Method and system for on-line non-destructive detection of internal moisture in a temperature protector

Abstract

The present application relates to the technical field of protector detection, and more particularly to a temperature protector internal micro moisture on-line nondestructive detection method and system, comprising dielectric spectrum scanning of temperature protectors passing through the production line one by one to obtain dielectric response spectrum; comparing and analyzing the dielectric response spectrum with a preset reference spectrum library to obtain a state deviation index; according to the comparison result of the state deviation index and a preset state threshold, the protectors are divided into two categories: one is directly released to the subsequent process, and the other is guided to a time domain reflection analysis station; for the temperature protectors guided to the time domain reflection analysis station, reflection time domain signals are obtained and analyzed to extract moisture characteristic parameters; the moisture characteristic parameters are input into a pre-calibrated conversion model to output moisture content estimates; and the state deviation index and the moisture content estimates are comprehensively analyzed to finally determine the moisture state of the temperature protectors according to a preset decision rule. Through the present application, the on-line detection efficiency and quantitative analysis accuracy can be ensured.

Description

Technical Field

[0001] This invention relates to the technical field of protector testing, and in particular to an online non-destructive testing method and system for micro-moisture inside a temperature protector. Background Technology

[0002] Temperature protectors are critical safety components in electrical equipment, and their internal sealing performance directly affects the reliability of equipment operation. During the production process, defects in the sealing process may cause trace amounts of moisture to enter the protector, leading to problems such as insulation degradation and corrosion of metal contacts, ultimately causing the protector to fail. Therefore, strict control of internal moisture levels is necessary during the online monitoring phase of the production line.

[0003] Existing non-destructive testing methods for moisture inside temperature protectors mainly employ methods based on the dielectric response principle. Detection technology based on broadband dielectric spectrum is used for rapid screening on production lines due to its fast response speed and ease of online integration. However, since the dielectric response is a comprehensive reflection of the properties of all materials inside the device, it is easily affected by multiple factors such as the shell material, internal structure, and filling medium. Its specific identification capability for moisture content is limited, and it is difficult to achieve high-precision quantitative judgment when used alone, which can easily lead to misjudgment or missed detection.

[0004] Therefore, there is an urgent need for a non-destructive testing method that can ensure both online detection efficiency and quantitative analysis accuracy, so as to achieve full screening of trace moisture inside temperature protectors without affecting the production rhythm. Summary of the Invention

[0005] This invention provides an online non-destructive testing method and system for micro-moisture inside a temperature protector, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for online non-destructive testing of internal moisture in a temperature protector, the method being used for online testing on a temperature protector production line, comprising:

[0008] Dielectric spectrum scanning was performed on each temperature protector passing through the production line to obtain the dielectric response spectrum, which was used to characterize the overall dielectric properties of the temperature protector.

[0009] The dielectric response spectrum is compared and analyzed with a preset reference spectrum library to obtain a state deviation index;

[0010] Based on the comparison results between the state deviation index and the preset state threshold, the protector is divided into two categories: directly allowing flow to subsequent processes or guiding it to the time domain reflectance analysis station.

[0011] For the temperature protector guided to the time-domain reflectometry analysis station, the reflection time-domain signal is acquired and analyzed to extract moisture characteristic parameters;

[0012] The moisture characteristic parameters are input into a pre-calibrated conversion model to output an estimated moisture content.

[0013] Based on the combined state deviation index and the estimated moisture content, the moisture status of the temperature protector is finally determined according to the preset decision rules.

[0014] Furthermore, the dielectric response spectrum is compared and analyzed with a preset reference spectrum library to obtain a state deviation index, including:

[0015] Multiple qualified temperature protectors that have been confirmed to be dried are selected from the current production batch, and their dielectric response spectra are collected. Through statistical analysis, a reference spectrum library and its normal fluctuation range for this batch are established.

[0016] For each temperature protector whose dielectric response spectrum is obtained by scanning, the distance between it and the reference spectrum in the reference spectrum library in the multidimensional feature space is calculated, and the distance is quantified as the state deviation index.

[0017] Furthermore, the dielectric spectrum scanning employs a wideband coupled sensor array positioned on the main conveyor path of the production line;

[0018] The time-domain reflectometry station is independently located beside the main transmission path and uses a common-path reflective terahertz time-domain spectroscopy system.

[0019] Furthermore, the extracted moisture characteristic parameters include:

[0020] Time-frequency analysis is performed on the reflected time-domain signal to extract the signal attenuation slope and phase delay change within a set terahertz frequency band as the moisture characteristic parameters.

[0021] Furthermore, the preset decision rules include the identification of moisture exceeding the standard state and abnormal states within the non-water category.

[0022] Furthermore, the wideband coupled sensor array is arranged in segments along the main conveying path of the production line for omnidirectional scanning of different surfaces of the temperature protector;

[0023] Each set of broadband coupled sensors integrates a ranging module and dynamically adjusts the signal coupling parameters based on the ranging results.

[0024] Furthermore, the preset benchmark spectral library is established based on qualified samples from the current production batch, and features are extracted by screening key frequency points that are sensitive to moisture.

[0025] The state deviation index is calculated using a weighted distance algorithm, where the weights of each feature dimension are determined based on moisture sensitivity.

