Low-dose X-ray nondestructive testing method and system for elevator steel wire rope

Abstract

The invention discloses a nondestructive testing method and system for a low-dose X-ray elevator steel wire rope. Comprising the following steps: step 1, carrying out X-ray transmission scanning on an elevator steel wire rope which continuously moves to generate an initial digital image set; step 2, acquiring displacement information of the steel wire rope and associating the displacement information with each frame of image in the initial digital image set; step 3, performing preprocessing and image enhancement on the associated initial digital image set to obtain enhanced image data; 4, analyzing, identifying and positioning the image data based on a deep learning algorithm, and segmenting internal and external defects of the steel wire rope; and step 5, according to the defect identification result, quantitatively calculating the metal cross sectional area loss rate and / or the safety coefficient of the steel wire rope, and generating a safety evaluation report. According to the invention, the detection range of the elevator steel wire rope extends from the surface to the internal structure, visual identification and evaluation of wire breakage, abrasion, corrosion and other defects are realized, and the method has the advantage of high accuracy.

Description

Technical Field

[0001] This invention relates to the field of special equipment safety inspection technology, and in particular to a method and system for non-destructive testing of elevator wire ropes using low-dose X-rays. Background Technology

[0002] As a critical load-bearing component of elevator systems, the health of elevator wire ropes directly affects the operational safety of elevators. Due to long-term exposure to alternating stress, bending, and environmental corrosion, wire ropes can develop defects such as broken wires, wear, corrosion, and deformation. Internal defects, in particular, are difficult to detect and pose significant safety hazards. Currently, conventional inspection methods for elevator wire ropes mainly rely on manual visual inspection and electromagnetic testing. Manual visual inspection is highly subjective, unable to detect internal damage, and inefficient. Electromagnetic testing methods (such as magnetic flux leakage testing) are sensitive to surface and near-surface defects, but have limited effectiveness in detecting deep internal broken wires, corrosion of non-ferromagnetic materials, and the condition of the rope core. Furthermore, they are susceptible to the effects of wire rope vibration and lifting. Therefore, there is an urgent need for a non-destructive testing technology that can intuitively and accurately detect internal and external defects in wire ropes and achieve automated, quantitative assessment. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for non-destructive testing of elevator wire ropes using low-dose X-rays. This invention extends the testing scope of elevator wire ropes from the surface to the internal structure, achieving accurate visual identification and precise quantitative assessment of defects such as broken wires, wear, and corrosion, with the advantage of high accuracy.

[0004] The technical solution provided by this invention is as follows: A non-destructive testing method for elevator wire ropes using low-dose X-rays, comprising the following steps: Step 1: Use a low-dose X-ray source to perform a transmission scan on a continuously moving elevator steel cable, and use a digital detector array to receive the transmitted X-ray signals and generate an initial digital image set. Step 2: Synchronously acquire the displacement information of the wire rope through the motion control module, and associate the displacement information with each frame of the initial digital image set; Step 3: Preprocess and enhance the associated initial digital image set to obtain enhanced image data; Step 4: Analyze the enhanced image data based on deep learning algorithms to identify, locate, and segment the internal and external defects of the wire rope; Step 5: Based on the defect identification results, quantitatively calculate the metal cross-sectional area loss rate and / or safety factor of the wire rope, and generate a safety assessment report.

[0005] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes includes the following pretreatment: Multiple frames of images were acquired by a digital detector array under X-ray-free conditions, and then averaged to obtain a dark-field image. ; A flat-field image is obtained by averaging multiple frames of images acquired by a digital detector array in a uniform X-ray field without elevator cables. ; The original grayscale images in the initial digital image set are calibrated using the following formula. Perform standardization correction to obtain the corrected image. : ; In the formula: Represents the pixel coordinates in the image. It is the average value of the difference between the flat field image and the dark field image across the entire image.

[0006] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes includes a three-dimensional reconstruction sub-step after preprocessing and before image enhancement: Based on the displacement information associated with each frame of the image, assign precise spatial coordinates to each frame of the image: ; In the formula, It is the initial position of each frame in the initial digital image set. It is the image frame number. It is the displacement per frame; The corrected images are stacked in spatial order to reconstruct three-dimensional volume data representing the spatial structure of the wire rope. : .

