Intelligent bonding copper wire equipment control method and system based on Internet of Things

Abstract

The invention discloses an intelligent bonding copper wire equipment control method and system based on the Internet of Things, and relates to the technical field of semiconductor manufacturing. An intelligent bonding copper wire equipment control system based on the Internet of Things comprises copper wire bonding equipment, an edge computing gateway, a cloud platform and a communication module. According to the invention, a three-level cooperative intelligent control system combining real-time control of an equipment end, intelligent prediction of an edge layer and dynamic adjustment of parameters, and long-term optimization of a cloud end and remote intervention of experts is innovatively constructed, and a hierarchical uploading and distributed cooperative control strategy of real-time data and historical data is combined; data integrity, real-time performance, intelligence and system expandability are effectively considered, and an intelligent copper wire bonding control overall solution for complex scenes is formed.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor manufacturing technology, and in particular to an Internet of Things-based intelligent copper bonding wire equipment control method and system. Background Art

[0002] Copper wire bonding technology is a key process in semiconductor packaging and microelectronics assembly, widely used for electrical connections between internal electrodes and lead frames within chips. With the continuous improvement of chip integration and packaging density, the stability of the copper wire bonding process, solder joint quality, and equipment operating status are placing higher demands on yield and product reliability. Traditional bonding equipment primarily relies on preset process parameters and single-point sensor feedback for process control. This makes it difficult to perceive the multi-factor coupling effects of the welding process in real time, resulting in large fluctuations in welding quality. Existing control methods are particularly difficult to quickly adapt to changes in materials, equipment wear, or the environment, further impacting welding stability and product consistency.

[0003] At present, some advanced copper wire bonding equipment is equipped with sensors for key parameters such as pressure and ultrasonic energy, and is combined with PLC or dedicated controllers to achieve basic closed-loop control. Some high-end equipment has added visual inspection functions for solder joints for offline quality assessment after welding. In terms of equipment networking, existing IoT platforms mostly focus on functions such as remote equipment monitoring, production data statistics, and fault alarms. A small number of companies have attempted to upload welding process data to the cloud and perform simple trend analysis. However, existing technologies generally have problems such as a single data source, insufficient real-time feedback capabilities, lack of multimodal information fusion analysis, and disconnection between cloud analysis and equipment control. The overall level of intelligence is low, making it difficult to meet the needs of high-precision and high-reliability welding processes. Summary of the Invention

[0004] A method for controlling an intelligent copper bonding wire device based on the Internet of Things, comprising:

[0005] Pressure, temperature, ultrasonic power, current, vibration and microscopic image acquisition devices are installed on the copper wire bonding equipment to synchronously collect multi-physical quantity data and solder joint images during the welding process;

[0006] Receive multi-physical quantity data through the edge computing gateway and perform time alignment, filtering and denoising, feature extraction, and abnormal data removal;

[0007] The pre-trained welding quality prediction model is called in the edge computing gateway, taking multi-physical sensor data, process parameters, and solder joint image features as input, and outputting solder joint quality scores and defect categories.

[0008] Adjust process parameters in real time based on the solder joint quality score and defect category, and send adjustment instructions to the equipment controller via the edge gateway;

[0009] Upload welding process data and welding results to the cloud platform, which builds a long-term optimization model based on big data analysis, historical process parameters and expert rules, and periodically updates it to the edge device;

[0010] Combining real-time closed-loop control, edge intelligent prediction and long-term cloud optimization, a three-level collaborative intelligent control system is formed, which dynamically adapts to changes in materials, equipment status and environment, and improves welding stability and finished product quality.

[0011] As a preferred technical solution of the present invention, the sensor includes:

[0012] Pressure sensors, ultrasonic energy sensors, temperature sensors, current sensors and vibration sensors installed on the bonding head or bonding arm;

[0013] A microscopic imaging device is arranged in the welding area, which obtains the microscopic morphology of the welding point and forms a multimodal data source by combining multi-physical quantity data.

[0014] As a preferred technical solution of the present invention, the preprocessing of the edge computing gateway includes:

[0015] Synchronize and align multi-physical quantity data based on timestamps;

[0016] De-noising of pressure, ultrasonic energy and current data by Kalman filtering or wavelet transform;

[0017] Perform edge enhancement, contrast enhancement, and key area segmentation of solder joints on microscopic images;

[0018] Combine data integrity detection with sensor health assessment to eliminate or compensate for abnormal and missing data.

[0019] As a preferred technical solution of the present invention, the welding quality prediction model includes:

[0020] CNN image feature extraction network to obtain solder joint geometry, surface roughness and fusion area features;

[0021] LSTM time series analysis network, which analyzes the temporal trends and mutation characteristics of multi-physical quantity data;

[0022] The feature fusion layer combines the output features of CNN and LSTM, and integrates the process parameters to form the solder joint quality score and defect classification results.

