An automatic identification and detection system for defects in the inner wall of air conditioning ducts

Abstract

This invention discloses an automatic identification and detection system for defects in the inner wall of air conditioning ducts, relating to the field of air conditioning duct inspection technology. The system comprises a detection unit and an analysis unit, which form a bidirectional closed-loop collaborative structure through a data transmission module. The detection unit includes a fusion detection module, an anti-interference module, and an adjustment support. The fusion detection module integrates multiple detection sub-modules to achieve complementary signal calibration, the anti-interference module filters interference and adaptively adjusts detection parameters, and the adjustment support adapts to different pipe diameters and protects the pipes. The analysis unit includes an identification and grading module and a trend prediction module, relying on algorithms to achieve automatic defect identification and grading, trend prediction, and early warning. The analysis unit's feedback guides the detection unit to dynamically adjust detection parameters and strategies, achieving deep collaboration between detection and analysis, and adapting to the automated detection needs of inner wall defects in air conditioning ducts under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of air conditioning duct inspection technology, specifically to an automatic identification and detection system for defects in the inner wall of air conditioning ducts. Background Technology

[0002] As a core component of air conditioning systems, the condition of the inner wall of air conditioning ducts directly affects the cooling efficiency and operational safety of the system. In commercial central air conditioning and old air conditioning systems, ducts are subject to high temperature and humidity and media erosion conditions for a long time, which can easily lead to various defects such as cracks, corrosion and weld blockages on the inner wall. If these defects are not detected and dealt with in a timely and accurate manner, they can easily cause refrigerant leaks, compressor burnout and other failures. Therefore, it is particularly important to conduct efficient and accurate detection of defects on the inner wall of air conditioning ducts.

[0003] In existing technologies for detecting defects in the inner walls of air conditioning ducts, the detection and analysis stages are independent of each other, lacking a closed-loop collaborative design. The defect data collected at the detection end is easily affected by environmental interference and has blind spots. The analysis end cannot achieve intelligent classification and trend prediction of defects based on accurate data. At the same time, the detection parameters and strategies are fixed and cannot be dynamically adjusted according to the actual defect situation. As a result, the detection technology cannot balance the accuracy, intelligent efficiency, and adaptability to operating conditions, and cannot meet the actual needs of comprehensive, accurate, and efficient detection of defects in the inner walls of air conditioning ducts under complex operating conditions.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic identification and detection system for defects in the inner wall of air conditioning ducts, so as to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, this invention provides an automatic identification and detection system for defects in the inner wall of air conditioning ducts, comprising a fusion detection module, an anti-interference module, an adjustment bracket, an identification and grading module, a trend prediction module, and a data transmission module. The fusion detection module, anti-interference module, and adjustment bracket collaboratively constitute a detection unit; the identification and grading module, trend prediction module, and data transmission module collaboratively constitute an analysis unit. The detection unit and analysis unit form a closed-loop collaborative structure through bidirectional linkage via the data transmission module. The fusion detection module includes an ultrasonic detection submodule, an eddy current detection submodule, and a stress wave detection submodule; the anti-interference module includes a noise acquisition sensor and a filtering submodule; the identification and grading module includes a CNN algorithm submodule and a grading submodule; and the trend prediction module includes… It includes an LSTM algorithm submodule and an early warning submodule. The detection unit collects raw data on defects in the inner wall of the air conditioning duct and filters interference signals. The analysis unit analyzes and processes the raw defect data and outputs the analysis results. The analysis unit feeds back the analysis results to the detection unit to guide the detection unit in adjusting the detection parameters and detection strategies. It realizes bidirectional closed-loop collaboration between the detection unit and the analysis unit, breaking the limitations of the separation of detection and analysis in existing technologies. They complement each other's technical shortcomings. Through the collaborative cooperation of various modules and submodules, the accuracy and intelligence level of defect detection are significantly improved. At the same time, a balance is achieved between detection accuracy, detection efficiency, and energy consumption. It adapts to the defect detection needs of air conditioning ducts under complex working conditions and solves the technical pain point that existing technologies cannot achieve accuracy, efficiency, and low consumption when using any one unit alone.

[0007] Furthermore, the ultrasonic testing submodule, eddy current testing submodule, and stress wave testing submodule work together to collect detection signals of different types of defects on the inner wall of the pipeline. The collected detection signals are then calibrated using a fusion algorithm to output accurate raw defect data. Through the collaborative work of multiple detection submodules, the blind spots of single detection methods are filled, enabling comprehensive collection of different types of defects on the inner wall of the pipeline. This improves the completeness and accuracy of the raw defect data, providing high-quality data support for subsequent defect analysis and avoiding the problems of missed or misjudged defects caused by single detection methods.

[0008] Furthermore, the noise acquisition sensor collects interference signals in the detection environment in real time, the filtering submodule filters the interference signals, and automatically adjusts the detection parameters of the fusion detection module to ensure the stability of the detection signal; it realizes adaptive anti-interference in the detection process, effectively filters various interference signals in the environment, avoids the distortion of detection data caused by interference signals, improves the system's operational stability under complex working conditions, and expands the system's adaptability to various scenarios.

[0009] Furthermore, a flexible buffer structure is provided at the end of the adjusting bracket. The adjusting bracket automatically adjusts the spacing between the detection heads according to the inner diameter of the pipe. The flexible buffer structure avoids collision between the detection head and the inner wall of the pipe. This enables adaptive detection of pipes with different diameters without the need to replace the detection equipment, reducing detection costs. At the same time, the flexible buffer structure protects the inner wall of the pipe, avoiding secondary damage to the pipe during the detection process, and adapts to the detection needs of curved pipes.

