Nondestructive testing method for welding seam of steel structure

Abstract

The invention discloses a nondestructive testing method for a welding seam of a steel structure. The method comprises the steps that firstly, morphology data of the surface of a building steel structure are obtained through a high-precision laser scanner and an electromagnetic induction technology; coating thickness distribution and weld surface roughness are analyzed based on the morphology data, and a surface information matrix is constructed; then automatically adjusting ultrasonic wave beam parameters of the ultrasonic phased array by using the adaptive parameter regulation and control model and executing detection to obtain a preliminary ultrasonic echo signal; the time-frequency domain difference between the preliminary echo and the reference signal is calculated through an echo signal compensation model, and secondary compensation is completed to obtain an optimized ultrasonic echo signal; fusing the optimized echo signal, the morphology data and the surface information matrix to construct a detection data set, and adopting a pre-trained steel structure detection model to realize defect detection; and finally, evaluating the confidence of a detection result by adopting a three-dimensional confidence evaluation model, performing grading, and performing local re-detection and parameter adjustment on a low-confidence region to form closed-loop optimization. According to the method, the surface state of the steel structure can be adaptively matched, the accuracy of echo signals and the defect detection precision are improved, and the detection reliability is guaranteed through three-dimensional confidence evaluation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ultrasonic testing, and particularly relates to an ultrasonic nondestructive testing method for internal defects of a building steel structure weld, and particularly relates to a nondestructive testing method for optimizing detection result determination by using a three-dimensional confidence evaluation model. BACKGROUND

[0002] Steel structures are widely used in the fields of buildings, bridges, machinery and the like due to advantages such as high strength and convenient construction, and a weld, as a key part of the connection of the steel structure, directly affects the overall safety and service life of the steel structure due to internal defects (such as cracks, pores, slag inclusions and the like) of the weld. At present, the ultrasonic phased array technology has become one of the mainstream technologies for nondestructive testing of the weld of the steel structure due to a wide detection range and high sensitivity, and some patents also attempt to improve the detection accuracy and result determination, for example, the patent with the publication number CN120177618A proposes a two-dimensional confidence calculation scheme based on model probability values combined with signal-to-noise ratios, trying to improve the reliability of the detection result, but the existing technology still has many deficiencies, specifically as follows: Lack of adaptability of parameter adjustment: in the traditional detection and some improved schemes, the ultrasonic wave beam parameters (such as focal length, probe angle, frequency and the like) of the ultrasonic phased array are mostly fixedly set or only adjusted based on a single surface feature, without fully considering the differences in the coating thickness distribution and the weld surface roughness of different steel structure surfaces, and the over-thickness of the coating or the over-standard roughness will cause uneven ultrasonic wave energy attenuation and propagation path deviation, and then affect the authenticity of the preliminary ultrasonic echo signal, and hide error hazards for subsequent defect identification and confidence evaluation.

[0003] Insufficient echo signal compensation: the existing technology mostly relies on a single test block echo signal for simple amplitude compensation, without accurately correcting the time-frequency domain features (such as phase shift and time delay difference) of the preliminary echo and the reference signal, resulting in signal distortion problems of the compensated echo signal, especially the insufficient restoration degree of the reflection signal of a small defect such as a <0.5mm pore, which is easy to cause defect omission or misjudgment.

[0004] Low data fusion degree: most detection methods only perform defect detection based on limited data of the ultrasonic echo signal or the echo signal combined with the surface morphology, without deeply fusing the steel structure surface information matrix such as the coating thickness distribution, the roughness distribution and the detection process data, which is difficult to fully reflect the correlation between the internal state of the weld and external influencing factors, and limits the accuracy of defect positioning and size measurement.

[0005] The confidence evaluation mechanism has significant defects: the existing confidence evaluation scheme, such as the two-dimensional calculation of the model probability values combined with the signal-to-noise ratios in the patent CN120177618A, has the problems of single dimension and disconnection with the actual defect reliability: Ignoring the influence of defect geometric characteristics: this kind of scheme only takes signal-to-noise ratio as the core index on the signal level, without considering the key role of the geometric characteristics of the defect itself on the detection reliability. For example, small size defects (<0.5mm) are naturally caused by weak reflected energy, resulting in low signal-to-noise ratio, which is easily misjudged as a low confidence area according to the existing scheme. While large size defects (>5mm) have high actual reliability even if the signal-to-noise ratio is slightly low due to surface interference. The existing scheme cannot distinguish such differences, resulting in a confidence evaluation result that does not match the actual state of the defect.

