Automatic detection system and method for inner cooling channel type salt core piston

Abstract

The invention relates to the technical field of piston manufacturing and detection, in particular to an automatic detection system and method for an inner cooling channel type salt core piston. The objective of the invention is to solve the problems of low efficiency, unstable detection precision, incapability of completely covering complex inner cooling channel defects and the like caused by dependence on manual operation in an existing inner cooling channel type salt core piston detection process. The system comprises an industrial X-ray detection module, an AI intelligent analysis module, a data acquisition and transmission module, a PLC cooperative control module, a virtual mapping and prediction module and a data management and tracing module, through deep fusion of industrial CT / DR equipment and an AI deep learning algorithm, automation and standardization of defect detection are realized, the detection precision and efficiency are improved, the labor cost is reduced, and the detection efficiency is improved. And meanwhile, full-process tracing and process optimization of detection data are realized, and technical support is provided for large-scale high-quality production of the salt-core piston.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of piston manufacturing and detection, and particularly to an automatic detection system and method for an inner-cooling-channel salt core piston. BACKGROUND

[0002] With the continuous upgrading of engine technology, the heat dissipation performance and structural stability of the piston are increasingly required. The salt core piston with an inner-cooling-channel design can effectively reduce the working temperature of the piston and improve the thermal efficiency of the engine, and its production has been continuously increasing in recent years. The inner-cooling-channel of the salt core piston is formed by placing a salt core pressed from salt into a mold cavity and then forming it through a casting process. The integrity and cleanliness of the channel directly affect the performance and service life of the piston, so defects such as pores, salt residues, lack of meat, step difference, and blockage of the salt core channel need to be strictly detected.

[0003] The existing channel detection of the salt core piston mainly relies on industrial CT or DR (digital radiographic imaging) technology combined with manual analysis. The specific process includes: first, obtaining two-dimensional or three-dimensional images of the internal structure of the piston through multi-angle DR scanning or CT tomography; then, a technician marks the location, type, and size of the defect on the image based on professional experience and attention, such as observing the position deviation of the oil way top through the DR image, or calculating the center deviation of the oil way and the outer circle of the piston on the CT image; finally, comparing the detection results with the design drawings to determine whether the defect exceeds the allowed range.

[0004] However, the existing detection technology has many significant defects: first, it is highly subjective, and the detection results depend on the experience and attention of the technician, and different personnel have different standards for judging the same defect, which can easily lead to missed detection; second, it is inefficient, and manually analyzing images frame by frame takes a long time, which cannot meet the real-time detection needs of large-scale production; third, it is prone to fatigue and misdiagnosis, and long-term observation of X-ray images can cause visual fatigue, significantly increasing the risk of missed detection, and when ultrasonic detection fails for defects that are not perpendicular to the incident direction, it is difficult for humans to fully identify defects through X-ray images; fourth, the precision is limited, and the density resolution of ordinary X-ray is low, making it difficult to detect small pores or trace salt residues with a diameter <0.3mm, and two-dimensional images have organization overlap problems, affecting the positioning accuracy of defects; fifth, data management is difficult, and manually recorded detection data lacks a standardized format, making it difficult to achieve full-process traceability and process optimization, and unable to locate the causes of defects through quantitative analysis. SUMMARY

[0005] In order to solve the technical defects put forward in the background art, the purpose of the present application is to provide an automatic detection system and method for an inner cooling channel type salt core piston, which effectively solves the problems of low efficiency, unstable detection accuracy and inability to fully cover complex inner cooling channel defects caused by relying on manual operation in the detection process of the existing inner cooling channel type salt core piston.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions: An automatic detection system for an inner cooling channel type salt core piston, comprising: An industrial X-ray detection module for multi-angle X-ray scanning of the piston salt core channel to collect real-time imaging two-dimensional or three-dimensional internal structure images; An AI intelligent analysis module in communication connection with the industrial X-ray detection module, comprising an AI host, a detection software and a display, the AI host being used for running the detection software and performing deep learning inference analysis on the collected images to identify defect types, positions and key parameters, and the display being used for displaying the determination results; A data acquisition and transmission module connected with the industrial X-ray detection module and the AI intelligent analysis module respectively, for transmitting the image data collected by the industrial X-ray detection module to the AI intelligent analysis module and a background data system; A PLC cooperative control module connected with the AI intelligent analysis module and communicating with a production line PLC and a mechanical arm actuator through a communication protocol, for controlling the mechanical arm actuator according to the determination results of the AI intelligent analysis module; A virtual mapping and prediction module connected with the data acquisition and transmission module and the data management module, for constructing a virtual mapping model of the production line and predicting defect risks based on historical data; A data management and traceability module for storing detection data, defect information, process parameters and determination results, and supporting data traceability and process analysis.