[0010] Furthermore, the CNN algorithm submodule analyzes the raw defect data output by the detection unit to identify the defect type, and the grading submodule classifies the defect level according to the defect parameters and industry standards; thus realizing automatic identification of defect type and automatic classification of defect level without the need for secondary manual judgment, reducing the professional threshold for inspection personnel, reducing the error of manual judgment, and improving the accuracy and efficiency of defect grading.

[0011] Furthermore, the LSTM algorithm submodule combines pipeline operating parameters and historical detection data to predict the speed of defect development, and the early warning submodule outputs early warning information on defect deterioration based on the prediction results; thus achieving accurate prediction and early warning of defect development trends, giving staff sufficient maintenance time, avoiding pipeline failures caused by defect expansion, and reducing maintenance costs and safety hazards.

[0012] Furthermore, the data transmission module enables bidirectional data transmission between the detection unit and the analysis unit, transmitting the raw defect data output by the detection unit to the analysis unit and feeding back the analysis results output by the analysis unit to the detection unit; ensuring real-time and stable data transmission between the detection unit and the analysis unit, ensuring the normal operation of the closed-loop collaborative structure, realizing bidirectional linkage between detection data and analysis results, and providing timely support for the dynamic adjustment of detection parameters and detection strategies.

[0013] Compared with the prior art, the beneficial effects of the present invention are:

[0014] 1. This invention constructs a two-way closed-loop collaborative structure between the detection unit and the analysis unit, breaking the limitation of traditional independent detection and analysis, realizing two-way linkage feedback between detection data acquisition and data analysis. The analysis unit outputs results based on the detection data and guides the detection unit to dynamically adjust parameters and strategies. The detection unit provides the analysis unit with accurate and interference-free raw data. This solves the core problem of the disconnect between detection and analysis from the architectural level, and greatly improves the overall detection efficiency of the system.

[0015] 2. This invention achieves complementary calibration of multimodal detection signals through a fusion detection module, filling the blind spots of single detection methods, comprehensively collecting defect signals of different types and depths, and filtering complex working condition interference in real time with an anti-interference module and adaptively adjusting detection parameters to effectively avoid data distortion; at the same time, the adjustable support can adapt to different pipe diameters, and the flexible buffer structure prevents pipe damage, adapting to various complex installation scenarios, significantly improving the integrity and accuracy of defect data and the system's working condition adaptability.

[0016] 3. This invention relies on the CNN algorithm to automatically identify defect types and standardize their classification, and combines the LSTM algorithm to integrate multi-dimensional data to complete defect trend prediction and graded early warning. No manual secondary judgment is required, enabling preventive defect handling and reducing system maintenance costs and operational risks. At the same time, it can dynamically optimize the detection strategy to achieve on-demand detection, balance detection accuracy, efficiency, and energy consumption, and generate standardized detection reports and complete data classification, archiving, and dual backup, providing reliable data support for the entire life cycle maintenance of air conditioning pipelines and improving the practicality and traceability of detection results. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an automatic identification and detection system for defects in the inner wall of air conditioning ducts. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This invention provides a technical solution: an automatic identification and detection system for defects in the inner wall of air conditioning ducts. This system is mainly used for detecting defects in the inner wall of commercial central air conditioning ducts, old air conditioning ducts, and other complex working conditions. It can achieve accurate acquisition, intelligent identification, grade classification, trend prediction, and dynamic optimization of detection strategies for defects, solving the technical pain points of existing technologies such as separation of detection and analysis, insufficient accuracy, and poor adaptability.

[0020] The core of this automatic identification and detection system for defects in the inner wall of air conditioning ducts includes a detection unit and an analysis unit. These two units achieve bidirectional linkage through a data transmission module, forming a closed-loop collaborative structure. The overall layout design of each core module and sub-module is designed to solve the problems of independent detection and analysis, poor data transmission, and poor coordination among modules in existing technologies. It is based on the actual needs of air conditioning duct defect detection and combines the application principles of multimodal detection technology, deep learning technology, and anti-interference technology to ensure that the system can adapt to complex working conditions and improve detection accuracy and intelligence.

[0021] The core modules of this system include a fusion detection module, an anti-interference module, an adjustment support identification and grading module, a trend prediction module, and a data transmission module. The fusion detection module, as the core data acquisition component of the detection unit, includes ultrasonic detection, eddy current detection, and stress wave detection sub-modules to comprehensively acquire detection signals of different types of defects on the inner wall of the pipeline. The anti-interference module includes a noise acquisition sensor and a filtering sub-module to filter interference signals in the detection environment and ensure the stability of the detection data. The adjustment support is used to adapt to pipelines of different diameters, avoiding damage to the inner wall of the pipeline during the detection process. The identification and grading module includes a CNN algorithm sub-module and a grading sub-module to analyze the detection data and achieve defect type identification and grading. The trend prediction module includes an LSTM algorithm sub-module and an early warning sub-module to predict the development trend of defects and output early warning information. The data transmission module enables bidirectional data transmission between the detection unit and the analysis unit, ensuring the normal operation of the closed-loop collaborative structure.

[0022] The unique technical approach of this solution lies in breaking away from the traditional design of independent operation of detection and analysis modules in existing technologies. Instead, it constructs a two-way closed-loop collaborative structure between the detection and analysis units, deeply integrating multimodal fusion detection technology with deep learning technology. Through the collaborative cooperation of various modules and sub-modules, it achieves full-process automation of detection data acquisition, interference filtering, defect analysis, trend prediction, and detection strategy optimization. At the same time, it improves the system's adaptability and scalability through modular design.