[0006] Ignoring the influence of environmental interference: environmental factors during the detection process, such as changes in ultrasonic wave propagation speed caused by high temperature environment, unstable probe coupling caused by vibration, and signal attenuation caused by dust attachment, will directly affect the signal-to-noise ratio. However, the existing scheme does not include environmental disturbances such as temperature fluctuations ΔT>15℃ and vibration amplitude A≥0.5mm as independent dimensions in confidence evaluation, mistakenly attributing the decrease in signal-to-noise ratio caused by the environment to unreliable defect recognition and triggering unnecessary local re-inspection, significantly increasing detection cost and time consumption.

[0007] Non-targeted closed-loop optimization mechanism: even if the existing technology identifies a low confidence area, it mostly uses a single way of fixed parameter re-inspection, without adjusting parameters and controlling the environment based on the root cause of low confidence, such as excessive surface coating thickness and large environmental temperature fluctuations. This results in low re-inspection efficiency, limited confidence improvement, and the risk of detection blind spots.

[0008] Therefore, there is an urgent need for a steel structure weld non-destructive testing method that can achieve adaptive parameter adjustment, accurate signal compensation, multi-source data fusion, and has three-dimensional confidence evaluation and targeted closed-loop optimization capability, to solve the core problems of parameter rigidity, signal distortion, confidence evaluation disconnection, and high re-inspection cost in existing technology. SUMMARY

[0009] The purpose of the present application is to overcome the problems of fixed parameters, insufficient signal compensation, low data fusion, and single dimension of confidence evaluation in existing steel structure weld ultrasonic non-destructive testing methods, and to provide a non-destructive testing method that can adaptively adjust detection parameters, improve echo accuracy, and improve defect detection precision, and realize confidence evaluation and closed-loop optimization based on three-dimensional models of defect characteristics, signal quality, and environmental interference and the analytic hierarchy process.

[0010] To achieve the above purpose, the present application provides a steel structure weld non-destructive testing method, comprising the following steps: S1, using a high-precision laser scanner and electromagnetic induction technology, collecting surface topography data of the building steel structure weld and the surrounding area, the surface topography data including three-dimensional geometric profile data and coating thickness data; S2, preprocessing the surface topography data, analyzing to obtain coating thickness distribution and weld surface roughness, constructing a surface information matrix according to spatial coordinates; S3, inputting the surface information matrix into an adaptive parameter control model, automatically adjusting the ultrasonic wave beam parameters of the ultrasonic phased array, controlling the ultrasonic phased array to detect the weld to obtain preliminary ultrasonic echo signals, and collecting detection environment data, the environment data including temperature fluctuations, vibration amplitudes, and dust concentrations; S4, based on the reference signal of the standard test block, calculating the time-frequency domain difference between the preliminary ultrasonic echo signal and the reference signal through an echo signal compensation model, performing secondary compensation on the preliminary ultrasonic echo signal, and obtaining an optimized ultrasonic echo signal; S5, fusing the optimized ultrasonic echo signal, surface topography data, surface information matrix, and environment data to construct a detection data set, inputting a pre-trained steel structure detection model to realize defect detection, and using a three-dimensional confidence evaluation model to quantitatively evaluate the comprehensive confidence of the detection results; the three-dimensional confidence evaluation model includes three evaluation dimensions of defect characteristics, signal quality, and environmental interference, and the comprehensive confidence value is calculated from the dimension scores and their weights; S6, adjusting the ultrasonic phased array parameters and performing local re-inspection on the low confidence area, synchronously optimizing the environmental interference control, repeating steps S3-S5 until the confidence meets the standard, and realizing closed-loop optimization.

[0011] Further, the sampling interval of the high-precision laser scanner in step S1 is not greater than 0.1 mm, and the electromagnetic induction technology calculates the coating thickness by combining the electromagnetic properties of the coating and the material calibration coefficient.

[0012] Further, in step S2, the dimension of the surface information matrix is consistent with the number of sampling points in the scanning area, and the core dimensions include coordinates, coating thickness, and roughness.

[0013] Further, in step S3, the adaptive parameter control model is constructed based on a BP neural network, the input is the surface information matrix feature, and the output is the focal length, probe angle, ultrasonic wave frequency, and pulse width of the ultrasonic phased array.