[0026] Furthermore, the preset state threshold is a grading threshold determined based on the statistical characteristics of the current batch and the moisture calibration experiment, used to classify the protector into three categories: normal, suspected, and abnormal flow.

[0027] Furthermore, the preset decision rules include:

[0028] If the estimated moisture content exceeds the threshold for excessive moisture, it is determined to be excessive moisture.

[0029] If the estimated moisture content does not exceed the standard but the deviation index of the state exceeds the normal range, it should be further compared with the internal abnormal reference range of the non-water category to distinguish between moisture and non-moisture abnormalities.

[0030] For situations that fall within the judgment boundary, initiate a review test or manual review process.

[0031] On the other hand, the present invention also provides an online non-destructive testing system for trace moisture inside a temperature protector, the system being used for online testing on a temperature protector production line, comprising:

[0032] The dielectric spectrum scanning module is used to scan the dielectric spectrum of each temperature protector passing through the production line to obtain the dielectric response spectrum, which is used to characterize the overall dielectric properties of the temperature protector.

[0033] The deviation analysis module is used to compare and analyze the dielectric response spectrum with a preset reference spectrum library to obtain a state deviation index.

[0034] The diversion control module is used to control the protector to either directly allow flow to subsequent processes or guide it to the time domain reflectance analysis station based on the comparison result between the state deviation index and the preset state threshold.

[0035] The time-domain reflectometry module is used to acquire and analyze the reflection time-domain signal of the temperature protector guided to the time-domain reflectometry analysis station, and extract moisture characteristic parameters.

[0036] The moisture estimation module is used to input the moisture characteristic parameters into a pre-calibrated conversion model and output the moisture content estimate.

[0037] The status determination module is used to combine the status deviation index and the moisture content estimate, and make a final determination on the moisture status of the temperature protector according to the preset decision rules.

[0038] The technical solution of this invention achieves the following technical effects: First, a dielectric spectroscopy scan is used to quickly screen all temperature protectors on the production line. Based on the state deviation index, product flow is differentiated, and only products suspected of moisture intrusion with abnormal states are guided to the time-domain reflectometry (TDRS) analysis station. Then, leveraging the strong specificity of TDRS for moisture, moisture characteristic parameters are extracted, ensuring the control requirements of full screening on the production line, avoiding the omission of potentially excessive moisture products, and preventing a decrease in production efficiency caused by high-precision testing of all products. Simultaneously, online detection efficiency and quantitative analysis accuracy are guaranteed. The state deviation index, reflecting the overall dielectric characteristic deviation of the product, is combined with the specific quantitative result for moisture, i.e., the moisture content estimate. Pre-set decision rules distinguish between characteristic deviations caused by moisture and those caused by non-moisture factors, compensating for the insufficient anti-interference capability of single dielectric spectroscopy detection, avoiding the limitation of a lack of overall perspective in single TDRS detection, and improving the reliability of the judgment results.

[0039] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating the online non-destructive testing method for micro-moisture inside the temperature protector of the present invention.

[0042] Figure 2 This is a structural block diagram of the online non-destructive testing system for micro-moisture inside the temperature protector of the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0045] like Figure 1 As shown, this invention provides an online non-destructive testing method for trace moisture inside a temperature protector. The method is applied to online testing on a temperature protector production line and specifically includes the following steps:

[0046] Step S100: Using a wideband coupled sensor array set on the main conveying path of the production line, the dielectric spectrum of each temperature protector passing through the production line is scanned to obtain the dielectric response spectrum, which is used to characterize the overall dielectric properties of the temperature protector.

[0047] Step S200: Compare and analyze the dielectric response spectrum with a preset reference spectrum library to obtain the state deviation index;

[0048] Step S300: Based on the comparison result between the state deviation index and the preset state threshold, the protector is classified into one that can be directly released to the subsequent process or guided to the time domain reflectance analysis station.

[0049] Step S400: For the temperature protector guided to the time domain reflectance analysis station, acquire and analyze the reflection time domain signal, and extract the moisture characteristic parameters;

[0050] Step S500: Input the moisture characteristic parameters into the pre-calibrated conversion model and output the moisture content estimate;

[0051] Step S600: Combining the state deviation index and the moisture content estimate, the moisture status of the temperature protector is finally determined according to the preset decision rules.

[0052] In this embodiment, all temperature protectors on the production line are first rapidly screened using dielectric spectroscopy scanning. Based on the state deviation index, the product flow is distinguished, and only products suspected of moisture intrusion with abnormal states are guided to the time domain reflectance analysis station. Then, the strong specificity of time domain reflectance technology for moisture is used to extract moisture characteristic parameters, ensuring the control requirements of full screening of the production line, avoiding the omission of potentially excessive moisture products, and avoiding the decrease in production efficiency caused by high-precision detection of all products. At the same time, it ensures the efficiency of online detection and the accuracy of quantitative analysis. The state deviation index, which reflects the deviation of the overall dielectric properties of the product, is combined with the specific quantitative result for moisture, i.e., the moisture content estimate. The preset decision rules distinguish between characteristic deviations caused by moisture and characteristic deviations caused by non-moisture factors, making up for the lack of anti-interference ability of single dielectric spectroscopy detection, avoiding the limitation of single time domain reflectance detection lacking an overall perspective, and improving the reliability of the judgment results.