[0007] Preferably, the industrial X-ray detection module adopts industrial CT or DR digital radiographic imaging technology, and the image density resolution is not less than 0.1mm.

[0008] Preferably, the detection software of the AI intelligent analysis module supports online model updating and replacement function.

[0009] Preferably, the data acquisition and transmission module adopts a high-speed Ethernet interface, and its data transmission delay is ≤100ms.

[0010] Preferably, the virtual mapping and prediction module adopts digital twin technology to construct a virtual production line model, and analyzes the correlation between historical defect data and process parameters through gradient boosting tree algorithm for prediction.

[0011] Preferably, the data management and traceability module supports data query and statistical analysis in at least one dimension of piston batch, detection time, and defect type.

[0012] An automatic detection method for an inner cooling channel type salt core piston, applied to the automatic detection system of any one of claims 1-6, the method comprising the following steps: S1, system building: deploying an industrial X-ray detection module and building a detection environment meeting the imaging requirements, configuring the hardware devices and detection software of the AI intelligent analysis module, building a data acquisition and transmission module, a PLC collaborative control module, a virtual mapping and prediction module, and a data management and traceability module, and completing the communication protocol matching and connection debugging between the modules; S2, data acquisition and labeling: randomly select piston samples in different production batches according to a preset proportion, and collect multi-angle X-ray images of each sample through the industrial X-ray detection module; at the same time, label all collected X-ray images to clearly mark the position, shape, type and size parameters of defects to form a labeled training data set; S3, AI model training and optimization: perform data enhancement processing on the labeled training data set, and train the selected AI model based on the enhanced data set until the model detection accuracy meets the preset standard; S4, real-time automatic detection and feedback: when the piston salt core is conveyed to the detection station through the production line, the industrial X-ray detection module automatically starts the scanning program, quickly collects the X-ray images of the salt core channel, and transmits the image data to the AI intelligent analysis module in real time through the data acquisition and transmission module; the AI intelligent analysis module calls the trained and optimized model to quickly infer and analyze the images, identifies whether there is a defect and the specific type, position and parameters of the defect; the system automatically judges the product qualified state according to the preset defect judgment standard, outputs the detection result through the display and indicator light, and uploads the complete detection data to the data management and traceability module.

[0013] Preferably, in step S3, the enhancement processing includes image rotation, scaling, cropping, and adding Gaussian noise operations.

[0014] Preferably, in step S3, the AI model adopts a convolutional neural network architecture, the learning rate is set to 0.001-0.01 during the training process, the number of iterations is 100-500 times, and the defect recognition accuracy after the model training is completed is not less than 99.5%.

[0015] Preferably, in step S4, after receiving the unqualified product judgment signal, the PLC cooperative control module automatically rejects through the mechanical arm actuator; at the same time, the virtual mapping and prediction module synchronously updates the virtual model data, combines real-time detection data and historical data to predict potential defect risks, and feeds back the prediction results to the production system.

[0016] In summary, the beneficial effects of the present application are: The present application effectively solves the problems of low efficiency, strong subjectivity and difficulty in meeting the needs of large-scale industrial production of traditional detection methods by organically combining the industrial X-ray detection module, the AI intelligent analysis module, the data acquisition and transmission module, the PLC cooperative control module, the virtual mapping and prediction module and the data management and traceability module, significantly improves the detection accuracy and production efficiency of the piston inner cooling channel, reduces human error, guarantees the stability and consistency of product quality, and provides strong support for the continuous improvement of production process through the correlation analysis and optimization suggestions of process parameters, has important engineering application value and economic benefits, and can be widely applied to the intelligent manufacturing field of key parts such as automobile engine pistons. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is the overall layout of the automatic detection system for the inner cooling channel type salt core piston of the present application; Figure 2 is the work flow chart of the automatic detection method for the inner cooling channel type salt core piston of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the present application.