[0007] The images in the three-dimensional volume data are used for subsequent image enhancement processing.

[0008] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes, wherein the image enhancement specifically refers to multi-scale enhancement based on wavelet transform, including: Discrete wavelet decomposition is performed on the image in the three-dimensional volume data to obtain low-frequency overview components and high-frequency components containing details in the horizontal, vertical, and diagonal directions. The high-frequency components are subjected to a nonlinear enhancement transformation, and the transformation formula is as follows: ; In the formula: Represents any detail coefficient. To enhance the intensity control factor, To adaptively adjust parameters The enhanced high-frequency components and low-frequency profile components are subjected to inverse wavelet transform to reconstruct the enhanced image.

[0009] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes employs a U-Net++ network architecture as its deep learning algorithm. This network fuses high-level semantic features from the i-th layer of the encoder with low-level detail features from the j-th layer of the decoder using a feature fusion function, and is trained using a weighted cross-entropy loss function to achieve pixel-level accurate segmentation of defects such as broken wires, wear, and corrosion. The formula for the feature fusion function is as follows: ; In the formula: For the high-level semantic features of the i-th layer of the encoder, For the low-level detail features of the j-th layer of the decoder, Represents the convolution fusion operation. For channel splicing.

[0010] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes, specifically the method for quantitatively calculating the metal cross-sectional area loss rate, is as follows: For any axial position of the wire rope Based on the defect segmentation binary mask image at this location, the actual effective metal area is calculated. ; And based on the effective metal area Calculate the metal cross-sectional area loss rate at this location. : ; In the formula: This is the nominal metallic cross-sectional area of ​​the wire rope.

[0011] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes uses a safety factor calculated using the following model: ; In the formula: The nominal tensile strength of the wire rope. For maximum working load, This is the dynamic load factor.

[0012] The aforementioned low-dose X-ray non-destructive testing method for elevator wire ropes further includes a life prediction step after the safety assessment step: Based on the metal cross-sectional area loss rate data obtained from this and historical tests, a linear degradation model was used to fit the performance degradation curve; the linear degradation model is as follows: ; In the formula: As the initial value, For degradation rate, For time; Estimating parameters using the least squares method and And calculate the remaining useful life: ; In the formula: It is the critical safety threshold. This represents the current loss rate.

[0013] The aforementioned system for non-destructive testing of elevator wire ropes using low-dose X-rays includes: A low-dose X-ray source is used to generate and emit an X-ray beam that penetrates the steel wire rope of the elevator to be inspected. A digital detector array, positioned opposite the low-dose X-ray source, is used to receive X-ray signals after penetrating the steel wire rope and convert them into digital image sequences. A motion control module is used to drive the elevator wire rope to pass through the irradiation area of ​​the X-ray beam at a constant speed during the detection process; The data acquisition and processing unit is connected to the digital detector array and the motion control module respectively, and is used to receive the digital image sequence and the displacement information of the wire rope; The defect intelligent identification module, integrated into the data acquisition and processing unit, is used to automatically identify, locate, and quantify internal and external defects of the wire rope based on the processed image data using deep learning algorithms.

[0014] Compared with the prior art, the present invention has the following significant advantages: 1. This invention achieves full coverage detection of internal and external defects in elevator wire ropes by combining low-dose X-ray transmission scanning with three-dimensional reconstruction technology. It can accurately identify various defects such as broken wires, wear, corrosion, and deformation, and eliminate hidden safety hazards.

[0015] 2. This invention employs preprocessing methods with dark field and flat field calibration, along with wavelet transform multi-scale image enhancement technology, to effectively reduce noise interference and highlight defect features. Relying on the U-Net++ deep learning network architecture, it achieves pixel-level accurate segmentation of defects, quantifies and calculates the metal cross-sectional area loss rate and safety factor, resulting in more accurate detection results. Furthermore, the entire process is automated, improving detection efficiency.

[0016] 3. This invention uses a low-dose X-ray source, which significantly improves on-site detection safety while ensuring detection penetration and avoiding the harm of excessive radiation; it realizes three-dimensional volume data reconstruction by associating displacement information, and can arbitrarily extract cross-section analysis, and realize the prediction of remaining lifetime by combining historical data and linear degradation model.