[0053] Specifically, the aforementioned two-tiered linkage mode of rapid screening using broadband dielectric spectroscopy and precise quantification using time-domain reflectometry is not a simple superposition of two detection methods. Instead, it achieves product diversion through initial screening using dielectric spectroscopy, concentrating high-precision detection resources on suspected abnormal products. This ensures that all products on the production line are covered by testing while minimizing the impact of high-precision testing on the production rhythm, achieving a balance between the efficiency of batch testing and the accuracy of individual testing—something difficult to achieve with a single detection method. The ability of dielectric spectroscopy to perceive the overall state of a product and the ability of time-domain reflectometry to specifically identify moisture complement each other, not only compensating for their own shortcomings but also generating a synergistic effect. By quickly identifying potential problematic products through dielectric spectroscopy and then locating and quantifying the root cause of the problem through time-domain reflectometry, no additional detection steps are required, simultaneously improving detection efficiency, analytical accuracy, and result reliability.

[0054] In some embodiments of the present invention, existing broadband coupled sensor arrays are mostly arranged in a fixed orientation, which can only cover part of the surface of the protector, resulting in the lack of local feature information in the dielectric response spectrum. In order to characterize the overall dielectric properties of temperature protectors that continuously pass through the production line, the specific implementation method is as follows:

[0055] Multiple sets of broadband coupling sensors are symmetrically arranged on both sides of the main conveyor path of the production line. Each set of sensors contains an independent transmitting unit and a receiving unit. The detection surface of the sensor is polished and kept perpendicular to the conveyor path.

[0056] Multiple sensors are arranged into three scanning sections along the conveying path: front, middle, and rear. The front scanning section has one set of sensors, which detects the front face of the protector and parts of the adjacent two sides. The middle scanning section has two sets of sensors, symmetrically distributed on both sides of the conveying path, which detect the main side, top, and bottom of the protector, respectively. The rear scanning section has one set of sensors, which detects the rear face of the protector and parts of the remaining two sides. This three-section layout achieves full-range coverage scanning of the temperature protector.

[0057] Each sensor integrates a miniature laser ranging module to acquire the instantaneous distance between the sensor and the surface of the protector in real time. This distance signal is transmitted to the closed-loop control module, which dynamically adjusts the output power of the sensor's transmitting unit and the gain coefficient of the receiving unit according to a preset distance reference value. For example, when the distance fluctuates within ±0.5mm of the preset reference value, the transmitting power and gain maintain the reference parameters. When the distance exceeds this range, the transmitting power and gain are adjusted in real time according to a preset proportional coefficient to ensure stable signal coupling strength. The wideband operating range of the sensor is set to 1MHz~10GHz. During the scanning process, it is excited point by point in 10kHz frequency steps. The wideband signal output by the transmitting unit is coupled to the temperature protector via a coupling probe. The receiving unit synchronously acquires the amplitude and phase information of the reflected signal. After being converted into a digital signal by a high-speed analog-to-digital converter chip, it is spliced ​​in frequency order to form a complete dielectric response spectrum. The dielectric response spectrum contains information on the dielectric constant and dielectric loss factor at different frequencies, which fully characterizes the overall dielectric properties of the temperature protector.

[0058] In this embodiment, a segmented, multi-angle sensor layout achieves omnidirectional scanning. Simultaneously, a ranging module and a closed-loop control mechanism are integrated to dynamically adjust coupling parameters, improving both scan coverage integrity and signal coupling stability, rather than simply increasing the number of sensors or adjusting the operating frequency. The omnidirectional scanning layout ensures the dielectric response spectrum covers the dielectric characteristics of the temperature protector from all directions, avoiding feature loss due to unscanned areas. The closed-loop control coupling adjustment mechanism counteracts interference from transmission line vibration and placement deviations, ensuring stable coupling effects for protectors in different locations and batches during scanning, guaranteeing the consistency of the dielectric response spectrum.

[0059] As a preferred embodiment of the above, in order to ensure the stability of the signal coupling strength, the dynamic adjustment logic of the closed-loop control module is specifically implemented as follows:

[0060] Step S101: The preset distance reference value is determined through the optimal coupling experiment between the sensor and the temperature protector. A standard temperature protector with the same material and structure as the production batch is selected. Under the static state of the conveyor line, the distance between the sensor and the surface of the protector is adjusted. The signal coupling strength at each frequency is collected at different distance points. The distance corresponding to the peak value of the coupling strength is used as the preset distance reference value. This reference value is stored in the parameter storage unit of the closed-loop control module.

[0061] Step S102: The closed-loop control module consists of a signal acquisition unit, a comparison unit, a parameter adjustment unit, and a drive unit. The miniature laser ranging module acquires the instantaneous distance between the sensor and the surface of the protector in real time at a sampling frequency adapted to the actual detection requirements. The signal acquisition unit converts the instantaneous distance signal into a digital signal and transmits it to the comparison unit. The comparison unit calculates the deviation between the instantaneous distance and the reference value and divides different adjustment ranges according to the magnitude of the deviation, including a stable range, a small deviation range, and a moderate deviation range.