[0019] Those skilled in the art should understand that in the disclosure of the present application, the terms "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, which are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the above terms cannot be understood as a limitation of the present application.

[0020] In the description of the present application, if the word "several" or the like is described, it means one or more, the meaning of multiple is two and more, greater than, less than, more than, etc. Understand as not including the number, above, below, within, etc. Understand as including the number. If the first, second, third is described, it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the sequence of indicated technical features.

[0021] The following will be described in detail in combination with the accompanying drawings Figures 1-2 Further detailed description will be made on an embodiment of the present application, an automatic detection system and method for an inner cooling channel type salt core piston.

[0022] An automatic detection system for an inner cooling channel type salt core piston, as shown in Figure 1 , 2 It includes an industrial X-ray detection module, an AI intelligent analysis module, a data acquisition and transmission module, a PLC collaborative control module, a virtual mapping and prediction module, and a data management and traceability module.

[0023] The industrial X-ray detection module is used for multi-angle X-ray scanning of the piston salt core channel, and collects real-time imaging two-dimensional or three-dimensional internal structure images; and the industrial X-ray detection module adopts industrial CT or DR digital radiographic imaging technology, and the image density resolution is not less than 0.1mm.

[0024] Further, the module is equipped with a double-source double-detector structure, wherein the main detector adopts a 16-bit high-sensitivity cesium iodide scintillator panel with a pixel size of 127μm, which can realize a spatial resolution of 0.05mm and ensure clear imaging of micro pores; the auxiliary detector adopts a low-dose fast scanning mode, cooperates with a C-arm scanning mechanism driven by a high-precision mechanical arm, and can complete 360° omnidirectional non-dead-angle scanning of the piston within 30 seconds to obtain a 1024×1024 pixel high-definition DR sequence image or a 0.1mm layer-thickness CT tomographic data. In order to guarantee the imaging quality, the module is built-in with an adaptive exposure control algorithm, which can automatically adjust the tube voltage and tube current according to the piston material and salt core density, reduce the X-ray dose, and improve the image signal-to-noise ratio to ≥50dB, effectively distinguishing the gray difference between salt residues and pores.

[0025] In this embodiment, the AI intelligent analysis module is in communication connection with the industrial X-ray detection module, including an AI host, a detection software and a display, the AI host is used for running the detection software and performing deep learning inference analysis on the collected images to identify the defect type, position and key parameters, and the display is used for displaying the judgment result.

[0026] Specifically, the detection software of the AI intelligent analysis module supports online model updating and replacement function. The AI host loaded with the detection software adopts 2 NVIDIA A100 GPU parallel computing architecture, is equipped with 1 TB cache and 256 GB DDR5 memory, and can process 8 X-ray image data streams simultaneously. The detection software kernel integrates multi-task deep learning model, and the bottom layer adopts improved U-Net++ architecture as the backbone network. The feature extraction capability for micro defects is enhanced by introducing attention mechanism and residual connection. For different defect types such as pores and salt residues, multi-scale feature fusion branches are constructed. The pore detection branch adopts double screening of circularity constraint and area threshold, and the salt residue identification branch analyzes the gray distribution characteristics through the spectral clustering algorithm to realize accurate identification of 0.02 mm³ micro salt residues. The software built-in defect labeling tool can automatically generate a structured detection report containing defect coordinates, size, type and confidence, and supports design drawing import comparison to automatically calculate the coaxiality deviation of oil channel center and piston outer circle and the position error of oil channel top.

[0027] In addition, it is worth noting that the software has a model management function, which can push an update package through a remote server to realize online upgrading and replacement of algorithm models. The upgrading process does not interrupt the detection process, ensuring production continuity.

[0028] In the embodiment, the data acquisition and transmission module is connected with the industrial X-ray detection module and the AI intelligent analysis module respectively, and is used to transmit the image data collected by the industrial X-ray detection module to the AI intelligent analysis module and the background data system.