[0014] Further, in step S4, the echo signal compensation model uses a wavelet transform algorithm to calculate the time-frequency domain difference, and the secondary compensation includes amplitude gain compensation, phase calibration, and time alignment.

[0015] Further, in the three-dimensional confidence evaluation model in step S5, the weight of the defect feature dimension is 0.4, the evaluation indexes include defect size, defect length-diameter ratio and defect edge definition; the weight of the signal quality dimension is 0.3, the evaluation indexes include signal-to-noise ratio, signal amplitude stability and signal integrity; the weight of the environmental interference dimension is 0.3, the evaluation indexes include temperature fluctuation, vibration amplitude and dust concentration. The fundamental purpose of non-destructive testing is to identify the true state and safety risk of defects in the weld. The geometric and morphological characteristics of the defects themselves are the core factors that determine whether the defects affect the load safety of the steel structure. Therefore, the weight of this dimension is set to the highest (0.4), which can ensure that the confidence evaluation is focused on the risk properties of the defects themselves, avoiding the masking or misjudgment of the true risk level of the defects due to non-core factors such as signal fluctuations and environmental interference, and ensuring that the evaluation results are directly linked to the safety requirements of the structure. Ultrasonic non-destructive testing relies on ultrasonic echo signals to extract defect information, and signal quality is the prerequisite for accurately identifying defect characteristics. If the signal is distorted, unstable or missing, even if the defect itself is high-risk, it cannot be accurately evaluated. Therefore, the weight of this dimension is set to 0.3, which is lower than the defect feature and higher than the environmental interference, which can ensure that the confidence evaluation is neither based on data nor reversed, and achieves a balance between data reliability and defect risk. Environmental factors (temperature, vibration, dust) of ultrasonic testing will directly affect the signal quality, but they are not the properties of the defects themselves. For example, the decrease of signal-to-noise ratio caused by high temperature and the instability of amplitude caused by vibration are external interference, not unreliable defect identification. Therefore, the weight of this dimension is set to 0.3 (the same as signal quality), which can quantify the influence of external interference on detection, avoid misjudging the signal deviation caused by the environment as low reliability of defects, and thus reduce unnecessary retesting and reduce detection cost.

[0016] Further, the defect feature (Fdefect) score in step S5 is calculated by weighting the defect size, aspect ratio, and edge definition, and the specific scoring standards are as follows: 1 point for a defect size > 2 mm, 0.7 points for 1 mm≤ size≤ 2 mm, 0.5 points for 0.5 mm≤ size< 1 mm, and 0.4 points for size< 0.5 mm; 1 point for a crack aspect ratio > 8, 0.8 points for 5 < aspect ratio≤ 8, 1 point for a pore aspect ratio of 0.8-1.2, and 0.7 points for 1.2 < aspect ratio≤ 1.5 or 0.5≤ aspect ratio< 0.8; 1 point for a defect edge definition without blur, 0.6 points for slight blur, and 0.3 points for severe blur. Through the standard of the larger the size, the higher the score, the degree of weakening of the defect on the bearing capacity of the weld is directly related, the difference in scores between large-size defects (high risk) and small defects (low risk) is clear, and the risk assessment deviation caused by size misjudgment is avoided. For cracks, the aspect ratio is large, easy to expand, and the pore aspect ratio is close to 1, and the risk is controllable, the scoring standards are differentiated, the safety risks of different types of defects are accurately distinguished, and the evaluation error caused by confusion of defect attributes is avoided. Through the standard of 1 point for clear and 0.3 points for severe blur, the reliability of defect boundary recognition is quantified, and a clear edge means a small error in defect type / size determination, and a high score; blur indicates that the determination may have deviation, and a low score provides a basis for the recognition accuracy level for subsequent re-inspection. Combined with the weights of the three, the single index is avoided, the Fdefect score is fully matched with the real risk of the defect and the recognition reliability, accurate defect dimension support is provided for three-dimensional confidence calculation, and the three-dimensional confidence calculation is provided.