[0023] Currently available technical documents mainly include an automatic detection device for defects on the inner wall of air conditioning copper pipes, a method and system for detecting defects on the inner wall of pipes, and a pipe defect detection system based on deep learning. These publicly available documents only achieve automatic detection of surface defects on the inner wall of copper pipes, using a single detection method without multimodal fusion detection or intelligent analysis and trend prediction modules, thus failing to achieve intelligent defect classification and trend prediction. They employ a single detection method combined with simple data processing, without constructing a closed-loop collaborative structure, resulting in weak anti-interference capabilities and poor adaptability. While deep learning technology is used for defect identification, it is not deeply integrated with multimodal fusion detection technology, nor does it achieve dynamic optimization of the detection strategy, making it difficult to balance detection accuracy and efficiency. The core difference between this technical solution and the aforementioned publicly available documents lies in the construction of a two-way closed-loop collaborative structure for detection and analysis, integrating multimodal fusion detection and deep learning technology to achieve fully automated detection and intelligent analysis. It also possesses adaptive anti-interference and adaptive pipe diameter adjustment capabilities, solving the technical pain points of existing publicly available technologies that cannot simultaneously achieve accuracy, efficiency, and adaptability.

[0024] The specific implementation process of this system is mainly divided into seven core steps. Taking the inspection of commercial central air conditioning pipelines as the overall application scenario, the pipelines have various diameters, exist in high temperature and high humidity environments, contain condensate and oil residues, generate interference signals due to the operation of multiple devices, and are prone to various types of defects such as cracks, corrosion, weld blockages, and pores on the inner wall of the pipelines. Some pipelines have been in use for a long time, and the development trend of defects is unclear. Therefore, it is necessary to achieve accurate detection, intelligent classification, and early warning.

[0025] Step 1: System Initialization: The system initialization step is the foundation of the entire testing process. It mainly involves starting up each module of the system, calibrating parameters, clearing data, and establishing communication links. This ensures that each module and sub-module can work together normally, laying the foundation for subsequent testing. In existing technologies, the testing system often directly enters the testing phase after startup without sufficient initialization and calibration. This leads to parameter deviations in each module and unstable communication links, thus affecting the accuracy of the testing data and the stability of the system operation. Therefore, a system initialization step is needed to eliminate parameter deviations, establish stable communication links, and ensure the consistency of all system components. Specific technical methods are as follows:

[0026] Power on the system's main power supply, then sequentially start the detection and analysis units. Perform self-tests on each core module and sub-module to confirm that each module is fault-free. Perform parameter calibration on the ultrasonic detection, eddy current detection, and stress wave detection sub-modules of the fusion detection module, setting initial detection parameters to ensure that the detection accuracy of each sub-module meets requirements. Calibrate the communication link of the data transmission module, establishing a bidirectional communication link between the detection and analysis units to ensure smooth data transmission. Initialize the CNN algorithm sub-module of the identification and grading module and the LSTM algorithm sub-module of the trend prediction module, loading the trained algorithm models and clearing historical detection data and analysis results to ensure the algorithm modules can operate normally. Perform initial position calibration on the adjustment bracket, setting the initial adjustment range to fit the minimum pipe diameter of the pipeline to be tested. Calibrate the noise acquisition sensor and filtering sub-module, setting initial filtering parameters to ensure effective filtering of common interference signals.

[0027] Example: In a commercial central air conditioning duct inspection scenario, the pipe diameters to be inspected range from several commonly used specifications. The ambient temperature is high, the relative humidity is high, and multiple air conditioning units and water pumps are operating nearby, which can easily generate vibration and electromagnetic interference. After the system starts up, it first initializes, starts the detection unit and analysis unit, and self-checks to confirm that all modules, including the fusion detection module and anti-interference module, are fault-free. It sets the initial detection frequency for the ultrasonic detection submodule, the initial sensitivity for the eddy current detection submodule, and the initial transmission power for the stress wave detection submodule. It calibrates the Bluetooth and 5G dual-mode communication links of the data transmission module to ensure that the data transmission delay between the detection unit and the analysis unit meets the requirements. It loads the trained CNN defect recognition model and LSTM trend prediction model and clears historical detection data. It adjusts the initial position of the adjusting bracket to fit the smallest pipe diameter. It calibrates the acquisition sensitivity of the noise acquisition sensor and sets the initial filtering threshold of the filtering submodule to ensure that it can filter noise signals generated by the vibration of the main unit and electromagnetic interference.

[0028] Compared to existing technologies, this step employs a multi-module synchronous self-testing and parameter calibration technique. It calibrates not only the detection module but also the algorithm module, communication module, and adjustment bracket. Simultaneously, a bidirectional communication link is established and an optimized algorithm model is loaded, avoiding detection deviations and system failures caused by insufficient initialization in existing technologies. Through this step, the system's post-startup stability is significantly improved, the parameter deviations of each module are greatly reduced, and the stability of the communication link is enhanced. This ensures the accurate execution of subsequent testing work and effectively reduces detection errors and the probability of system downtime caused by improper initialization.