[0029] Specifically, the module adopts a high-speed Ethernet interface based on TCP / IP protocol, and integrates an edge computing preprocessing unit. Lossless compression and format conversion (conversion of original DICOM data to JPEG2000 standard format) are performed before image data transmission, ensuring that the data transmission delay is ≤50 ms and the bandwidth occupation is stable within 800 Mbps. To realize full-quantity data acquisition, the module is built-in with an industrial sensor data interface, which can synchronously collect real-time process parameters (such as tube voltage, tube current, and scanning time) of the X-ray detection module, mechanical arm positioning coordinates, and environmental temperature and humidity data, and bind them with image data through time stamp. The data transmission adopts an encrypted transmission protocol to prevent data leakage or tampering, and supports a breakpoint resume mechanism. When the network is interrupted and restored, the lost data can be automatically transmitted to ensure 100% data integrity. In addition, the module is equipped with a local cache unit, which can temporarily store detection data when the background data system fails, and the cache capacity supports local backup of 8 hours of full-load detection data.

[0030] In the embodiment, the PLC cooperative control module is connected with the AI intelligent analysis module, and communicates with the production line PLC and the mechanical arm actuator through a communication protocol, to control the mechanical arm actuator according to a determination result of the AI intelligent analysis module.

[0031] Specifically, the PLC cooperative control module adopts a Siemens S7-1500 series PLC as a main controller, configures a PROFINET industrial Ethernet interface to realize real-time communication with the production line PLC and the mechanical arm actuator, and adopts an OPCUA standard as a communication protocol, with a data update period ≤20 ms. The module receives a determination result signal (including a qualified / unqualified flag, a defect type code and a confidence value) sent by the AI intelligent analysis module through a Modbus TCP / IP protocol, and performs logical judgment in combination with a piston in-place signal and a station state signal sent by the production line PLC. When the determination result is unqualified, the PLC cooperative control module immediately sends a rejection instruction to the mechanical arm actuator, and the instruction contains a coordinate position of unqualified products on a tray (an X / Y / Z offset value fed back based on a vision positioning system) and a grasping posture parameter. To ensure action accuracy, the module is built-in with a motion trajectory planning algorithm, which can automatically adjust a motion path of the mechanical arm according to a piston model, to avoid collision with surrounding equipment, with a single rejection action period ≤3 seconds. Meanwhile, the module has a fault diagnosis and alarm function, can monitor a mechanical arm execution state in real time, and when abnormality such as material jamming or signal loss occurs, immediately triggers an audible and light alarm, sends a shutdown request to the production line PLC, and uploads a fault code to the data management and traceability module. In addition, the PLC cooperative control module supports real-time data interaction with the virtual mapping and prediction module, receives a defect risk early warning signal sent by the virtual mapping and prediction module, and according to a preset rule, adjusts scanning parameters of a detection station in advance or notifies an upstream process to make process parameter fine adjustment (such as adjusting a salt core pressing pressure or pouring temperature).

[0032] In the embodiment, the virtual mapping and prediction module is connected with the data acquisition and transmission module and the data management module, to build a virtual mapping model of the production line, and predict a defect risk based on historical data; the data management and traceability module is used to store detection data, defect information, process parameters and determination results, and supports data traceability and process analysis.

[0033] Specifically, the virtual mapping and prediction module adopts digital twinning technology to construct a full-factor virtual model of the production line, which is composed of a physical layer data interface, a three-dimensional modeling engine, and a prediction algorithm library. The physical layer data interface collects real-time operation parameters, equipment status, and environmental data of each station on the production line through OPCUA protocol, and synchronizes the collected data to the virtual model at a cycle of 10 ms. The three-dimensional modeling engine is developed based on the Unity3D platform, and restores the physical layout of the detection station in a 1:1 scale, including the C-arm motion trajectory of the industrial X-ray detection module, the mechanical arm grabbing path, and the delivery posture of the piston salt core. The model rendering precision reaches 0.01 mm, and can dynamically demonstrate each step of the detection process.