[0017] 4. This invention ensures that the wire rope moves at a constant speed and is synchronized with image acquisition through a motion control module, reducing the impact of jitter; the standardized calibration process eliminates the interference of detector noise and X-ray field inhomogeneity. Compared with the defects of electromagnetic detection method affected by lift-off effect, this invention has higher detection stability and is suitable for routine detection of elevator wire ropes under different working conditions. Detailed Implementation

[0018] The present invention will be further described below with reference to embodiments, but these embodiments are not intended to limit the scope of the invention.

[0019] Example 1: A non-destructive testing method for elevator wire ropes using low-dose X-rays, comprising the following steps: Step 1: Use a low-dose X-ray source to perform a transmission scan on a continuously moving elevator steel cable, and use a digital detector array to receive the transmitted X-ray signals and generate an initial digital image set. This step uses a low-dose X-ray source that meets safety standards. While ensuring penetration (enough to penetrate the steel wire rope body and internal core), it reduces the radiation dose and ensures the safety of on-site testing. During testing, the X-ray source continuously emits a uniform beam of radiation, while the motion control module drives the elevator steel wire rope through the radiation irradiation area at a constant speed, achieving continuous and thorough scanning of the entire length of the steel wire rope and avoiding localized missed detections.

[0020] A digital detector array is positioned opposite the X-ray source to precisely receive the X-ray signal after it penetrates the steel wire rope. The attenuation of X-rays differs when passing through defective parts of the steel wire rope (such as broken wires or corrosion) compared to normal parts, creating differentiated signals. The digital detector array converts this physical signal into an electrical signal in real time, and then performs analog-to-digital conversion to generate an original grayscale image containing structural information about the steel wire rope. Multiple consecutive frames of images together constitute an initial digital image set, providing basic data for subsequent defect analysis.

[0021] Step 2: Synchronously acquire the displacement information of the wire rope through the motion control module, and associate the displacement information with each frame of the initial digital image set; In this step, the motion control module is equipped with a high-precision rotary encoder, which works in conjunction with the traction device of the elevator wire rope to capture the motion state of the wire rope in real time. The encoder converts the mechanical displacement of the wire rope into electrical pulse signals, with each pulse corresponding to a fixed, minute displacement of the wire rope (single-frame displacement). This ensures that the accuracy of displacement measurement matches the accuracy of image acquisition, avoiding subsequent data misalignment due to displacement errors.

[0022] The system has a preset synchronization logic between the image acquisition trigger signal and the encoder pulse signal. When the encoder detects movement of the wire rope... At the same time, the digital detector array is synchronously triggered to acquire one frame of image. Simultaneously, the data acquisition and processing unit records the acquisition time and frame number of each frame of image. ) and the corresponding cumulative displacement, based on the initial position ( ) through the formula " Calculate the precise spatial coordinates corresponding to this frame of image ( ), and bind spatial coordinates to image data for storage.

[0023] Through this association step, discrete two-dimensional image frames are given clear spatial location attributes, solving the problem of ambiguity between the image and the actual location. Subsequently, based on this association relationship, multiple frames of images can be stacked in spatial order to achieve three-dimensional reconstruction. At the same time, the specific location of the defect in the axial direction of the wire rope can be accurately located, providing key spatial dimension support for defect quantitative analysis and safety assessment.

[0024] Step 3: Preprocess and enhance the associated initial digital image set to obtain enhanced image data; This step establishes a stable signal reference by systematically calibrating the detector to eliminate measurement errors caused by inherent noise and X-ray field inhomogeneities. The specific implementation is based on a physical model of X-ray attenuation; the intensity attenuation of X-rays after passing through the steel wire rope follows Beer-Lambert's law. ; In the formula: The intensity of the incident ray, Intensity of transmitted rays The mass decay coefficient of the material. For material density, The penetration thickness.

[0025] In actual testing, it is first necessary to acquire multiple frames of images under X-ray-free conditions and average them to obtain a dark-field image that accurately characterizes the detector's background noise and dark current. .

[0026] Subsequently, multiple frames of images were acquired and averaged under a uniform X-ray field and without a sample (no elevator steel cable) to obtain a flat-field image characterizing the spatial inconsistency of the system response. Based on these benchmark data, the original grayscale image is processed using a formula. The image is standardized to obtain the corrected image. : ; In the formula: Represents the pixel coordinates in the image. It is the average value of the difference between the flat field image and the dark field image, used to restore the overall signal strength to a reasonable level.