[0029] Step Two: Detection Unit Deployment Steps: The detection unit deployment steps primarily involve installing and fixing the detection unit on the air conditioning duct to be inspected, and adjusting its position. This ensures the detection unit can fully cover the area of ​​the duct to be inspected, adapt to the pipe diameter and installation environment, and provide a stable detection posture for defect data collection. In existing technologies, detection devices are mostly fixed or handheld portable, resulting in cumbersome deployment processes, inability to adapt to pipes of different diameters and installation locations, and unstable detection posture after deployment, easily leading to blind spots or distorted detection data. Therefore, a scientific detection unit deployment process is needed to achieve rapid adaptation and stable installation of the detection unit, ensuring comprehensive detection range and stable detection posture. Specific technical methods are as follows:

[0030] First, a site survey is conducted on the pipeline to be inspected to determine its diameter, direction, installation location, and surrounding environment, and the area to be inspected is marked. Based on the pipeline diameter, the distance between the inspection heads is automatically adjusted by adjusting the telescopic structure of the support, ensuring that the inspection heads of the ultrasonic, eddy current, and stress wave detection sub-modules of the integrated inspection module can fit against the inner wall of the pipeline. Simultaneously, the flexible buffer structure at the end of the support is adjusted to ensure flexible contact between the inspection head and the inner wall of the pipeline, avoiding collision damage. The inspection unit is then fixed to the pipeline, ensuring a secure fixation to prevent displacement due to pipeline vibration during inspection. The inspection angle of the inspection unit is adjusted to ensure that the inspection head can fully cover the area to be inspected, with no blind spots. After deployment, the inspection unit self-test is restarted to confirm that the inspection unit is deployed in place, its inspection posture is stable, and each inspection sub-module can collect signals normally.

[0031] Example: In a commercial central air conditioning duct inspection scenario, the ducts to be inspected include main pipes and branch pipes of different diameters. Some pipes are installed inside the ceiling, surrounded by air conditioning units and pipe supports. A small amount of dust and condensation is present on the pipe surface. First, the on-site ducts are surveyed, marking the areas to be inspected for the main and branch pipes, and determining the pipe diameter. For pipes of different diameters, the automatic extension and retraction of the supports is adjusted to adjust the spacing between the inspection heads, ensuring that the inspection heads of the ultrasonic inspection submodule, eddy current inspection submodule, and stress wave inspection submodule are in close contact with the inner wall of the pipe. The flexible buffer structure maintains tight contact with the inner wall of the pipe, avoiding damage to the oxide layer on the pipe surface. The inspection unit is then fixed to the pipe support with clips, ensuring a secure fixation and preventing displacement caused by vibrations from the air conditioning unit. The inspection angle of the inspection unit is adjusted so that the inspection head can cover all parts of the inner wall of the pipe, avoiding blind spots in areas prone to defects such as welds and bends. After deployment, the inspection unit self-test is initiated to confirm that each inspection submodule can acquire signals normally, with stable inspection posture and no displacement or obstruction issues.

[0032] Compared to existing technologies, this step employs adaptive pipe diameter adjustment and flexible buffer deployment. The adjustment bracket automatically adjusts the spacing between the detection heads according to the pipe diameter, eliminating the need for manual replacement of the detection equipment. Simultaneously, the flexible buffer structure prevents collision damage between the detection head and the pipe's inner wall, solving the problems of existing detection devices being unable to adapt to pipes of different diameters and easily damaging the pipes after deployment. Through the implementation of this step, the deployment efficiency of the detection unit is significantly improved, and its adaptability is greatly expanded, enabling it to adapt to air conditioning pipes of different diameters and installation locations. At the same time, the stability of the detection posture is improved, and the detection blind zone is reduced, ensuring accurate subsequent defect data collection and reducing the risk of secondary damage to the pipes during the detection process.

[0033] Step 3: Defect Data Acquisition and Anti-interference Processing: This is the core detection step. It primarily involves acquiring raw data on defects on the inner wall of the pipeline through various sub-modules of the fusion detection module. Simultaneously, an anti-interference module filters out interference signals from the environment and performs complementary calibration on the acquired raw data, outputting accurate raw defect data to provide high-quality data support for subsequent defect identification and classification. In existing technologies, pipeline defect detection often employs single detection methods, resulting in blind spots and an inability to comprehensively acquire data on different types of defects. Furthermore, interference signals in the detection environment can easily distort the detection data, affecting the accuracy of subsequent defect analysis. Therefore, multi-modal fusion acquisition and adaptive anti-interference processing are needed to achieve comprehensive and accurate defect data acquisition. Specific technical methods are as follows:

[0034] The integrated detection module comprises three sub-modules: ultrasonic testing, eddy current testing, and stress wave testing. These three modules work collaboratively. The ultrasonic testing sub-module collects detection signals for defects such as deep cracks and thinning of the pipe's inner wall. The eddy current testing sub-module collects detection signals for defects such as surface and near-surface corrosion and scratches on the pipe's inner wall. The stress wave testing sub-module collects detection signals for defects such as defects under the insulation layer and micro-weld plugs on the pipe's inner wall. During data acquisition, the noise acquisition sensor in the anti-interference module collects interference signals from the detection environment in real time, including interference signals generated by equipment vibration, electromagnetic interference, temperature and humidity changes, and condensate and oil residue in the pipe. The filtering sub-module filters the collected interference signals. Analysis is performed, and interference signals are filtered out using an adaptive filtering algorithm. Simultaneously, the detection parameters of each submodule in the fusion detection module are automatically adjusted, including the detection frequency of the ultrasonic detection submodule, the sensitivity of the eddy current detection submodule, and the transmission power of the stress wave detection submodule, to ensure the stability of the detection signals. The detection signals acquired by the three submodules are fused and calibrated. Invalid signals are eliminated using a fusion algorithm, and the detection blind spots of each submodule are complemented to output accurate raw defect data, including the defect location, signal strength, and size-related parameters. The processed raw defect data is transmitted in real-time to the analysis unit via the data transmission module and simultaneously stored in the system's local database for subsequent querying and analysis.

