared with the preset ownership value Ow0, and the device ownership status is judged based on the comparison result. The adjustment process of the blockchain evidence storage scheme is then corrected based on the judgment result. When Ow≤Ow0, the ownership status of the device is determined to be clear, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc'=Jc×(1-Kf×(Ow0-Ow) / Ow0), where Kf is the ownership correction coefficient. When Ow > Ow0, the device ownership status is determined to be complex, and the adjustment process of the blockchain evidence storage scheme is corrected. The correction includes: adjusting the node adjustment coefficient Jc to Jc', and setting Jc' = Jc × (1 + Kf × (Ow - Ow0) / Ow0), where Kf is the ownership correction coefficient.

10. A system applied to the multimodal data acquisition robot dog inspection method as described in any one of claims 1-9, characterized in that, include: The multi-source acquisition module is used to collect equipment status parameters and inspection data from multiple sources in the target boiler room. The inspection plan generation module is used to generate basic inspection plans based on equipment status parameters; The plan push module is used to push basic inspection plans; The inspection feedback module is used to calibrate the basic inspection plan based on the environmental hazard factor data in the multi-source inspection data, optimize the calibration process of the basic inspection plan based on the equipment operation soundprint data in the multi-source inspection data, acquire segmented inspection data during the inspection process, and generate a blockchain evidence storage plan based on the segmented inspection data. The evidence storage correction module is used to obtain the equipment aging index and adjust the blockchain evidence storage scheme according to the equipment aging index. It is also used to correct the adjustment process of the blockchain evidence storage scheme according to the equipment ownership coefficient in the equipment status parameters. The evidence storage solution push module is used to push blockchain evidence storage solutions.

Images