[0027] The output of the above preprocessing steps forms the data basis for all subsequent analyses, and its calibration quality directly determines the accuracy and reliability of the final detection results.

[0028] After preprocessing and before image enhancement, a 3D reconstruction sub-step is also included: The corrected images are stacked in spatial order to reconstruct three-dimensional volume data representing the spatial structure of the wire rope. : .

[0029] To evaluate the reconstruction quality, a spatial resolution metric is introduced: ; in, The smallest distinguishable feature size, This spatial resolution enhances the data's dimensionality, allowing analysis to move beyond a few cross-sections and enabling the extraction and analysis of cross-sections and longitudinal profiles at any location, thus achieving a complete digital representation of the wire rope's spatial structure.

[0030] To address the characteristics of weak defect features and susceptibility to noise interference in complex wire rope structures, this step employs a multi-scale enhancement algorithm based on wavelet domain analysis. First, the signal-to-noise ratio of the image is defined: ; In the formula: For signal variance, This represents the noise variance.

[0031] In practice, the corrected image is first subjected to two-dimensional discrete wavelet decomposition: ; In the formula: It carries the overall information of the image (low-frequency components). , , It contains detailed features (high-frequency components) in the horizontal, vertical, and diagonal directions, respectively.

[0032] Then, a nonlinear enhancement transformation is applied to the detailed features. The transformation formula is as follows: ; In the formula: Represents any detail coefficient. To enhance the intensity control factor, To adaptively adjust parameters.

[0033] Finally, the image is reconstructed and enhanced using inverse wavelet transform: ; This processing effectively highlights defect edges and texture features, providing enhanced image data for subsequent identification.

[0034] Step 4: Analyze the enhanced image data based on deep learning algorithms to identify, locate, and segment the internal and external defects of the wire rope; This step employs an improved U-Net++ network architecture to achieve accurate defect identification and segmentation. This network fuses high-level semantic features from the i-th layer of the encoder with low-level detail features from the j-th layer of the decoder using a feature fusion function, and is trained using a weighted cross-entropy loss function to achieve pixel-level accurate segmentation of defects such as broken wires, wear, and corrosion. The formula for the feature fusion function is as follows: ; In the formula: For the high-level semantic features of the i-th layer of the encoder, For the low-level detail features of the j-th layer of the decoder, Represents the convolution fusion operation. For channel splicing.

[0035] The network calculates the class probability for each pixel using the softmax function: ; In the formula: For the predicted category of pixel i, For category indexing, The total number of categories, The unnormalized score is the output of the network.

[0036] The training process uses a weighted cross-entropy loss function: ; In the formula: These are class weights, used to address class imbalance issues. This represents the total number of pixels.

[0037] This step completes the essential transformation from image features to semantic information, providing a direct computational basis for subsequent quantitative analysis.

[0038] Step 5: Based on the defect identification results, quantitatively calculate the metal cross-sectional area loss rate and / or safety factor of the wire rope, and generate a safety assessment report.

[0039] This step transforms the segmentation results into quantifiable engineering safety indicators, with the core being the calculation of the metal cross-sectional area loss rate.

[0040] First, for any axial position of the wire rope Extract the cross-sectional binary mask image at this location from the deep learning segmentation results. In the mask, areas with a pixel value of 1 represent the effective metal area of ​​the steel wire rope, while areas with a pixel value of 0 represent defective areas such as broken wires, corrosion, and wear. Then, based on the physical parameters of the digital detector array and the imaging magnification ratio, the actual physical area corresponding to a single pixel is calculated. Finally, the total number of valid pixels with a value of 1 in the binary mask image is counted using the formula. The actual effective metal area at this axial position was calculated. .

[0041] For three-dimensional defect features, the voxel integration method can be used to perform a weighted average of the effective areas of axially adjacent sections, further reducing local errors. ; In the formula: Represents the entire three-dimensional volume data. This is a binary mask for the defect region.

[0042] Based on effective metal area Calculate the metal cross-sectional area loss rate at this location. : ; In the formula: This is the nominal metallic cross-sectional area of ​​the wire rope.

[0043] Introducing a safety factor assessment model for wire ropes: ; In the formula: The nominal tensile strength of the wire rope. For maximum working load, This is the dynamic load factor.