[0062] Step S103: The parameter adjustment unit has built-in preset power adjustment coefficients and gain adjustment coefficients, which are determined through calibration experiments. Under different distance deviation conditions, the transmit power and receive gain are gradually adjusted, and the adjustment ratio that keeps the coupling strength within a reasonable range of the peak value corresponding to the reference value is recorded. The adjustment coefficients are obtained through statistical fitting. When the instantaneous distance is in a stable range, the parameter adjustment unit outputs the reference transmit power and reference gain, and the drive unit maintains the current parameters of the transmit unit and the receive unit. When the instantaneous distance is in a small deviation range, the parameter adjustment unit calculates the target transmit power and target gain according to the preset formula, and the drive unit converts the digital adjustment signal into an analog signal through the digital-to-analog converter circuit, and adjusts the output of the power amplifier of the transmit unit and the programmable gain amplifier parameters of the receive unit in real time. When the instantaneous distance is in a moderate deviation range, the parameter adjustment unit calculates the target parameters according to the doubled adjustment coefficient, and at the same time sends a distance deviation warning signal to the production line control module to prompt the inspection of the conveyor line operation status or the placement of the protector.

[0063] Step S104: During the adjustment process, the receiving unit collects the amplitude of the coupled signal in real time, and the amplitude signal is fed back to the comparison unit to form a secondary closed-loop verification. The comparison unit compares the feedback amplitude with the reference amplitude. If the feedback amplitude exceeds the reasonable range of the reference amplitude, the parameter adjustment unit fine-tunes the target transmission power and target gain according to a preset ratio until the feedback amplitude returns to the allowable range, ensuring that the signal coupling strength remains stable.

[0064] In this embodiment, the reference distance and adjustment coefficient are calibrated experimentally, and graded adjustment is implemented according to the deviation range. Combined with secondary feedback verification, fine control is achieved to ensure the stability of signal coupling strength. The graded adjustment logic makes the parameter adjustment more in line with the actual distance fluctuation, avoiding over-adjustment or under-adjustment. The secondary feedback verification further corrects the adjustment deviation, ensuring that the coupling strength is stable in the optimal range, so that the dielectric response spectrum collected under different distance fluctuation conditions has consistent signal quality, and reduces the impact of distance interference on the dielectric property characterization.

[0065] In practical implementation, as one example, existing benchmark spectral libraries often use universal spectra across batches, failing to consider the subtle differences in material formulations and process parameters between different production batches, leading to inherent deviations between the benchmark spectrum and the actual detected spectrum. Simultaneously, the comparison process often employs simple distance calculations of dielectric parameters across the entire frequency range, failing to distinguish the varying sensitivity of different frequency characteristics to moisture. Dielectric fluctuations caused by non-moisture factors are easily misjudged as state deviations, affecting the accuracy of the indicators. Therefore, this embodiment establishes a dedicated benchmark spectral library for the current batch and improves the specific response of the state deviation indicator to moisture factors by screening key features and weighted multi-dimensional distance calculations. The specific implementation is as follows:

[0066] Step S201: Select a sufficient quantity of temperature protectors that have undergone drying treatment and whose performance is qualified from the current production batch. The drying treatment must be carried out in a constant temperature and humidity environment to ensure that there is no trace amount of moisture residue inside. Perform dielectric spectrum scanning on each of the selected qualified products to obtain their respective dielectric response spectra. Extract the dielectric constant and dielectric loss factor of multiple key frequency points in each spectrum. The selection of key frequency points is determined based on the influence law of moisture on dielectric properties. Prioritize the selection of characteristic frequencies within the moisture-sensitive frequency band and exclude frequency points that have weak moisture response and are easily interfered with by the shell material.

[0067] Step S202: Through statistical analysis, the extracted feature parameters are processed to calculate the mean and standard deviation of the dielectric constant and dielectric loss factor at each key frequency point. The mean is used as the feature parameter value of the reference spectrum, and the range of the mean plus or minus three times the standard deviation is used as the normal fluctuation range of the batch. The reference parameters and fluctuation ranges of all key frequency points are integrated to construct a reference spectrum library exclusive to the current batch.

[0068] Step S203: For the dielectric response spectrum of each temperature protector obtained by online scanning, firstly, the dielectric constant and dielectric loss factor of the corresponding key frequency points are extracted according to the same rules to form the feature vector to be detected; at the same time, the reference parameters of the corresponding key frequency points are retrieved from the reference spectrum library to form the reference feature vector. The two together constitute a multi-dimensional feature space; in order to highlight the weight of the moisture-sensitive feature, the weight coefficient of each key frequency point is determined through calibration experiments. The higher the moisture sensitivity of the frequency point, the larger the weight coefficient. The calibration experiment needs to inject a trace amount of moisture into the standard sample, record the change amplitude of the dielectric parameters at different frequency points, and determine the weight coefficient according to the proportion of the change amplitude.

[0069] Step S204: The weighted Euclidean distance algorithm is used to calculate the distance between the feature vector to be detected and the reference feature vector in the multidimensional feature space. During the calculation, the difference of each feature dimension is first multiplied by the corresponding weight coefficient, and then the sum of squares is calculated and the square root is taken. The weighted distance obtained is the state deviation index.