[0017] Further, the signal quality (Fsignal) score in step S5 is calculated by weighting the signal-to-noise ratio, signal amplitude stability, and signal integrity, and the specific score standards are as follows: 1 point for SNR>60dB, 0.8 points for 55dB<SNR≤60dB, 0.5 points for 50dB<SNR≤55dB, 0.3 points for SNR<50dB; 1 point for σ<5%, 0.7 points for 5%≤σ<10%, 0.4 points for 10%≤σ<15%, and 0.2 points for σ≥15%; 1 point for signal integrity C=1, 0.8 points for 0.8≤C<1, 0.5 points for 0.5≤C<0.8, and 0.3 points for C<0.5. The proportion of effective defect echoes and noise in the signal is accurately quantified according to the SNR value interval, the effective signal strength is determined, the score is high for high SNR, and the accuracy of defect feature extraction is ensured; the score is low for low SNR, and it is prompted that noise may mask defect information, thereby providing a basis for whether the signal needs to be optimized subsequently. The signal amplitude fluctuation degree is reflected by the σ value interval, the smaller the fluctuation, the higher the score, which indicates that the detection data has good repeatability and is less affected by random interference, and defect size / type misjudgment caused by unstable signal is avoided; the larger the fluctuation, the lower the score, which prompts that the data reliability is insufficient. Whether the signal is complete is determined according to the C value interval, C=1 represents that the signal has no missing and can fully capture the whole feature of the defect; C<0.5 represents that the signal is broken / missing and may miss the key information of the defect, the data integrity risk is prompted by the score difference, and false conclusions based on incomplete signals are prevented. Through the weight fusion of the three, the one-sidedness of a single index is avoided, the support ability of the signal for accurate detection of defects is comprehensively reflected, the data reliability is ensured, the precise signal dimension support is provided for three-dimensional confidence evaluation, and misjudgment / omission caused by signal problems is reduced.

[0018] Further, the environmental interference score (Fenv) in step S5 is calculated by weighting the temperature fluctuation, vibration amplitude, and dust concentration, and the specific score standards are as follows: 1 point for ΔT<5℃, 0.8 points for 5℃≤ΔT<10℃, 0.4 points for 10℃≤ΔT<15℃, and 0.2 points for ΔT>15℃; 1 point for A<0.1mm, 0.7 points for 0.1mm≤A<0.3mm, 0.4 points for 0.3mm≤A<0.5mm, and 0.2 points for A≥0.5mm; 1 point for D<5mg / m 3 , 0.8 points for 5mg / m 3 ≤D<10mg / m 3 , 0.5 points for 10mg / m 3 ≤D<20mg / m 3 , and 0.2 points for D≥20mg / m 30.3 points. Temperature fluctuation is graded to quantify the impact of temperature on acoustic characteristics. The ΔT range accurately reflects the interference intensity of temperature on ultrasonic wave propagation. Low-temperature fluctuations result in stable acoustic characteristics of steel, with minimal signal impact, leading to a high score. High-temperature fluctuations easily cause abnormal ultrasonic wave propagation speed / attenuation, resulting in a low score. This helps distinguish between signal deviations caused by temperature interference and defects themselves. Range A quantifies the interference of vibration on probe coupling. Low vibrations result in stable probe contact and no additional signal fluctuations, leading to a high score. High vibrations easily cause the probe to detach from the coupling surface, causing sudden changes in signal amplitude, resulting in a low score. This helps avoid misjudging vibration-induced signal instability as a defect feature. Range D reflects the impact of dust on coupling. Low dust levels result in normal couplant performance and unobstructed ultrasonic wave propagation, leading to a high score. High dust levels easily form an obstruction layer, causing signal attenuation, resulting in a low score. This helps correct the misleading assessment of defects caused by weak signals due to dust. By integrating the weights of these three factors, the one-sidedness of assessing a single environmental factor is avoided, and the overall impact of the testing environment on signal quality is comprehensively reflected. This provides a complete basis for confidence correction, thereby reducing unnecessary retesting caused by misjudgment due to environmental interference and lowering testing costs.

[0019] Furthermore, the confidence level mentioned in step S5 is calculated using the formula Conf=0.4×Fdefect + 0.3×Fsignal +0.3×Fenv, and the criterion for determining the low confidence level region is a confidence level < 0.5; the parameter adjustment mentioned in step S6 is based on the surface information matrix of the low confidence level region and the environmental data for fine-tuning, and the environmental interference control includes adding dust covers, shock-absorbing pads or temperature compensation modules.