[0044] Based on this quantitative data, a safety assessment report is generated, which includes key indicators such as the metal cross-sectional area loss rate distribution curve and the broken wire distribution spectrum, providing a comprehensive and objective technical basis for maintenance decisions.

[0045] Furthermore, following the safety assessment step, a lifetime prediction step is also included: Based on the metal cross-sectional area loss rate data obtained from this and historical tests, a linear degradation model was used to fit the performance degradation curve; the linear degradation model is as follows: ; In the formula: As the initial value, For degradation rate, For time; Estimating parameters using the least squares method and : ; ; In the formula: For the detection time point, This corresponds to the metal cross-sectional area loss rate value. This represents the number of data points. Based on this, the remaining useful life is predicted as follows: ; In the formula: It is the critical safety threshold. This represents the current loss rate.

[0046] To quantify prediction uncertainty, confidence interval calculation is introduced: ; In the formula: for Distribution quantiles This represents the standard error.

[0047] This predictive model extends security management from static assessment based on a single detection to dynamic early warning throughout the entire lifecycle. Users can rationally arrange maintenance plans based on the prediction results, optimizing operation and maintenance costs while ensuring security.

[0048] Example 2: This example provides a system for implementing the method described in Example 1, comprising: A low-dose X-ray source is used to generate and emit an X-ray beam that penetrates the elevator steel wire rope under inspection. The low-dose X-ray source is a compact microfocus X-ray source (model: MFX-160), with a tube voltage adjustment range of 30-160kV, a tube current of 50-500μA, and a focal size of ≤10μm. It adopts a pulsed emission mode, with a single irradiation dose ≤0.1mGy / h (compliant with GB18871-2002 radiation safety standards). The X-ray beam is a fan-shaped beam with a divergence angle of 15°, ensuring that a clear transmission signal can be obtained when penetrating elevator steel wire ropes with a diameter ≤60mm (common specification 8×19S+FC-15.5mm), while also ensuring the safety of on-site inspection personnel.

[0049] A digital detector array, positioned opposite the low-dose X-ray source, is used to receive X-ray signals after penetrating the steel wire rope and convert them into a digital image sequence. It employs a flat-panel amorphous silicon digital detector (model: DDA-2000) with an effective detection area of ​​200mm × 200mm, pixel size of 100μm × 100μm, and a total of 2048 × 2048 pixels. It features 16-bit grayscale (dynamic range 0-65535), an adjustable frame rate of 10-30fps, and is synchronously triggered with the X-ray source to ensure precise matching between the acquisition of each frame and the displacement of the steel wire rope. It receives the penetrating X-ray signals and converts them into a digital image sequence in real time.

[0050] The motion control module is used to drive the elevator wire rope to pass through the X-ray beam irradiation area at a constant speed during the detection process. It consists of a high-precision rotary encoder (model: E6B2-CWZ6C), a stepper motor driver (model: DM542), and a PLC controller (model: S7-1200). The rotary encoder has a resolution of 1000 lines / revolution and is coaxially connected to the wire rope traction wheel. A single pulse corresponds to a wire rope displacement Δz=0.1mm, with a measurement accuracy of ±0.01mm. The PLC controller communicates with the data acquisition unit via the Modbus protocol, driving the stepper motor to move the wire rope through the X-ray irradiation area at a constant speed of 0.5-2m / min, thereby achieving stable control of the motion state and real-time acquisition of displacement signals.

[0051] The data acquisition and processing unit is connected to the digital detector array and motion control module respectively, and is used to receive the digital image sequence and the displacement information of the wire rope. It adopts an industrial control computer (CPU: Intel Core i7-12700H, memory 32GB, hard disk 1TB SSD), equipped with a high-speed image acquisition card (model: PCIe-1433), with a sampling rate ≥1GB / s, and supports synchronous reception of signals from the digital detector array and motion control module. It has built-in acquisition software developed based on LabVIEW, which can store the initial digital image set, displacement information and related data in real time, and perform preprocessing, 3D reconstruction, image enhancement and other algorithm calculations.