[0070] This embodiment establishes a dedicated benchmark spectral library based on the current batch of products, screens key moisture-sensitive features and assigns differentiated weights, and obtains the state deviation index through weighted multidimensional distance calculation. The dedicated benchmark spectral library for the current batch eliminates benchmark deviations caused by differences in materials and processes between different batches, making the comparison more targeted. The key feature screening and weighted calculation highlight the influence of moisture-sensitive frequency bands and reduce the interference of dielectric changes caused by non-moisture factors such as slight fluctuations in shell material on the index. Weighted multidimensional distance calculation can more accurately quantify the degree of deviation between the spectrum to be detected and the benchmark spectrum, making the state deviation index more reflective of changes in dielectric properties caused by moisture intrusion.

[0071] In a specific implementation, as one example, if the diversion logic only sets up two options: release and fine measurement, lacking an intermediate transition level, some products with slight deviations but not due to moisture factors are easily over-guided to the time-domain reflectometry station, increasing unnecessary inspection costs. Meanwhile, some products with critical deviations may be missed due to a single threshold judgment. Therefore, this embodiment establishes a batch-specific hierarchical threshold system, combined with a three-level diversion logic and an independently set time-domain reflectometry station, ensuring inspection accuracy while avoiding impacting the efficiency of the main conveyor line. The specific implementation is as follows:

[0072] The determination of preset state thresholds relies on the characteristics of the current batch's test data. Before the formal testing of the current batch of products, a selection of qualified products that have passed the S200 test are chosen, and the distribution of their state deviation index is statistically analyzed. Combined with the results of moisture calibration experiments, including the deviation index corresponding to standard samples injected with different trace moisture contents, three threshold intervals are defined to form a graded threshold system. The first threshold is the 95th percentile of the deviation index of qualified products in the current batch, and the second threshold is the deviation index corresponding to the critical value of trace moisture in the moisture calibration experiment, i.e., the minimum moisture content that may cause subsequent failure risks. The two thresholds divide the deviation range into normal interval, suspected interval, and abnormal interval.

[0073] After the online temperature protector calculates the state deviation index through S200, it is compared with the classification threshold. If the index falls within the normal range, the dielectric properties of the product are determined to be normal, the diversion actuator of the main conveyor remains in place, and the product is directly released to the subsequent process; if the index falls within the suspected range, the product is determined to have potential abnormalities but not reach the high-risk level, the diversion actuator activates lateral guidance to guide the product to the time domain reflectance analysis station for precise verification, and at the same time, an identifiable suspected label is affixed to the product surface; if the index falls within the abnormal range, the product is determined to have a high risk of moisture intrusion, the diversion actuator quickly guides the product to a separate abnormal product temporary storage area, and simultaneously sends an early warning signal to the production line control module to prompt the staff to handle it first.

[0074] The time-domain reflectometry (TDAR) analysis station is independently located beside the main conveyor path and connected to the main conveyor line via a short-distance transition conveyor line. The conveyor line maintains the same speed as the main conveyor line to avoid product jamming or collisions. An independent positioning device is installed within the station, automatically positioning the product to the detection reference position upon entry, ensuring consistency in TDAR detection. After detection, if the deviation is determined to be due to factors other than moisture, the product is returned to the main conveyor line via a return conveyor line for subsequent processes. If the moisture content is determined to be excessive, the product is transferred to the abnormal product storage area, achieving a coordinated effect of continuous flow on the main conveyor line, accurate verification of suspected products, and rapid isolation of abnormal products.

[0075] This embodiment constructs a graded threshold system based on the deviation distribution of the current batch and moisture calibration experiments, designs a three-level diversion logic, and coordinates with an independent and interconnected time-domain reflectometry station to achieve a balance between detection accuracy and production efficiency. Specifically, the batch-specific graded threshold system ensures that the judgment criteria are highly compatible with the characteristics of the current batch of products, avoiding misjudgments caused by general thresholds; the three-level diversion logic can accurately distinguish products of different risk levels, reduce the occupation of the time-domain reflectometry station by slightly deviating products, and ensure the continuous flow efficiency of the main conveyor line; the independently set time-domain reflectometry station is linked with the main conveyor line through transition and return conveyor lines, which does not interfere with the main production rhythm and ensures that suspected products are fully verified.

[0076] In some embodiments of the present invention, for the temperature protector guided to the time-domain reflectometry analysis station, it is necessary to obtain its internal moisture characteristic information. Existing common-path reflective terahertz time-domain spectroscopy systems are susceptible to minor surface tilts, localized stains, or transmission positioning deviations during detection, leading to unstable optical path coupling and baseline drift in the reflected time-domain signal. Furthermore, time-frequency analysis often performs uniform processing across the entire set terahertz frequency band without further screening for the most moisture-sensitive sub-bands. Signal changes caused by non-moisture factors such as minor inhomogeneities in the shell material or fluctuations in the density of the internal filling medium are easily included in the feature extraction range, resulting in insufficient specificity of moisture characteristic parameters and affecting subsequent quantitative accuracy. To solve these problems, this embodiment improves the quality of the reflected time-domain signal and the moisture specificity of characteristic parameters by optimizing the optical path stability control of the common-path system and combining it with precise screening of moisture-sensitive sub-bands. The specific implementation is as follows:

[0077] Step S401: The optical path structure of the common-path reflective terahertz time-domain spectroscopy system is equipped with a real-time calibration module. The system includes a terahertz pulse emitter, a polarization beam splitter, a collimating lens, a reflector, and a terahertz detector. The collimating lens and the reflector are both equipped with a micro-electric adjustment mechanism, and the system integrates a laser positioning unit. The emission direction of the laser positioning unit is coaxial with the propagation direction of the terahertz pulse. Before acquiring the reflected time-domain signal, the laser positioning unit emits a positioning laser to the surface of the temperature protector, receives the reflected laser signal, and calculates the tilt angle and the distance from the center of the protector surface. Based on this, the focal length of the collimating lens and the angle of the reflector are adjusted by the micro-electric adjustment mechanism to ensure that the terahertz pulse is perpendicularly incident on the surface of the protector, thus ensuring stable optical path coupling.

[0078] Step S402: The operating frequency band of the terahertz pulse transmitter is set to the preset terahertz frequency band, the transmission power is kept constant, and the detection distance is fixed to the optimal coupling distance verified by experiments; during the signal acquisition process, the terahertz detector synchronously acquires the reflected time domain signal at the appropriate sampling frequency, the acquisition duration covers the complete terahertz pulse cycle, and the original signal is filtered to remove high-frequency noise.

[0079] Step S403: Time-frequency analysis uses wavelet transform algorithm to process the reflected time-domain signal. First, the preset terahertz frequency band is divided into multiple continuous sub-bands. The bandwidth of each sub-band is determined according to the absorption characteristics of water on terahertz waves. Moisture-sensitive sub-bands are screened through moisture calibration experiment: different amounts of trace moisture are injected into standard samples, and the signal response in each sub-band is collected. Sub-bands whose signal change amplitude is linearly correlated with the moisture content are selected as moisture-sensitive sub-bands, and sub-bands with weak signal response or no significant correlation with moisture content are excluded.

[0080] Step S404: In the moisture-sensitive sub-frequency band, feature extraction is performed on the signal after time-frequency analysis: the linear fitting slope of the signal amplitude change with time in the sub-frequency band is calculated as the signal attenuation slope; by comparing the phase difference between the reflected signal and the reference signal in the sub-frequency band, the reference signal being the reflected signal of the dry standard sample, the cumulative change of phase over time is statistically analyzed as the phase delay change, and the above two parameters together constitute the moisture characteristic parameters.

[0081] This embodiment achieves real-time optical path calibration through laser positioning and electric adjustment. Combined with moisture calibration experiments, sensitive sub-frequency bands are selected for feature extraction, improving optical path stability and feature specificity. The real-time optical path calibration mechanism can counteract interference caused by protector surface tilt and positioning deviation, ensuring baseline stability of the reflected time-domain signal and reducing signal distortion. The selection of moisture-sensitive sub-frequency bands focuses feature extraction on signal changes induced by moisture, reducing interference from non-moisture factors on feature parameters. The signal attenuation slope and phase delay change are both extracted based on the moisture-sensitive sub-frequency bands, and their variation patterns have a clear correlation with moisture content.

[0082] In a specific implementation, as one example, the construction of the conversion model is based on the characteristics of the current production batch. A standard temperature protector with the same material and structure as the current batch is selected, and samples with multiple gradient moisture contents are prepared. Each sample is injected with a trace amount of moisture through precise measurement. After standing until the moisture is evenly distributed, time-domain reflectometry is performed on each gradient moisture sample in sequence, and the corresponding signal attenuation slope and phase delay change are collected. At the same time, the actual moisture content of the sample is determined by the laboratory standard detection method. Each set of moisture characteristic parameters is associated with the corresponding actual moisture content to form a complete calibration dataset.

[0083] The conversion model is constructed using a weighted fitting algorithm, assigning weights according to the sensitivity of moisture characteristic parameters. Correlation analysis is used to process the signal attenuation slope and phase delay change in the calibration dataset, calculating the correlation coefficients between these two parameters and the actual moisture content. A higher correlation coefficient indicates stronger sensitivity to moisture content. The weight allocation ratio is determined based on the correlation coefficients, with parameters of higher sensitivity receiving greater weights to ensure that the model prioritizes responding to moisture-induced characteristic changes during calculation. Based on the weighted characteristic parameters and the actual moisture content, a nonlinear fitting algorithm is used to construct the conversion model. During the fitting process, the model coefficients are iteratively optimized to minimize the estimation error until the deviation between the model's prediction results and the actual moisture content on the validation set meets the preset accuracy requirements.

[0084] The conversion model incorporates a dynamic calibration mechanism that operates on a fixed testing cycle. After a set number of products have been tested, a portion of the products whose moisture content has been precisely tested in the laboratory are selected as calibration samples. Their moisture characteristic parameters and laboratory test values ​​are input into the conversion model, and the deviation between the model's current estimate and the laboratory test value is calculated. If the deviation exceeds the allowable range, the model's fusion weights and fully connected layer parameters are updated using incremental training. This eliminates the need to retrain the entire model, ensuring that the model's adaptability remains consistent.