[0020] Beneficial effects of this invention: Parameter adaptation enhances detection adaptability: By driving the adaptive parameter control model through the surface information matrix, the ultrasonic phased array parameters are accurately matched with the surface condition of the steel structure (coating thickness, roughness), avoiding signal distortion caused by fixed parameters; Precise signal compensation reduces errors: The secondary compensation mechanism based on time-frequency domain differences can specifically correct the amplitude, phase, and time delay errors of the initial echo. The optimized echo signal is closer to the real defect reflection characteristics, reducing the omission and misjudgment of minor defects. Multi-source data fusion improves detection accuracy: By integrating multiple sources of information such as ultrasonic echo, surface morphology, surface information and environmental data, the internal and external influencing factors of the weld are comprehensively reflected. Compared with single echo data detection, the defect positioning accuracy is improved and the defect size measurement error is controlled within a small range. Three-dimensional confidence assessment improves the reliability of results: The weights are determined by the analytic hierarchy process, and the confidence is comprehensively assessed from three dimensions: defect characteristics, signal quality, and environmental interference. This avoids the one-sidedness of single-dimensional assessment, improves the accuracy of confidence assessment, and can more accurately identify low-reliability areas. Closed-loop optimization mechanism to eliminate detection blind spots: Based on the three-dimensional confidence assessment results, parameters are fine-tuned and the environment is controlled for the surface condition and environmental interference in low confidence areas. The confidence level is achieved through re-inspection, so that the overall confidence level of the detection results is ≥95%. High degree of automation and improved efficiency: From data acquisition, parameter adjustment, defect detection to 3D confidence assessment and closed-loop optimization, all processes are automated. Compared with detection methods that rely on manual parameter adjustment and single-dimensional assessment, detection efficiency is greatly improved and human error is reduced. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the specific steps in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] This embodiment specifically discloses a non-destructive testing method for steel structure welds, including the following steps: Step 1: Obtain surface morphology data of the steel structure A high-precision laser scanner combined with electromagnetic induction technology is used to perform a comprehensive scan of the weld area and surrounding area of ​​the target building's steel structure. High-precision laser scanners are used to collect three-dimensional topographic data of steel structure surfaces and obtain the geometric contours of welds, such as weld height, width, and surface undulation information. Electromagnetic induction technology is used to detect the electromagnetic properties of coatings on the surface of steel structures, and in combination with preset coating material parameters, coating thickness data is obtained indirectly. During the scanning process, the sampling interval is controlled to be no greater than 0.1 mm to ensure that the resolution of the morphology data meets the requirements of subsequent analysis.

[0023] Step 2: Construct the surface information matrix The morphology data obtained in step 1 is preprocessed, including noise removal and data stitching, and then a surface information matrix is ​​constructed through the following analysis: Coating thickness distribution analysis: Based on the electromagnetic property data of the coating obtained by electromagnetic induction, the coating thickness at each sampling point is calculated using a calibration formula, such as coating thickness = electromagnetic signal attenuation × material calibration coefficient, and a coating thickness distribution map is generated. Weld surface roughness analysis: Based on the three-dimensional topographic data of laser scanning, roughness parameters such as the arithmetic mean deviation of the weld area (Ra) and the maximum height of the profile (Rz) are calculated, and a roughness distribution map is generated. Surface information matrix construction: Align the coating thickness distribution map and the roughness distribution map according to spatial coordinates, and construct a two-dimensional surface information matrix with coordinates, coating thickness and roughness as the core dimensions. The matrix dimensions are consistent with the number of sampling points in the scanning area.

[0024] Step 3: Acquire preliminary ultrasound echo signals Parameter adaptive adjustment: The surface information matrix constructed in step 2 is input into the pre-trained adaptive parameter control model. This model is based on a BP neural network. It takes surface information matrix features, such as average coating thickness and average roughness, as input and ultrasonic phased array beam parameters, such as focal length, probe angle, ultrasonic frequency, and pulse width, as output to automatically generate detection parameters that match the current surface condition of the steel structure. Ultrasonic phased array testing: Based on the beam parameters output by the adaptive parameter adjustment model, the ultrasonic phased array probe is controlled to scan and inspect the weld seams of the steel structure. After the ultrasonic waves penetrate the weld seams, they are reflected to form preliminary ultrasonic echo signals. The time-domain waveform and amplitude information of the echo signals are recorded by a data acquisition card. At the same time, environmental data, including temperature, vibration amplitude, and dust concentration, are collected synchronously.