[0035] Example: In the scenario of inspecting commercial central air conditioning pipes, the inner wall of the pipe to be inspected may have defects such as slight corrosion, micro-cracks and local weld blockages. The inspection environment may also have mechanical interference caused by the vibration of the air conditioning unit, electromagnetic interference caused by surrounding electrical equipment, and a small amount of condensate and oil residue inside the pipe, which may interfere with the inspection signal. The fusion detection module is activated by three detection sub-modules: the ultrasonic detection sub-module collects signals of deep micro-cracks and wall thinning in the inner wall of the pipe; the eddy current detection sub-module collects signals of slight corrosion and scratches on the inner surface of the pipe; and the stress wave detection sub-module collects signals of micro-defects and local weld plugs under the insulation layer of the inner wall of the pipe. A noise acquisition sensor collects interference signals in real time from host vibration, electromagnetic interference, condensate, and oil contamination. The filtering sub-module filters these interference signals, automatically increasing the detection frequency of the ultrasonic detection sub-module and enhancing the sensitivity of the eddy current detection sub-module to ensure that the detection signals are not interfered with. A fusion algorithm calibrates the signals collected by the three sub-modules, eliminating invalid signals caused by condensate interference and supplementing defect signals under the insulation layer that the ultrasonic detection sub-module cannot collect. It outputs accurate raw data containing parameters related to defect location, signal strength, and size. This raw data is transmitted to the analysis unit via the data transmission module and simultaneously stored in a local database, providing support for subsequent defect identification and classification.

[0036] Compared with existing technologies, this step employs a multimodal fusion acquisition and adaptive anti-interference collaborative processing technique, combining ultrasonic testing, eddy current testing, and stress wave testing to compensate for each other's detection blind spots. Simultaneously, through noise acquisition and adaptive filtering, interference signals are filtered in real time, and detection parameters are dynamically adjusted. This solves the problems of detection blind spots, weak anti-interference capabilities, and data distortion inherent in existing single-detection methods. Through the implementation of this step, the comprehensiveness and accuracy of raw defect data are significantly improved, enabling the comprehensive acquisition of defect data of different types and depths. The impact of interference signals on the detection data is greatly reduced, and invalid signals are decreased, providing high-quality data support for subsequent defect identification and classification. The reliability of defect data is enhanced, avoiding the problems of missed or misjudged defects due to data distortion.

[0037] Step Four: Defect Identification and Grading: This step is the core of intelligent analysis. It primarily utilizes the identification and grading module of the analysis unit to analyze the accurate raw data output from the defect data acquisition and anti-interference processing steps. This process identifies defect types, classifies defect levels, and scores data reliability, providing a basis for subsequent trend prediction and detection strategy optimization. Current pipeline defect detection technologies can only identify the presence of defects, but cannot automatically identify defect types or classify their levels. This requires secondary manual judgment, which is prone to misjudgments and omissions, and cannot verify data reliability, affecting the reliability of subsequent analysis results. Therefore, deep learning algorithms are needed to achieve automatic defect identification, grading, and data reliability verification, improving the intelligence and reliability of the analysis. Specific technical methods are as follows:

[0038] After receiving the raw defect data from the data transmission module, the analysis unit activates the CNN algorithm submodule of the identification and grading module. This submodule extracts and analyzes features from the raw defect data, and, combined with a trained defect sample model, automatically identifies the defect type, including common defects such as cracks, corrosion, weld plugs, porosity, and slag inclusions. After identification, the grading submodule automatically classifies the defect into three levels—minor, moderate, and severe—based on the defect's size parameters, distribution location, and industry standards. Simultaneously, the reliability scoring unit of the identification and grading module scores the raw defect data based on its reliability, providing a score from 0 to 100, according to the data's stability, integrity, and residual interference signals. A higher score indicates higher reliability. The analysis results, including defect type, defect level, and reliability score, are fed back to the detection unit via the data transmission module and stored in the system database, providing a basis for subsequent trend prediction and detection strategy optimization.

[0039] Example: In a commercial central air conditioning duct inspection scenario, the raw data output from the defect data acquisition and anti-interference processing steps contains information related to three types of defects: microcracks at pipe welds, minor corrosion on the inner wall of the pipe, and localized weld plugging. The analysis unit activates the CNN algorithm submodule to extract features from this raw data. Combining this with a trained defect sample model (including cracks, corrosion, and weld plugging), it automatically identifies the three defect types: microcracks, minor corrosion, and localized weld plugging. The grading submodule, based on the width of the microcracks, the area of ​​the minor corrosion, and the degree of the localized weld plugging, and in accordance with industry standards, classifies microcracks and minor corrosion as minor defects and localized weld plugging as moderate defects. The reliability scoring unit scores the raw data for the three defects. Microcracks and minor corrosion detection data have less interference and higher integrity, receiving scores of 96 and 95 respectively. Localized weld plug detection data has a small amount of interference, receiving a score of 92. The analysis results, including defect type, grade, and reliability score, are fed back to the detection unit and stored in the system database, providing support for subsequent trend prediction and detection strategy optimization.