[0085] During online testing, the moisture characteristic parameters of the product to be tested are input into the calibrated conversion model. The conversion model calculates according to the preset parameters and directly outputs the estimated moisture content.

[0086] In this embodiment, batch adaptive construction ensures that the model parameters are highly matched with the characteristics of the current batch of products, reducing the systematic bias caused by the general model; the weighted fitting logic highlights the role of highly sensitive parameters and reduces the interference of single parameter noise on the estimation results; the dynamic calibration mechanism can correct the estimation bias caused by environmental, equipment and other factors in real time, maintaining the long-term detection accuracy of the model; the final output moisture content estimate has high adaptability to the current detection scenario.

[0087] In practical implementation, if there is no distinction between situations where the state deviation is abnormal but the moisture content is not excessive, and situations where the moisture content is excessive but the state deviation is abnormal, it is easy to treat non-moisture abnormal products as having excessive moisture content, or to miss the accurate control of products with excessive moisture content. Therefore, it is necessary to construct a hierarchical decision rule with two parameters linked together, embedding the identification logic of moisture and non-moisture abnormality classification, and calibrating the decision threshold in combination with the characteristics of the current batch to achieve accurate judgment. The specific implementation is as follows:

[0088] The preset decision rules are constructed based on the current batch's testing and verification data to determine the moisture content exceeding threshold and the state deviation grading threshold. The moisture content exceeding threshold is determined through laboratory failure verification. A standard sample consistent with the current batch is selected, injected with different moisture contents, and subjected to accelerated aging experiments. The lowest moisture content that causes insulation degradation or contact corrosion in the protector is recorded as the moisture content exceeding threshold. Simultaneously, the state deviation grading threshold set by S300 is used to supplement the state deviation reference range for non-moisture-related anomalies. This range is derived from statistical analysis of testing data on non-moisture defect samples from the current batch, such as uneven shell material or misaligned filling medium. The decision rules are divided according to a two-parameter combination scenario, with the specific logic as follows:

[0089] After obtaining the state deviation index and moisture content estimate of the product to be judged through online detection, the threshold ranges corresponding to the two are first compared. If the state deviation falls within the normal range and the moisture content estimate is lower than the exceeding threshold, the moisture state is judged to be normal and the product is released to the subsequent process. If the moisture content estimate is higher than the exceeding threshold, even if the state deviation is normal, a review process is initiated. The dielectric response spectrum and reflection time domain signal of the product are retrieved again, and the detection calculations from S200 to S500 are repeated to eliminate accidental detection errors. If the moisture content still exceeds the standard after the review, it is judged to be excessive moisture.

[0090] If the deviation falls into the suspected or abnormal range, but the estimated moisture content is below the threshold, the deviation is further compared with the non-water classification anomaly reference range. If the deviation meets the reference range and the product has no other appearance defects, it is judged as an internal anomaly of the non-water classification, marked, and transferred to a special maintenance station. If the deviation exceeds the reference range, a second time domain reflectance test is initiated to eliminate the error in the extraction of moisture characteristic parameters. If the moisture estimate is still normal after the second test, the non-water classification anomaly judgment is maintained.

[0091] If the deviation from the standard falls into the abnormal range and the estimated moisture content is higher than the threshold, it is directly judged as a state of excessive moisture and transferred to the abnormal product temporary storage area. The test data of the product, including dielectric response spectrum, moisture characteristic parameters, and estimated results, are linked at the same time to facilitate subsequent quality traceability. For suspected boundary values ​​such as moisture content close to the threshold or deviation from the standard being at the border between suspected and abnormal ranges, a manual review channel is added. Combining product appearance inspection and rapid laboratory verification, the judgment results are ensured to be reliable.

[0092] During online judgment, the dual-parameter data of the product to be tested is automatically retrieved, and the decision-making scenario is matched step by step according to the above logic. The judgment result and corresponding processing instructions are output in real time. The judgment result and processing suggestions are synchronously fed back to the production line control module to guide the action of the diversion execution mechanism.

[0093] In this embodiment, the dual-parameter linkage decision and identification logic can accurately distinguish between excessive moisture and internal anomalies in non-water categories, avoiding improper handling caused by misjudgment; the review and manual review channels compensate for the uncertainty of boundary value judgment and ensure the reliability of the results; the threshold calibrated based on the characteristics of the current batch makes the decision rules adapt to batch differences, and the judgment results are highly consistent with the actual product status.

[0094] Based on the same inventive concept as the online non-destructive testing method for trace moisture inside the temperature protector in the foregoing embodiments, this invention also provides an online non-destructive testing system for trace moisture inside the temperature protector, such as... Figure 2 As shown, the system is used for online testing on a temperature protector production line. The system includes:

[0095] The dielectric spectrum scanning module is used to scan the dielectric spectrum of each temperature protector passing through the production line to obtain the dielectric response spectrum, which is used to characterize the overall dielectric properties of the temperature protector.

[0096] The deviation analysis module is used to compare and analyze the dielectric response spectrum with a preset reference spectrum library to obtain a state deviation index.