[0034] The prediction algorithm library integrates LSTM time series prediction model and random forest classification algorithm, and establishes a defect risk prediction model by analyzing the correlation between defect types and process parameters in historical detection data. When the same type of defect (such as salt core blockage) occurs for three consecutive batches and the occurrence rate exceeds 0.5%, the system automatically triggers an early warning, predicts the number and impact range of possible defects within the next two hours, and generates process adjustment suggestions.

[0035] The data management and traceability module is built based on a distributed database architecture, and adopts a three-level storage strategy: real-time detection data is stored in a high-performance SSD array, historical data is automatically migrated to a NAS storage system, and archived data is transferred to a tape library for long-term storage. The module develops a special data query engine, supporting multi-dimensional retrieval: single product complete detection report (including X-ray original image, defect annotation map, and AI analysis heat map) can be queried by piston batch number, defect type occurrence rate trend in the past month can be calculated, or the influence of different operator equipment adjustment parameters on detection results can be analyzed by production shift. To meet the quality traceability requirements, the module uses blockchain technology to store key detection data, generates a unique hash value and timestamp for each detection record, ensuring data cannot be tampered with; at the same time, it supports integration with ERP system, automatically synchronizes production order information, and realizes full life cycle data traceability from raw material storage to finished product delivery. In addition, the module has built-in data visualization tools that can generate defect distribution heat map, equipment operation OEE report, and AI model accuracy trend curve, providing data support for process optimization.

[0036] An automatic detection method for an inner cooling channel type salt core piston, comprising the following steps: S1, system construction: deploy an industrial X-ray detection module and build a detection environment that meets the imaging requirements, configure hardware devices and detection software for the AI intelligent analysis module, build data acquisition and transmission module, PLC collaborative control module, virtual mapping and prediction module, and data management and traceability module, complete communication protocol matching and connection debugging between modules; S2, data collection and labeling: randomly select piston samples in different production batches according to the preset proportion, collect multi-angle X-ray images of each sample through high-precision industrial X-ray detection module, and comprehensively cover key parts of the salt core inner cooling channel; collect piston samples of various defects such as gas holes, salt residues, meat defects, segment differences and blockages, collect X-ray images of defect parts clearly; a professional labeling team uses an image labeling tool to label all collected X-ray images, clearly mark the location, shape, type and size parameters of the defects, and form a labeled training data set; S3, AI model training and optimization: the labeled training data set is subjected to data enhancement processing, including image rotation, scaling, cropping and adding Gaussian noise operation, expanding the data set size and diversity; based on the enhanced data set, the selected AI model is trained, the training parameters such as learning rate and iteration number are set, the model weight parameters are adjusted in real time through the loss function, the cross-validation method is used to evaluate the model detection performance, the training strategy is dynamically adjusted to avoid overfitting and underfitting problems, until the model detection accuracy meets the preset standard; S4, real-time automatic detection and feedback: when the piston salt core is transported to the detection station through the production line, the industrial X-ray detection module automatically starts the scanning program, quickly collects the X-ray images of the salt core channel, and transmits the image data to the AI intelligent analysis module in real time through the data collection and transmission module; the AI intelligent analysis module calls the trained and optimized model to quickly infer and analyze the image, identifies whether there is a defect and the specific type, location and parameters of the defect; the system automatically judges the product qualified state according to the preset defect judgment standard, outputs the detection result through the display and indicator light, and uploads the complete detection data to the data management and traceability module; after receiving the unqualified product judgment signal, the PLC cooperative control module immediately triggers the mechanical arm to perform the rejection action, while the virtual mapping and prediction module synchronously updates the virtual model data, combines the real-time detection data and historical data to predict the potential defect risk, and feeds back the prediction result to the production system.

[0037] Further, in step S3, the AI model adopts a convolutional neural network architecture, the learning rate is set to 0.001-0.01 during the training process, the iteration number is 100-500 times, and the defect recognition accuracy of the model after training is not less than 99.5%.