[0040] Compared to existing technologies, this step employs a combination of CNN algorithms and grading standards to achieve automatic defect type identification and grade classification. It also adds a data reliability scoring unit to verify the reliability of the detection data, solving the problems of existing technologies that cannot automatically identify defect types, classify defect grades, require secondary manual judgment, and cannot verify data reliability. Through the implementation of this step, the intelligence level of defect identification and grading is significantly improved, eliminating the need for manual intervention, reducing errors from manual judgment, and enhancing the accuracy of defect identification and grading. The data reliability scoring effectively filters out low-reliability data, providing a reliable basis for subsequent analysis and avoiding analytical biases caused by low-reliability data, thus improving the overall analytical reliability of the system.

[0041] Step 5: Defect Trend Prediction Step: The defect trend prediction step is the core step for achieving early warning. It mainly uses the trend prediction module of the analysis unit, combined with the analysis results of the defect identification and grading steps, pipeline operating parameters, and historical detection data, to predict the development speed and deterioration time of defects, outputting early warning information to provide maintenance guidance for personnel. In existing technologies, pipeline defect detection can only achieve real-time detection and analysis of defects, but cannot predict the development trend of defects. Personnel cannot know the deterioration of defects in advance and can only carry out repairs after the defects have expanded, increasing maintenance costs and safety hazards. Therefore, it is necessary to use time-series prediction algorithms, combined with multi-dimensional data, to achieve accurate prediction and early warning of defect trends. Specific technical means are as follows:

[0042] The trend prediction module receives the analysis results, such as defect type and defect level, output by the identification and grading module. Simultaneously, it accesses pipeline operating parameters stored in the system database, including pipeline operating years, medium type, operating pressure, and operating temperature, as well as historical inspection data, including past defect detection records and defect development status. The module then activates the LSTM algorithm submodule, which performs time-series analysis on the defect level, pipeline operating parameters, and historical inspection data to establish a defect development trend model and predict the defect development speed, including corrosion thinning rate and crack propagation rate. Based on the predicted defect development speed and the current defect level, the module predicts the time of defect deterioration and determines the warning level. The warning submodule outputs defect deterioration warning information, including defect location, current level, predicted deterioration time, and warning level, based on the prediction results and warning level. This warning information is fed back to the detection unit via the data transmission module and simultaneously pushed to the personnel's terminal devices, providing maintenance guidance. The prediction results are also stored in the system database for subsequent tracking and verification.

[0043] Example: In a commercial central air conditioning duct inspection scenario, the defect identification and classification steps identify minor cracks at the pipe welds as minor defects, slight corrosion on the inner wall of the pipe as minor defects, and localized weld blockages as moderate defects. The trend prediction module calls upon the pipeline's operating parameters, including a 5-year operating history, refrigerant as the medium, and standard operating pressure and temperature, as well as historical inspection data. No cracks or corrosion were found during last year's inspection, and the localized weld plugging was a minor defect. The LSTM algorithm submodule performs time-series analysis on this data, establishing a defect development trend model. It predicts that the propagation rate of microcracks is slow, the development rate of minor corrosion is moderate, and the deterioration rate of localized weld plugging is relatively fast. Based on the prediction results, it is predicted that microcracks and minor corrosion will not deteriorate to moderate defects within one year, while localized weld plugging will deteriorate to severe defects within three months. Therefore, the localized weld plugging is classified as a Level 1 warning, and the microcracks and minor corrosion as Level 3 warnings. The warning submodule outputs warning information, specifying the location of the localized weld plug, its current level, predicted deterioration time, and Level 1 warning level, pushing it to the staff's terminal to remind them to perform timely repairs. Simultaneously, the prediction results are stored in the system database for subsequent tracking and verification.

[0044] Compared to existing technologies, this step employs an LSTM time-series prediction algorithm combined with multi-dimensional data. It integrates defect levels, pipeline operating parameters, and historical inspection data to achieve accurate prediction and tiered early warning of defect development trends. This solves the problem that existing technologies cannot predict defect development trends or provide early warnings. Through this step, staff can anticipate the deterioration of defects, promptly carry out maintenance work, and prevent pipeline failures caused by defect expansion, such as refrigerant leaks and compressor burnouts. This reduces maintenance costs and safety hazards. Furthermore, the prediction results provide a basis for optimizing subsequent inspection strategies, enabling on-demand inspection and improving inspection efficiency.

[0045] Step Six: Dynamic Optimization of Detection Strategy: This step is the core of achieving closed-loop collaboration. It primarily uses the reliability scores from the defect identification and grading steps and the prediction results from the defect trend prediction step to feed back to the detection unit. This guides the detection unit to adjust detection parameters and frequency, achieving dynamic optimization of the detection strategy and balancing detection accuracy, efficiency, and energy consumption. In existing technologies, the detection parameters and frequency of the detection system are fixed and cannot be dynamically adjusted according to defect conditions and data reliability, leading to insufficient detection accuracy, low efficiency, or excessive energy consumption. Therefore, closed-loop feedback is needed to dynamically optimize the detection strategy based on analysis results, achieving accurate and on-demand detection. Specific technical methods are as follows:

[0046] The detection unit receives defect identification and grading results, data reliability scores, and defect trend prediction results from the analysis unit. For defect areas with data reliability scores below a set threshold, the detection unit activates the anti-interference module and the fusion detection module, adjusting detection parameters, including increasing the detection frequency of the ultrasonic detection submodule, enhancing the sensitivity of the eddy current detection submodule, and adjusting the transmission power of the stress wave detection submodule. Simultaneously, the detection range of the area is expanded for secondary data acquisition until the data reliability score reaches the set threshold. Based on the defect trend prediction results, for severe defects predicted to deteriorate rapidly and first-level warning defects, the detection frequency and number of tests in the area are increased to ensure timely tracking of defect development. For minor defects predicted to develop slowly and third-level warning defects, the detection frequency and number of tests are appropriately reduced to lower system energy consumption. After adjustments, the detection unit continues detection work according to the optimized detection strategy, while storing the adjusted detection strategy parameters in the system database to form a detection strategy optimization record for future optimization and reference.