[0097] The diversion control module is used to control the protector to either directly allow flow to subsequent processes or guide it to the time domain reflectance analysis station based on the comparison result between the state deviation index and the preset state threshold.

[0098] The time-domain reflectometry module is used to acquire and analyze the reflection time-domain signal of the temperature protector guided to the time-domain reflectometry analysis station, and extract moisture characteristic parameters.

[0099] The moisture estimation module is used to input the moisture characteristic parameters into a pre-calibrated conversion model and output the moisture content estimate.

[0100] The status determination module is used to combine the status deviation index and the moisture content estimate, and make a final determination on the moisture status of the temperature protector according to the preset decision rules.

[0101] The system described above in this invention can effectively realize the online non-destructive detection method for trace moisture inside the temperature protector, and the technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0102] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for online non-destructive testing of micro-moisture inside a temperature protector, characterized in that, The method is applied to online testing on a temperature protector production line, and the method includes: Dielectric spectrum scanning was performed on each temperature protector passing through the production line to obtain the dielectric response spectrum, which was used to characterize the overall dielectric properties of the temperature protector. The dielectric response spectrum is compared and analyzed with a preset reference spectrum library to obtain a state deviation index; Based on the comparison results between the state deviation index and the preset state threshold, the protector is divided into two categories: directly allowing flow to subsequent processes or guiding it to the time domain reflectance analysis station. For the temperature protector guided to the time-domain reflectometry analysis station, the reflection time-domain signal is acquired and analyzed to extract moisture characteristic parameters; The moisture characteristic parameters are input into a pre-calibrated conversion model to output an estimated moisture content. Based on the combined state deviation index and the estimated moisture content, the moisture status of the temperature protector is finally determined according to the preset decision rules.

2. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 1, characterized in that, The dielectric response spectrum is compared and analyzed with a preset reference spectrum library to obtain a state deviation index, including: Multiple qualified temperature protectors that have been confirmed to be dried are selected from the current production batch, and their dielectric response spectra are collected. Through statistical analysis, a reference spectrum library and its normal fluctuation range for this batch are established. For each temperature protector whose dielectric response spectrum is obtained by scanning, the distance between it and the reference spectrum in the reference spectrum library in the multidimensional feature space is calculated, and the distance is quantified as the state deviation index.

3. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 2, characterized in that, The dielectric spectrum scanning employs a wideband coupled sensor array positioned on the main conveyor path of the production line. The time-domain reflectometry station is independently located beside the main transmission path and uses a common-path reflective terahertz time-domain spectroscopy system.

4. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 3, characterized in that, The extracted moisture characteristic parameters include: Time-frequency analysis is performed on the reflected time-domain signal to extract the signal attenuation slope and phase delay change within a set terahertz frequency band as the moisture characteristic parameters.

5. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 4, characterized in that, The preset decision rules include the identification of excessive moisture levels and abnormal states within the non-water category.

6. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 3, characterized in that, The wideband coupled sensor array is arranged in segments along the main conveying path of the production line for omnidirectional scanning of different surfaces of the temperature protector. Each set of broadband coupled sensors integrates a ranging module and dynamically adjusts the signal coupling parameters based on the ranging results.

7. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 6, characterized in that, The preset benchmark spectral library is established based on qualified samples from the current production batch, and features are extracted by screening key frequency points that are sensitive to moisture. The state deviation index is calculated using a weighted distance algorithm, where the weights of each feature dimension are determined based on moisture sensitivity.

8. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 1, characterized in that, The preset state threshold is a grading threshold determined based on the statistical characteristics of the current batch and the moisture calibration experiment, used to classify the protector into three categories: normal, suspected, and abnormal flow.

9. The online non-destructive testing method for trace moisture inside a temperature protector according to claim 5, characterized in that, The preset decision rules include: If the estimated moisture content exceeds the threshold for excessive moisture, it is determined to be excessive moisture. If the estimated moisture content does not exceed the standard but the deviation index of the state exceeds the normal range, it should be further compared with the internal abnormal reference range of the non-water category to distinguish between moisture and non-moisture abnormalities. For situations that fall within the judgment boundary, initiate a review test or manual review process.

10. An online non-destructive testing system for trace moisture inside a temperature protector, the system being used for online testing on a temperature protector production line, characterized in that... include: The dielectric spectrum scanning module is used to scan the dielectric spectrum of each temperature protector passing through the production line to obtain the dielectric response spectrum, which is used to characterize the overall dielectric properties of the temperature protector. The deviation analysis module is used to compare and analyze the dielectric response spectrum with a preset reference spectrum library to obtain a state deviation index. The diversion control module is used to control the protector to either directly allow flow to subsequent processes or guide it to the time domain reflectance analysis station based on the comparison result between the state deviation index and the preset state threshold. The time-domain reflectometry module is used to acquire and analyze the reflection time-domain signal of the temperature protector guided to the time-domain reflectometry analysis station, and extract moisture characteristic parameters. The moisture estimation module is used to input the moisture characteristic parameters into a pre-calibrated conversion model and output the moisture content estimate. The status determination module is used to combine the status deviation index and the moisture content estimate, and make a final determination on the moisture status of the temperature protector according to the preset decision rules.

Images