[0038] And in step S4, to further improve the intelligent level and detection efficiency of the detection system, the system also has the function of dynamic detection parameter self-adaptive adjustment. When the virtual mapping and prediction module predicts that the occurrence rate of a certain type of defect has an upward trend according to real-time detection data and historical trend analysis, it will immediately send a parameter adjustment suggestion to the industrial X-ray detection module. After receiving the suggestion, the industrial X-ray detection module makes fine adjustments to the scanning parameters of the next piston to be detected without interrupting the current detection process. For example, if the risk of porosity defects increases, the tube voltage is appropriately increased to 150kV, the tube current is increased to 8mA, and the layer thickness is adjusted to 0.08mm to enhance the penetration and imaging clarity of small pores. At the same time, the AI intelligent analysis module also receives the risk warning information synchronously, automatically adjusts the detection branch model of this type of defect to high sensitivity mode, temporarily relaxes the confidence threshold to 90%, and ensures that potential risk defects are not missed. After continuously detecting 50 pistons and the occurrence rate of this type of defect returns to normal, the system automatically adjusts all parameters back to standard values, achieving dynamic optimization and closed-loop control of the detection process.

[0039] In addition, during real-time automatic detection, the data management and traceability module stores the detection data of each piston in real time, including original X-ray images, AI analysis results, process parameters, equipment status and environmental data, etc. When quality problems are found in the client or downstream process, the complete detection file of the product can be quickly retrieved in the data management and traceability module through the unique two-dimensional code or batch number on the piston, including X-ray images, defect annotation maps, AI model version, key parameters such as tube voltage / tube current at the time, and even the virtual scene of the production line at the time can be traced back through the virtual mapping and prediction module to accurately restore the detection process, providing comprehensive data support for root cause analysis of quality problems.

[0040] For the case where the same type of defect occurs in three consecutive batches and the occurrence rate exceeds 0.5%, the data management and traceability module will automatically trigger the process analysis report generation process, correlate the defect data with the process parameters of the upstream process (such as salt core making, pouring, die casting) (such as salt core moisture content, pouring speed, mold temperature) through the built-in data analysis engine, generate a process improvement report containing defect trend chart, key parameter influence degree sorting and optimization suggestions, and push it to the production management system and related process engineers terminal, realizing whole-chain quality control from detection to process optimization.

[0041] In summary, the automatic detection system and method for the inner cooling channel type salt core piston provided by the present application realizes high-precision, high-efficiency, and intelligent detection and quality control of the piston salt core inner cooling channel defects through the organic combination of the industrial X-ray detection module, the AI intelligent analysis module, the data acquisition and transmission module, the PLC collaborative control module, the virtual mapping and prediction module, and the data management and traceability module. The system not only has excellent defect recognition and data analysis capabilities, can generate detailed structured detection reports and accurately calculate geometric size deviations, but also ensures fast, safe, and complete data transmission through the high-speed Ethernet interface, edge computing preprocessing, and encrypted transmission protocol.

[0042] The system and method effectively solve the problems of low efficiency, strong subjectivity, and difficulty in meeting large-scale industrial production requirements of traditional detection methods, significantly improve the detection accuracy and production efficiency of the piston inner cooling channel, reduce human error, guarantee the stability and consistency of product quality, and provide strong support for continuous improvement of the production process through correlation analysis and optimization suggestions of process parameters, have important engineering application value and economic benefits, and can be widely applied to the intelligent manufacturing field of key components such as automobile engine pistons.

[0043] The embodiments of the specific implementation are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, wherein the same components are denoted by the same reference numerals. Therefore, equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. An automatic detection system for internal cooling channel type salt core pistons, characterized in that, include: The industrial X-ray inspection module is used to perform multi-angle X-ray scanning of the piston salt core channel and acquire real-time two-dimensional or three-dimensional internal structure images. The AI ​​intelligent analysis module is communicatively connected to the industrial X-ray inspection module and includes an AI host, inspection software, and a display. The AI ​​host is used to run the inspection software and perform deep learning inference analysis on the acquired images to identify the defect type, location, and key parameters. The display is used to show the judgment results. The data acquisition and transmission module is connected to the industrial X-ray detection module and the AI ​​intelligent analysis module respectively, and is used to transmit the image data acquired by the industrial X-ray detection module to the AI ​​intelligent analysis module and the background data system. The PLC collaborative control module is connected to the AI ​​intelligent analysis module and communicates with the production line PLC and the robotic arm actuator through a communication protocol. It is used to control the robotic arm actuator according to the judgment result of the AI ​​intelligent analysis module. The virtual mapping and prediction module, connected to the data acquisition and transmission module and the data management module, is used to construct a virtual mapping model of the production line and predict defect risks based on historical data. The data management and traceability module is used to store inspection data, defect information, process parameters and judgment results, and supports data traceability and process analysis.