[0047] Example: In a commercial central air conditioning duct inspection scenario, the inspection unit receives the results fed back by the analysis unit. The confidence score of local weld plugs is 92, which is lower than the set threshold of 95. Local weld plugs are predicted to be medium defects that deteriorate rapidly, which is a level 1 warning. The confidence scores of microcracks and slight corrosion are 96 and 95, respectively, which are higher than the set thresholds. They are predicted to be slight defects that develop slowly, which is a level 3 warning. For areas with localized weld plugging, the detection unit activates the anti-interference module, improves the filtering accuracy of the filtering submodule, and adjusts the detection parameters of the fusion detection module. This increases the detection frequency of the ultrasonic detection submodule and the transmission power of the stress wave detection submodule, expanding the detection range of the localized weld plugging area. Secondary data acquisition is then performed until the data reliability score reaches 95 or higher. For the first-level warning of localized weld plugging, the detection frequency in that area is increased from once a month to once a week to promptly track the deterioration of the weld plug. For the third-level warning of microcracks and slight corrosion, the detection frequency is reduced from once a month to once every two months to reduce system energy consumption. After adjustments, the detection unit continues to inspect the pipeline according to the optimized detection strategy, storing the adjusted detection parameters and frequency in the system database to form an optimization record.

[0048] Compared to existing technologies, this step employs a closed-loop feedback dynamic optimization approach to the detection strategy. Based on data reliability scores and defect trend predictions, it adjusts detection parameters and frequency in real time, achieving dynamic adaptation of the detection strategy. This solves the problems of fixed detection strategies in existing technologies, which cannot balance detection accuracy, efficiency, and energy consumption. Through this step, the reliability of detection data is further improved, the probability of missed or false defects is significantly reduced, detection efficiency is increased, and system energy consumption is reduced, achieving an optimal balance between detection accuracy, efficiency, and energy consumption. The adaptability of the detection strategy is significantly enhanced, enabling dynamic adjustments based on changes in defect conditions and the detection environment, thus improving the system's practicality and economy.

[0049] Step Seven: Output and Archiving of Inspection Results: This step is the final stage of the entire inspection process. It primarily involves organizing the results related to defect identification and grading, trend prediction, and inspection strategy optimization, outputting a standardized inspection report, and archiving all inspection data and analysis results for easy retrieval, traceability, and analysis later. In existing technologies, inspection results are often simple data records without standardized reports, and data archiving is not standardized. This hinders quick retrieval and traceability of inspection records, which is detrimental to long-term pipeline maintenance and defect analysis. Therefore, standardized report output and standardized archiving are needed to improve the practicality and traceability of inspection results. Specific technical methods are as follows:

[0050] The analysis unit organizes relevant data such as defect identification and grading results, defect trend prediction results, detection strategy optimization results, raw defect data, and anti-interference processing records. Following industry standards and system preset formats, it generates standardized inspection reports. These reports include pipeline information, inspection environment information, inspection time, defect type, defect location, defect level, data reliability score, defect trend prediction results, early warning information, and detection strategy optimization status. The inspection reports are transmitted to staff terminals via a data transmission module, and paper copies are printed for project acceptance, quality control, and maintenance reference. All inspection data and analysis results, including raw defect data, anti-interference processing records, defect identification and grading results, trend prediction results, detection strategy optimization records, and inspection reports, are categorized and archived according to inspection time and pipeline number, stored in both the system's local database and cloud database for dual data backup and security. A data query and traceability system is established, allowing staff to quickly query and trace relevant inspection records using keywords such as pipeline number and inspection time, providing data support for long-term pipeline maintenance and defect analysis.

[0051] Example: In a commercial central air conditioning duct inspection scenario, after the inspection process is completed, the analysis unit organizes all the data from the inspection and generates a standardized inspection report. The report clearly states the duct number, installation location, inspection time, temperature and humidity of the inspection environment, the type, location, and level of the three defects found, the data reliability score, the defect trend prediction results and warning level, and the specific content of the inspection strategy optimization. The inspection report is transmitted to the staff's terminal, and a paper version is printed for the maintenance and management of the central air conditioning duct. The original defect data, anti-interference processing records, defect identification and classification results, trend prediction results, inspection strategy optimization records, and inspection reports are classified according to the inspection time and duct number and stored in the system's local database and cloud database for dual backup. Staff can then quickly query the current and past inspection records of a duct by its duct number to trace the development of defects and the optimization process of inspection strategies, providing data support for long-term duct maintenance and defect analysis.

[0052] Compared to existing technologies, this step employs standardized report output and dual-backup archiving techniques to generate standardized inspection reports containing multi-dimensional information. It also achieves categorized archiving and dual backup of inspection data, establishing a data query and traceability system. This solves the problems of non-standardized inspection reports, disorganized data archiving, and inability to quickly query and trace data in existing technologies. Through the implementation of this step, the practicality of the inspection results is significantly improved. The standardized report can be directly used for project acceptance, quality control, and maintenance reference. The standardized data archiving and convenient query and traceability provide comprehensive data support for long-term pipeline maintenance and defect analysis. Simultaneously, dual data backup ensures data security, prevents data loss, and improves the reliability and usability of the system.