[0025] Step 4: Obtain optimized ultrasound echo signals Reference signal acquisition: A standard test block with the same material and welding process as the steel structure to be tested is used. The test block contains defects of known size, such as φ2mm pores and 5mm long cracks. The ultrasonic echo signal of the standard test block is acquired under the same testing environment and used as a reference signal. Time-frequency domain difference calculation: The preliminary ultrasonic echo signal and the reference signal obtained in step 3 are input into the echo signal compensation model. The model uses the wavelet transform algorithm to decompose the two signals in the time-frequency domain, calculates the amplitude difference, phase difference and time delay difference between the two in different frequency bands, and generates a time-frequency domain difference matrix. Secondary compensation: Based on the time-frequency domain difference matrix, the initial ultrasonic echo signal is corrected by frequency band division, the amplitude attenuation region is compensated for gain, the phase shift region is calibrated, and the time delay deviation region is time aligned, finally obtaining the optimized ultrasonic echo signal.

[0026] Step 5: Defect Detection and 3D Confidence Assessment Detection dataset construction: The optimized ultrasonic echo signal obtained in step 4, the surface morphology data in step 1, the surface information matrix in step 2, and the environmental data collected in step 3 are correlated according to spatial coordinates to construct a multi-source fusion detection dataset; Defect Detection: Input the detection dataset into a pre-trained steel structure detection model, which is constructed based on a Convolutional Neural Network (CNN). Through the training process, the model learns the mapping relationships of multi-source data features, weld defect types (cracks, pores, slag inclusions), and defect sizes, and outputs the location, type, size (length / depth), aspect ratio, and edge clarity information of the defects; Three-dimensional Confidence Evaluation Model Construction and Weight Determination: Use the Analytic Hierarchy Process (AHP) to construct a three-dimensional confidence evaluation model for defect features - signal quality - environmental interference, and determine the weights of each dimension: Defect Feature Dimension: The weight is 0.4, and the included indicators are defect size (length / depth), defect aspect ratio (e.g., crack aspect ratio > 5, pore aspect ratio ≈ 1), and defect edge clarity; Signal Quality Dimension: The weight is 0.3, and the included indicators are signal-to-noise ratio (SNR), signal amplitude stability (σ), and signal integrity (C); Environmental Interference Dimension: The weight is 0.3, and the included indicators are temperature fluctuation (ΔT), vibration amplitude (A), and dust concentration (D); Calculation of Scores for Each Dimension: Defect Feature Score (Fdefect): According to the defect size, if the size > 2mm, get 1 point; if 1mm ≤ size ≤ 2mm, get 0.7 points; if 0.5mm ≤ size < 1mm, get 0.5 points; if size < 0.5mm, get 0.4 points, aspect ratio, e.g., if crack aspect ratio > 8, get 1 point; if 5 < aspect ratio ≤ 8, get 0.8 points; if pore aspect ratio is 0.8 - 1.2, get 1 point; if 1.2 < aspect ratio ≤ 1.5 or 0.5 ≤ aspect ratio < 0.8, get 0.7 points, edge clarity, e.g., if the edge is clear without blur, get 1 point; if slightly blurred, get 0.6 points; if severely blurred, get 0.3 points. Use equal weights, that is, the weight of each indicator is 1 / 3, and calculate Fdefect by weighted calculation; Signal Quality Score (Fsignal): According to the signal-to-noise ratio, if SNR > 60dB, get 1 point; if 55dB < SNR ≤ 60dB, get 0.8 points; if 50dB < SNR ≤ 55dB, get 0.5 points; if SNR < 50dB, get 0.3 points, signal amplitude stability, if σ < 5%, get 1 point; if 5% ≤ σ < 10%, get 0.7 points; if 10% ≤ σ < 15%, get 0.4 points; if σ ≥ 15%, get 0.2 points, signal integrity, if C = 1 and there is no signal loss, get 1 point; if 0.8 ≤ C < 1, get 0.8 points; if 0.5 ≤ C < 0.8, get 0.5 points; if C < 0.5, get 0.3 points. Use equal weights, that is, the weight of each indicator is 1 / 3, and calculate Fsignal by weighted calculation; Environmental disturbance score (Fenv): Based on temperature fluctuation, ΔT < 5℃ earns 1 point, 5℃ ≤ ΔT < 10℃ earns 0.8 points, 10℃ ≤ ΔT < 15℃ earns 0.4 points, and ΔT > 15℃ earns 0.2 points; vibration amplitude A < 0.1mm earns 1 point, 0.1mm ≤ A < 0.3mm earns 0.7 points, 0.3mm ≤ A < 0.5mm earns 0.4 points, and A ≥ 0.5mm earns 0.2 points; dust concentration D < 5mg / m³ earns 0.2 points. 3 1 point, 5mg / m 3 ≤D<10mg / m 3 0.8 points, 10mg / m 3 ≤D<20mg / m 3 0.5 points, D≥20mg / m 3 A score of 0.3 is obtained by using equal weighting, i.e., each indicator has a weight of 1 / 3, and the weighted average is used to calculate Fenv. Confidence level calculation and rating: Confidence score calculation: The confidence score of each defect detection result is calculated according to the formula Conf=0.4×Fdefect + 0.3×Fsignal + 0.3×Fenv; Confidence level classification: Preset confidence thresholds: high confidence: confidence ≥ 0.8, medium confidence: 0.5 ≤ confidence < 0.8, low confidence: confidence < 0.5. The detection area is classified into the corresponding level according to the confidence value.