[0052] The intelligent defect identification module, integrated into the data acquisition and processing unit, is used to automatically identify, locate, and quantify internal and external defects of the wire rope based on deep learning algorithms in the processed image data. Integrated into the data acquisition and processing unit, the hardware features an NVIDIA RTX 3090 graphics card (24GB VRAM) to accelerate deep learning computation; the software utilizes an intelligent identification algorithm developed based on the PyTorch framework, incorporating a U-Net++ network model trained on over 5000 wire rope defect samples (including typical defects such as broken wires, wear, and corrosion). The model inference speed is ≥20fps, and the defect segmentation accuracy is ≥95%, enabling automatic defect identification, location, and quantification analysis.

[0053] The system connection and workflow are as follows: Mechanical Installation and Calibration: Fix the low-dose X-ray source and digital detector array relative to each other on the detection bracket, with a spacing of 500mm, ensuring that the X-ray beam penetrates perpendicularly through the center of the steel wire rope; adjust the traction wheel of the motion control module to fit against the elevator steel wire rope, rigidly connect the rotary encoder to the traction wheel, and calibrate the single-pulse displacement. =0.1mm; detector calibration is completed through dark field and flat field acquisition to ensure image signal stability.

[0054] Signal synchronization and data acquisition: After the system is started, the PLC controller drives the stepper motor to move the wire rope at a constant speed. The rotary encoder collects the displacement signal in real time and transmits it to the data acquisition unit. When the cumulative displacement reaches Δz, the acquisition unit synchronously triggers the X-ray source to emit rays and the detector array to acquire images. Each frame of image is automatically bound to the corresponding spatial coordinates to form an initial digital image set with spatiotemporal labels.

[0055] Data processing and defect identification: The data acquisition and processing unit performs dark-field-flat-field normalization correction on the initial image set and completes the reconstruction of three-dimensional volume data by combining displacement information; after highlighting the defect features through the wavelet transform multi-scale enhancement algorithm, the image data is input into the defect intelligent identification module; the U-Net++ network model performs pixel-level segmentation on the image, outputs a defect binary mask image, and then calculates the metal cross-sectional area loss rate and safety factor, and generates a safety assessment report containing the LMA distribution curve.

[0056] Lifespan prediction and result output: The system calls historical test data, fits the performance degradation curve through a linear degradation model, and calculates the remaining lifespan; the final test results are stored on the local hard drive in the form of a report, and can also be transmitted to the elevator maintenance management platform via Ethernet to provide data support for maintenance decisions.

[0057] Example 3: This example is based on the method of Example 1 and a specific effect test is conducted.

[0058] I. Experimental Samples and Equipment Parameters (a) Experimental Samples Three typical specifications of elevator wire rope were selected (covering common working conditions), and three sets of defect samples (including standard defects and simulated real defects) were set for each specification. The specific parameters are shown in Table 1. Table 1

[0059] (II) The experimental equipment parameters are shown in Table 2: Table 2

[0060] II. Experimental Results and Analysis (a) Defect identification accuracy Nine groups of samples (three specifications × three types of defects) were repeatedly tested (each group of samples was tested 10 times). The defect identification accuracy rate (number of correctly identified defects / actual number of defects × 100%) was calculated. The results are shown in Table 3. Table 3

[0061] Analysis: The average defect identification accuracy of the three specifications of steel wire rope all exceeded 96%, with the highest accuracy for broken wire defects (≥97.8%), mainly due to the significant difference in grayscale in X-ray images of broken wires; the accuracy for corrosion defects was relatively low, but still ≥95.3%, affected by the uneven grayscale distribution in the corrosion area, and some small pitting corrosion was easily interfered with by noise.

[0062] (ii) Measurement accuracy of metal cross-sectional area loss rate (LMA) Using the nominal loss rate of the standard defect sample as a benchmark, the deviations between the measured values ​​and the nominal values ​​obtained by this method are compared, and the results are shown in Table 4: Table 4

[0063] Analysis: The absolute deviation of the metal cross-sectional area loss rate measurement is ≤0.3%, and the relative deviation is ≤3.9%, meeting the engineering testing requirements (allowable relative deviation ≤5%). The measurement deviation of wear and corrosion defects is slightly higher than that of wire breakage defects. Because the boundaries of wear depth and corrosion area are difficult to define precisely in the image, further optimization is needed through 3D reconstruction and image enhancement techniques.