[0023] As a preferred technical solution of the present invention, after receiving the defect category and quality score output by the prediction model, the edge computing gateway automatically calls the preset parameter adjustment strategy according to the defect type and risk level, and adjusts the pressure, ultrasonic energy or welding time in real time; wherein, the parameter adjustment rules are dynamically optimized in combination with the historical welding parameter library, the expert process rule library and the adaptive compensation model.

[0024] As a preferred technical solution of the present invention, the cloud platform periodically receives welding process data, weld quality scores, and defect classification results uploaded by the edge computing gateway, and builds a long-term parameter optimization model based on big data analysis, expert knowledge, and reinforcement learning algorithms;

[0025] The optimization model combines the equipment operating status, material batch differences and environmental parameter changes to form a dynamic parameter recommendation strategy for the entire life cycle, and regularly updates it to the edge computing gateway through the parameter delivery interface.

[0026] As a preferred technical solution of the present invention, the method further includes equipment health prediction and fault warning steps, specifically:

[0027] Real-time monitoring of equipment operation vibration characteristics, ultrasonic system power curves, and temperature rise changes of key components through the edge computing gateway;

[0028] Combining historical fault records and expert rules, we build equipment health score and remaining life prediction models;

[0029] The prediction results are uploaded to the cloud platform in real time and combined with expert remote diagnosis opinions to form a closed loop of equipment health management.

[0030] As a preferred technical solution of the present invention, the edge computing gateway and the device controller realize high real-time data interaction through EtherCAT or TSN protocol;

[0031] The edge computing gateway and the cloud platform transmit historical data through the MQTT protocol, or transmit real-time status data through the OPC-UA protocol, realizing hierarchical uploading and distributed collaborative control of historical data and real-time data.

[0032] As a preferred technical solution of the present invention, the method supports remote expert intervention and parameter optimization collaborative mode. Experts remotely access the equipment operating status, real-time quality scores and defect classification results through the cloud platform, and combine the historical data of the welding process with expert rules to remotely issue parameter fine-tuning instructions or temporary control strategies to achieve expert-equipment collaborative decision-making and emergency response.

[0033] An intelligent copper wire bonding equipment control system based on the Internet of Things, comprising:

[0034] Copper wire bonding equipment, equipped with multiple sensors and microscopic imaging modules, is used to collect multi-physical quantity signals and solder joint images during the welding process in real time;

[0035] Edge computing gateways receive device sensor data, perform data preprocessing, AI reasoning analysis, and dynamic adjustment of process parameters, and issue control instructions to device controllers.

[0036] The cloud platform receives data uploaded by devices and edge gateways, performs remote monitoring, big data analysis, long-term optimization of process parameters and equipment health prediction, and periodically issues optimization strategies;

[0037] Communication module: devices communicate with edge gateways via EtherCAT or TSN, and edge gateways communicate with cloud platforms via MQTT or OPC-UA, enabling hierarchical data transmission and remote collaborative control.

[0038] The system forms a three-level collaborative intelligent control system that combines real-time control on the device side, intelligent prediction and adaptive parameter adjustment on the edge, and long-term parameter optimization and equipment health prediction on the cloud side.

[0039] The present invention has the following advantages:

[0040] The present invention simultaneously deploys pressure, temperature, ultrasonic power, current, and vibration sensors, and combines them with a microscopic imaging module to synchronously collect dynamic physical quantities and weld micromorphology data during the welding process, forming a multimodal collaborative perception system, which significantly improves the comprehensive perception of the welding status and the ability to identify abnormalities.

[0041] The present invention presets a solder joint quality prediction model in the edge computing gateway. The model integrates the time series characteristics of multiple physical quantities, microscopic image characteristics and real-time process parameters to realize real-time scoring and defect classification of solder joint quality. The prediction results directly drive the edge gateway to dynamically adjust key process parameters, realizing a prediction-driven adaptive control closed loop. Compared with traditional fixed parameters or simple feedback control, the response is more timely and the control is more precise. The edge gateway and the device controller are directly linked through a high-real-time industrial protocol, combining real-time prediction results with dynamic parameter compensation strategies to form an edge-device real-time closed-loop control chain, minimizing external network transmission delays, achieving millisecond-level control response, and ensuring real-time and stable control of the welding process.

[0042] The present invention constructs a long-term parameter optimization model on the cloud platform that combines historical data analysis, expert knowledge rules and reinforcement learning algorithms. The model forms a multi-scenario adaptive parameter recommendation strategy based on different material batches, equipment operating status, environmental conditions and historical process effects, and periodically sends it to the edge gateway to ensure that equipment parameters are dynamically optimized and self-evolved with production conditions.