[0027] Step 6: Closed-loop optimization Local re-detection trigger: For the low-confidence regions identified in step 5, their spatial coordinates are automatically marked, and a re-detection instruction is generated; Secondary parameter adjustment: Based on the surface information matrix of low confidence areas, such as areas with thicker coatings, higher roughness, and environmental interference data, such as areas with ΔT>10℃ and vibration amplitude A>0.3mm, the output parameters of the adaptive parameter control model are fine-tuned, such as increasing the ultrasonic frequency to improve resolution, adjusting the probe angle to reduce surface reflection interference, and adding a temperature compensation module to offset the effects of temperature fluctuations. Closed-loop verification: Perform local re-inspection of low-confidence areas according to the adjusted parameters, and simultaneously optimize environmental interference control, such as adding dust covers and shock-absorbing pads. Repeat steps 3-5 until the detection result confidence of the area is ≥0.5, medium confidence or above, and complete the closed-loop optimization. If the confidence still cannot meet the standard after adjusting the parameters multiple times, a manual review prompt will be output.

[0028] Finally, the method described in this application is merely a preferred embodiment and is not intended to limit the scope of protection of this invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A non-destructive testing method for welds in steel structures, characterized in that, Includes the following steps: S1. A high-precision laser scanner and electromagnetic induction technology are used to collect surface morphology data of the weld seams and surrounding areas of the building steel structure. The surface morphology data includes three-dimensional geometric contour data and coating thickness data. S2, preprocess the surface morphology data, analyze the coating thickness distribution and weld surface roughness, and construct a surface information matrix by aligning spatial coordinates; S3, input the surface information matrix into the adaptive parameter control model, automatically adjust the ultrasonic beam parameters of the ultrasonic phased array, control the ultrasonic phased array to detect the weld to obtain the preliminary ultrasonic echo signal, and at the same time collect the detection environment data, including temperature fluctuation, vibration amplitude, and dust concentration. S4. Based on the reference signal of the standard test block, the time-frequency domain difference between the preliminary ultrasonic echo signal and the reference signal is calculated through the echo signal compensation model. The preliminary ultrasonic echo signal is then compensated for in the second time to obtain the optimized ultrasonic echo signal. S5, the optimized ultrasonic echo signal, surface morphology data, surface information matrix and environmental data are integrated to construct a detection dataset, which is then input into a pre-trained steel structure detection model to achieve defect detection. A three-dimensional confidence evaluation model is used to perform a comprehensive confidence quantitative evaluation of the detection results. The three-dimensional confidence evaluation model includes three evaluation dimensions: defect characteristics, signal quality and environmental interference. Its comprehensive confidence value is calculated from the scores of each dimension and their weights. S6. Adjust the ultrasonic phased array parameters for the low-confidence region and perform local re-examination, while simultaneously optimizing environmental interference control. Repeat steps S3-S5 until the confidence level is met to achieve closed-loop optimization.

2. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, The sampling interval of the high-precision laser scanner mentioned in step S1 is no greater than 0.1 mm, and the electromagnetic induction technology calculates the coating thickness by combining the electromagnetic properties of the coating with the material calibration coefficient.

3. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, The dimensions of the surface information matrix mentioned in step S2 are consistent with the number of sampling points in the scanning area, and the core dimensions include coordinates, coating thickness and roughness.

4. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, The adaptive parameter control model described in step S3 is constructed based on a BP neural network. The input is the surface information matrix features, and the output is the focal length of the ultrasonic phased array, the probe angle, the ultrasonic frequency, and the pulse width.

5. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, The echo signal compensation model in step S4 uses wavelet transform algorithm to calculate time-frequency domain differences. The secondary compensation includes amplitude gain compensation, phase calibration and time alignment.

6. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, In the three-dimensional confidence assessment model described in step S5, the weight of the defect feature dimension is 0.4, and the assessment indicators include defect size, defect aspect ratio, and defect edge sharpness; the weight of the signal quality dimension is 0.3, and the assessment indicators include signal-to-noise ratio, signal amplitude stability, and signal integrity; the weight of the environmental interference dimension is 0.3, and the assessment indicators include temperature fluctuation, vibration amplitude, and dust concentration.

7. The non-destructive testing method for steel structure welds according to claim 6, characterized in that, The score of the defect feature (Fdefect) described in step S5 is obtained by weighted calculation of the weights of defect size, aspect ratio, and edge sharpness. The specific scoring criteria are as follows: if the defect size > 2mm, it gets 1 point; if 1mm ≤ size ≤ 2mm, it gets 0.7 points; if 0.5mm ≤ size < 1mm, it gets 0.5 points; if size < 0.5mm, it gets 0.4 points; if the crack aspect ratio > 8, it gets 1 point; if 5 < aspect ratio ≤ 8, it gets 0.8 points; if the aspect ratio of the air hole is 0.8 - 1.2, it gets 1 point; if 1.2 < aspect ratio ≤ 1.5 or 0.5 ≤ aspect ratio < 0.8, it gets 0.7 points; if the defect edge is clear without blur, it gets 1 point; if it is slightly blurred, it gets 0.6 points; if it is severely blurred, it gets 0.3 points.

8. The non-destructive testing method for steel structure welds according to claim 6, characterized in that, The score of the signal quality (Fsignal) described in step S5 is obtained by weighted calculation of the weights of signal-to-noise ratio, signal amplitude stability, and signal integrity. The specific scoring criteria are as follows: if SNR > 60dB, it gets 1 point; if 55dB < SNR ≤ 60dB, it gets 0.8 points; if 50dB < SNR ≤ 55dB, it gets 0.5 points; if SNR < 50dB, it gets 0.3 points; if σ < 5%, it gets 1 point; if 5% ≤ σ < 10%, it gets 0.7 points; if 10% ≤ σ < 15%, it gets 0.4 points; if σ ≥ 15%, it gets 0.2 points; if the signal integrity C = 1, it gets 1 point; if 0.8 ≤ C < 1, it gets 0.8 points; if 0.5 ≤ C < 0.8, it gets 0.5 points; if C < 0.5, it gets 0.3 points.

9. The non-destructive testing method for steel structure welds according to claim 6, characterized in that, The environmental disturbance score (Fenv) mentioned in step S5 is calculated by weighting temperature fluctuation, vibration amplitude, and dust concentration. The specific scoring criteria are as follows: ΔT < 5℃ = 1 point, 5℃ ≤ ΔT < 10℃ = 0.8 points, 10℃ ≤ ΔT < 15℃ = 0.4 points, ΔT > 15℃ = 0.2 points; A < 0.1mm = 1 point, 0.1mm ≤ A < 0.3mm = 0.7 points, 0.3mm ≤ A < 0.5mm = 0.4 points, A ≥ 0.5mm = 0.2 points; D < 5mg / m³ 3 1 point, 5mg / m 3 ≤D<10mg / m 3 0.8 points, 10mg / m 3 ≤D<20mg / m 3 0.5 points, D≥20mg / m 3 0.3 points.

10. The non-destructive testing method for steel structure welds according to claim 1, characterized in that, The confidence level described in step S5 is calculated according to the formula Conf = 0.4×Fdefect + 0.3×Fsignal + 0.3×Fenv. The determination criterion for the low confidence level region is that the confidence level < 0.5; the parameter adjustment described in step S6 is fine-tuned based on the surface information matrix and environmental data of the low confidence level region. The environmental interference control includes adding a dust-proof cover, a shock-absorbing pad, or a temperature compensation module.