[0064] (III) Comparison of Detection Efficiency Three specifications of steel wire ropes with a length of 10m were selected, and the detection time of this method was compared with that of traditional manual visual inspection and electromagnetic detection methods. The results are shown in Table 5. Table 5

[0065] Analysis: The average detection time of the method of this invention is only 17.6% of that of manual visual inspection and 36.3% of that of electromagnetic inspection, significantly improving detection efficiency. This is mainly because the method achieves full automation, requiring no manual intervention in image acquisition, processing, and defect identification, and can continuously scan without omissions.

[0066] (iv) Accuracy of safety factor calculation Based on the actual effective metal area Aactual obtained from the test, the safety factor of the wire rope was calculated and compared with the standard safety factor (calculated according to the requirements of GB / T 8918-2018 "Steel Wire Ropes for Important Purposes"). The results are shown in Table 6. Table 6

[0067] Analysis: The relative deviation of the safety factor calculation is ≤2.0%, indicating high accuracy and providing a reliable basis for elevator wire rope safety assessment. The deviation mainly stems from minor errors in the measurement of the metal cross-sectional area and errors in the dynamic load factor Kd (Kd was set to 1.2 in the experiment, which is basically consistent with the actual working conditions).

[0068] (v) Remaining life prediction accuracy Accelerated aging tests were conducted on wear defect samples (initial wear depth 0.8 mm, initial loss rate 6.2%) of 6×37S+IWR specification steel wire rope (15.5 mm). The metal cross-sectional area loss rate was measured every 100 hours for a total of 10 tests. Based on the data from the first 8 tests, a linear degradation model was used to predict the remaining life, and the results were compared with the actual remaining life. The results are shown in Table 7. Table 7

[0069] Analysis: As the amount of test data increases, the accuracy of remaining life prediction gradually improves. When the amount of test data is ≥7 times, the relative error is ≤2.4%, and the prediction results are reliable. The linear degradation model can fit the performance degradation law of wire rope wear defects well and is suitable for life prediction in engineering practice.

[0070] (vi) Verification of radiation dose safety Five radiation dose monitoring points were set up around the detection area (at distances of 1m, 2m, 3m, 4m, and 5m from the X-ray source). Radiation dose was monitored using a radiation dosimeter (model: FD-3013), and the results are shown in Table 8. Table 8

[0071] Analysis: The radiation doses around the detection area were all far below the limit (100 μGy / h) specified in GB18871-2002 "Basic Standards for Ionizing Radiation Protection and Radiation Source Safety". Furthermore, the X-ray source adopted a pulsed emission mode, which further reduced the radiation risk and ensured the safety of on-site detection.

[0072] In summary, this invention achieves a fundamental breakthrough in elevator wire rope detection capabilities through the innovative fusion of low-dose X-ray three-dimensional imaging and deep learning intelligent analysis. It extends the detection range from the surface to the internal structure, enabling for the first time accurate visual identification and precise quantitative assessment of defects such as broken wires, wear, and corrosion. By constructing an automated process, detection efficiency is improved. Furthermore, based on a quantitative model of metal cross-sectional area loss rate and a remaining life prediction algorithm, the safety management model shifts from post-processing to predictive maintenance, significantly enhancing elevator operational safety and maintenance economy.

Claims

1. A method for non-destructive testing of elevator steel wire ropes with low dose X-rays, characterized in that, Includes the following steps: Step 1: Use a low-dose X-ray source to perform a transmission scan on a continuously moving elevator steel cable, and use a digital detector array to receive the transmitted X-ray signals and generate an initial digital image set. Step 2: Synchronously acquire the displacement information of the wire rope through the motion control module, and associate the displacement information with each frame of the initial digital image set; Step 3: Preprocess and enhance the associated initial digital image set to obtain enhanced image data; Step 4: Analyze the enhanced image data based on deep learning algorithms to identify, locate, and segment the internal and external defects of the wire rope; Step 5: Based on the defect identification results, quantitatively calculate the metal cross-sectional area loss rate and / or safety factor of the wire rope, and generate a safety assessment report.

2. The low-dose X-ray non-destructive testing method of elevator steel wire ropes according to claim 1, characterized in that, The preprocessing includes: Acquire multiple images by the digital detector array without X-ray illumination and average to obtain a dark field image ; A flat-field image is obtained by averaging multiple frames of images acquired by a digital detector array in a uniform X-ray field without elevator cables. ; The original gray scale images in the initial digital image set are normalized by using the following calibration formula :​ ； wherein: represents the pixel coordinates in the image, is the full image average of the difference between the flat field image and the dark field image.