2. The automatic detection system for internal cooling channel type salt core pistons according to claim 1, characterized in that, The industrial X-ray inspection module uses industrial CT or DR digital X-ray imaging technology, and its image density resolution is not less than 0.1 mm.

3. The automatic detection system for internal cooling channel type salt core pistons according to claim 1, characterized in that, The detection software of the AI ​​intelligent analysis module supports online model updates and replacements.

4. The automatic detection system for internal cooling channel type salt core pistons according to claim 1, characterized in that, The data acquisition and transmission module uses a high-speed Ethernet interface, and its data transmission delay is ≤100ms.

5. The automatic detection system for internal cooling channel type salt core pistons according to claim 1, characterized in that, The virtual mapping and prediction module uses digital twin technology to construct a virtual production line model and uses the gradient boosting tree algorithm to analyze the correlation between historical defect data and process parameters for prediction.

6. The automatic detection system for internal cooling channel type salt core piston according to claim 1, characterized in that, The data management and traceability module supports data querying and statistical analysis based on at least one dimension: piston batch, inspection time, and defect type.

7. An automatic detection method for salt core pistons with internal cooling channels, characterized in that, Applied to the automatic detection system as described in any one of claims 1-6, the method comprises the following steps: S1. System Setup: Deploy the industrial X-ray inspection module and build an inspection environment that meets imaging requirements; configure the hardware and inspection software of the AI ​​intelligent analysis module; build the data acquisition and transmission module, PLC collaborative control module, virtual mapping and prediction module, and data management and traceability module; and complete the communication protocol matching and connection debugging between the modules. S2. Data Acquisition and Labeling: Piston samples are randomly selected from different production batches according to a preset ratio, and multi-angle X-ray images of each sample are acquired through an industrial X-ray inspection module; at the same time, all acquired X-ray images are labeled to clearly mark the location, shape, type and size parameters of defects, so as to form a labeled training dataset. S3. AI Model Training and Optimization: Perform data augmentation on the labeled training dataset, and train the selected AI model based on the augmented dataset until the model detection accuracy meets the preset standard. S4. Real-time Automatic Detection and Feedback: When the piston salt core is transported to the inspection station through the production line, the industrial X-ray inspection module automatically starts the scanning program, quickly acquiring X-ray images of the salt core channel. The image data is transmitted in real time to the AI ​​intelligent analysis module through the data acquisition and transmission module. The AI ​​intelligent analysis module calls the trained and optimized model to perform rapid reasoning analysis on the image, identifying whether there are defects and the specific type, location, and parameters of the defects. The system automatically judges the product's qualification status according to the preset defect judgment criteria, outputs the inspection results through the display and indicator lights, and uploads the complete inspection data to the data management and traceability module.

8. The automatic detection method for internal cooling channel type salt core piston according to claim 7, characterized in that, In step S3, the enhancement process includes image rotation, scaling, cropping, and adding Gaussian noise.

9. The automatic detection method for an internal cooling channel type salt core piston according to claim 7, characterized in that, In step S3, the AI ​​model adopts a convolutional neural network architecture. During training, the learning rate is set to 0.001-0.01, the number of iterations is 100-500, and the defect recognition accuracy after the model is trained is not less than 99.5%.

10. The automatic detection method for an internal cooling channel type salt core piston according to claim 7, characterized in that, In step S4, after receiving the non-conforming product judgment signal, the PLC collaborative control module will automatically remove the non-conforming product through the robotic arm execution mechanism; at the same time, the virtual mapping and prediction module will update the virtual model data synchronously, combine real-time detection data and historical data to predict potential defect risks, and feed the prediction results back to the production system.

Images