[0053] In summary, this automatic identification and detection system for defects in the inner wall of air conditioning ducts achieves excellent detection results in commercial central air conditioning duct inspection scenarios through the coordinated implementation of the above seven steps. It can comprehensively and accurately collect data on different types of defects in the inner wall of ducts, realize automatic defect identification, classification, trend prediction, and dynamic optimization of detection strategies, and output standardized inspection reports and complete data archiving. It solves the core pain points of existing technologies such as insufficient detection accuracy, low level of intelligence, poor adaptability, and separation of detection and analysis.

[0054] In addition to those mentioned above, the publicly available documents that can be cited in this technical solution include a pipeline inner wall defect detection device and method, and a pipeline defect detection system based on multimodal fusion. However, this method uses a single detection approach, does not involve deep learning technology or a closed-loop collaborative structure, and has weak anti-interference capabilities. Furthermore, it is not deeply integrated with deep learning technology, making it unable to predict defect trends or dynamically optimize detection strategies, resulting in low detection efficiency and a low level of intelligence.

[0055] The core differences between this technical solution and all the aforementioned publicly available documents are as follows: First, it constructs a two-way closed-loop collaborative structure between the detection unit and the analysis unit, realizing full-process automation of detection data acquisition, interference filtering, defect analysis, trend prediction, and detection strategy optimization, breaking the traditional design of separating detection and analysis in existing technologies. Second, it deeply integrates multimodal fusion detection technology with deep learning technology, combining ultrasonic detection, eddy current detection, and stress wave detection to complement each other's detection blind spots. At the same time, it uses CNN and LSTM algorithms to achieve accurate defect identification, classification, and trend prediction, improving the system's intelligence level. Third, it adds an adaptive anti-interference module and an adaptive adjustment bracket, realizing anti-interference under complex working conditions and adaptability to pipes of different diameters, solving the problems of poor adaptability and weak anti-interference capability of existing technologies. Fourth, it realizes dynamic optimization of the detection strategy, adjusting detection parameters and detection frequency according to data reliability and defect trend prediction results, balancing detection accuracy, efficiency, and energy consumption. Fifth, it generates standardized detection reports and realizes dual data backup and archiving, establishing a data query and traceability system, improving the practicality and traceability of detection results.

[0056] The unique technical approach of this solution lies in its use of a closed-loop collaborative structure as the core, integrating multiple technologies such as multimodal fusion detection, deep learning, adaptive anti-interference, and adaptive adjustment. This enables intelligent, accurate, and efficient detection of defects in the inner wall of air conditioning ducts throughout the entire process, solving the technical pain points of existing publicly available technologies that cannot simultaneously achieve accuracy, efficiency, adaptability, and practicality. Its technical concept and implementation methods are different from existing publicly available technologies.

[0057] The detection accuracy of this system is significantly improved compared to existing publicly available technologies. The misjudgment rate of defect identification and classification is greatly reduced, the detection efficiency is improved, energy consumption is reduced, and the scope of application is expanded. It can adapt to the pipeline inspection needs of various scenarios such as commercial central air conditioning and old air conditioning, providing reliable technical support for the long-term maintenance of air conditioning pipelines. It has broad application prospects and practical value.

Claims

1. An air conditioner pipeline inner wall defect automatic identification detection system, characterized in that: The system includes a fusion detection module, an anti-interference module, an adjustment bracket, an identification and grading module, a trend prediction module, and a data transmission module. The fusion detection module, anti-interference module, and adjustment bracket work together to form a detection unit, while the identification and grading module, trend prediction module, and data transmission module work together to form an analysis unit. The detection unit and the analysis unit form a closed-loop collaborative structure through bidirectional linkage via the data transmission module. The fusion detection module includes an ultrasonic detection submodule, an eddy current detection submodule, and a stress wave detection submodule. The anti-interference module includes a noise acquisition sensor and a filtering submodule. The identification and grading module includes a CNN algorithm submodule and a grading submodule. The trend prediction module includes an LSTM algorithm submodule and an early warning submodule. The detection unit collects raw data on defects in the inner wall of the air conditioning duct and filters interference signals. The analysis unit analyzes and processes the raw defect data and outputs analysis results. The analysis unit feeds back the analysis results to the detection unit to guide the detection unit in adjusting detection parameters and detection strategies.

2. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The ultrasonic testing submodule, eddy current testing submodule, and stress wave testing submodule work together to collect detection signals of different types of defects on the inner wall of the pipeline. The collected detection signals are then calibrated using a fusion algorithm to output accurate original defect data.

3. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The noise acquisition sensor collects interference signals in the detection environment in real time. The filtering submodule filters the interference signals and automatically adjusts the detection parameters of the fusion detection module to ensure the stability of the detection signal.

4. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The end of the adjusting bracket is equipped with a flexible buffer structure. The adjusting bracket automatically adjusts the spacing between the detection heads according to the inner diameter of the pipe. The flexible buffer structure prevents the detection heads from colliding with the inner wall of the pipe.

5. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The CNN algorithm submodule analyzes the raw defect data output by the detection unit to identify the defect type, and the grading submodule classifies the defect level according to the defect parameters and industry standards.

6. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The LSTM algorithm submodule combines pipeline operating parameters and historical detection data to predict the rate of defect development, and the early warning submodule outputs early warning information on defect deterioration based on the prediction results.

7. The automatic identification and detection system for defects in the inner wall of air conditioning ducts as described in claim 1, characterized in that: The data transmission module enables bidirectional data transmission between the detection unit and the analysis unit, transmitting the raw defect data output by the detection unit to the analysis unit and feeding back the analysis results output by the analysis unit to the detection unit.

Images