3. The method for non-destructive testing of elevator steel wire ropes with low dosage of X-rays according to claim 2, characterized in that, After preprocessing and before image enhancement, a 3D reconstruction sub-step is also included: Based on the displacement information associated with each frame of the image, assign precise spatial coordinates to each frame of the image: ； wherein is an initial position of each image in the initial digital image set, is an image frame number, is a single frame displacement amount; The plurality of frames of the corrected images are stacked in spatial order to reconstruct three-dimensional volume data representing the spatial structure of the steel wire rope : ； The images in the three-dimensional volume data are used for subsequent image enhancement processing.

4. The low-dose X-ray non-destructive testing method of elevator steel wire ropes according to claim 3, characterized in that, The image enhancement specifically refers to multi-scale enhancement based on wavelet transform, including: Discrete wavelet decomposition is performed on the image in the three-dimensional volume data to obtain low-frequency overview components and high-frequency components containing details in the horizontal, vertical, and diagonal directions. The high-frequency components are subjected to a nonlinear enhancement transformation, and the transformation formula is as follows: ； In the formula: represents any detail coefficient, is an intensity control factor, is an adaptive adjustment parameter The enhanced high-frequency components and low-frequency profile components are subjected to inverse wavelet transform to reconstruct the enhanced image.

5. The method for non-destructive testing of elevator steel wire ropes with low dosage of X-rays according to claim 1, characterized in that, The deep learning algorithm employs the U-Net++ network architecture. This network fuses high-level semantic features from the i-th layer of the encoder with low-level detail features from the j-th layer of the decoder using a feature fusion function, and is trained using a weighted cross-entropy loss function to achieve pixel-level accurate segmentation of defects such as broken wires, wear, and corrosion. The formula for the feature fusion function is as follows: ； In the formula: For the high-level semantic features of the i-th layer of the encoder, For the low-level detail features of the j-th layer of the decoder, Represents the convolution fusion operation. For channel splicing.

6. The method for non-destructive testing of elevator steel wire ropes with low dosage of X-rays according to claim 1, characterized in that, The specific method for quantitatively calculating the metal cross-sectional area loss rate is as follows: For any axial position of the steel wire rope , based on the position of the defect segmentation binary mask image, calculate its actual effective metal area ; and based on the effective metal area , the metal cross-sectional area loss rate at the location is calculated : ； In the formula: A is the nominal metallic cross-sectional area of the wire rope.

7. The method of claim 6, wherein the low dose X-ray non-destructive testing of elevator steel wire ropes is performed by using a portable X-ray device. The safety factor is calculated using the following model: ； wherein: is the nominal tensile strength of the steel wire rope, is the maximum working load, is the dynamic load factor.

8. The method for non-destructive testing of elevator steel wire ropes with low dosage of X-rays according to claim 7, characterized in that, Following the safety assessment step, a lifetime prediction step is also included: Based on the metal cross-sectional area loss rate data obtained from this and historical tests, a linear degradation model was used to fit the performance degradation curve; the linear degradation model is as follows: ； wherein: is an initial value, is a degradation rate, is time; Estimating parameters using the least squares method and And calculate the remaining useful life: ； In the formula: is a critical safety threshold, is a current loss rate.

9. System for implementing the method for non-destructive testing of elevator steel ropes with low doses of X-rays according to any one of claims 1-8, characterized in that, include: A low-dose X-ray source is used to generate and emit an X-ray beam that penetrates the steel wire rope of the elevator to be inspected. A digital detector array, positioned opposite the low-dose X-ray source, is used to receive X-ray signals after penetrating the steel wire rope and convert them into digital image sequences. A motion control module is used to drive the elevator wire rope to pass through the irradiation area of ​​the X-ray beam at a constant speed during the detection process; The data acquisition and processing unit is connected to the digital detector array and the motion control module respectively, and is used to receive the digital image sequence and the displacement information of the wire rope; The intelligent defect identification module, integrated into the data acquisition and processing unit, is used to automatically identify, locate, and quantify internal and external defects of the wire rope based on the processed image data using deep learning algorithms.

Images