[0043] The present invention further integrates equipment operation data, key component health status, historical fault records and expert knowledge base to construct an equipment health score and remaining life prediction model, realizing three-dimensional linkage monitoring and predictive maintenance of welding quality, equipment health and process parameters. The cloud platform also supports remote experts to view equipment status, historical data and current defect classification results in real time, and remotely issue parameter fine-tuning or emergency strategies to realize expert-equipment remote collaborative intelligent decision-making.

[0044] The present invention innovatively constructs a three-level collaborative intelligent control system that combines real-time control on the device side, intelligent prediction and dynamic parameter adjustment at the edge layer, and long-term cloud-based optimization with remote expert intervention. It combines hierarchical uploading of real-time and historical data with a distributed collaborative control strategy, effectively taking into account data integrity, real-time performance, intelligence, and system scalability, forming an overall intelligent copper wire bonding control solution for complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only schematic diagrams of the present invention. Those skilled in the art can also derive other drawings based on the provided drawings without inventive effort.

[0046] Figure 1 This is a schematic structural diagram of an IoT-based intelligent copper wire bonding equipment control system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the objectives, technical solutions, and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] Example 1, a method for controlling an intelligent copper bonding wire device based on the Internet of Things, comprising the following steps:

[0049] Step S1: Arrange pressure, temperature, ultrasonic power, current, vibration and microscopic image acquisition devices on the copper wire bonding equipment to synchronously acquire multi-physical quantity data and solder joint images during the welding process;

[0050] The sensor comprises:

[0051] Pressure sensors, ultrasonic energy sensors, temperature sensors, current sensors, and vibration sensors installed on the welding head or bond arm; microscopic imaging devices arranged in the welding area, which obtain the microscopic morphology of the solder joint and combine multi-physical quantity data to form a multimodal data source to provide basic data for subsequent quality prediction and dynamic parameter adjustment.

[0052] Step S2: Receive multi-physical quantity data through the edge computing gateway, and perform time alignment, filtering and denoising, feature extraction, and abnormal data removal;

[0053] The preprocessing of the edge computing gateway includes:

[0054] S201: Time synchronization and alignment:

[0055] The sampling frequencies and trigger times of the various sensors and microscopic imaging modules of the equipment vary. To ensure effective comparison and fusion analysis of multi-source data, the edge computing gateway first performs point-by-point synchronization alignment based on the timestamp information according to the welding cycle or weld point number, ensuring that the sensor data and microscopic image data within the same time window correspond one-to-one, thereby constructing an accurate time series feature set.

[0056] S202: Signal filtering and denoising:

[0057] Considering that factors such as equipment movement, ultrasonic vibration, and electromagnetic interference during the copper wire bonding process can cause noise interference to signals such as pressure, ultrasonic energy, and current, the edge computing gateway adopts the following denoising strategies based on sensor characteristics and signal features:

[0058] a. Pressure and ultrasonic energy signals: Wavelet decomposition + threshold denoising method is used to retain key fluctuation information and filter out high-frequency noise;

[0059] b. Current signal: Kalman filtering is used to predict and correct the signal curve in real time to improve the smoothness of the current change trend;

[0060] c. Vibration signal: Combining empirical mode decomposition (EMD) with instantaneous frequency analysis, this method distinguishes normal equipment operation vibration from abnormal impact signals, providing high-quality input for equipment health assessment.

[0061] S203: Image enhancement and feature area extraction:

[0062] For microscopic image data, the edge computing gateway performs the following image enhancement and region extraction operations on the original image:

[0063] Improve the contrast between the fusion zone and surrounding areas of the solder joint through adaptive histogram equalization;

[0064] Combine the Laplace operator with Canny edge detection to accurately identify the outline of the solder joint and the fusion boundary;

[0065] Automatically segment the solder joint core area, transition area and surrounding area based on morphological analysis and region growing algorithm;

[0066] Geometric features such as solder joint area, shape factor, and fusion boundary integrity are extracted, and combined with information such as grayscale distribution and texture direction to form image feature vectors, providing high-quality image features for subsequent multimodal feature fusion.

[0067] S204: Data integrity detection and abnormal data elimination:

[0068] To ensure the consistency and reliability of multi-physics data and image data, the edge computing gateway performs the following integrity checks and abnormal data removal:

[0069] For multi-sensor data within the same time window, the data integrity and credibility are judged based on time series continuity verification, physical quantity reasonable interval constraints and multi-sensor association logic rules;

[0070] For transient anomalies or mutation signals, we combine the Three Sigma principle with short-term trend anomaly detection to identify and mark abnormal data points;

[0071] For long-period loss or signal abnormality intervals, time series interpolation or trend extrapolation compensation is used to supplement missing data or correct abnormal points based on the changing trends of the previous and subsequent normal data segments.

[0072] S205: Sensor health assessment and dynamic weight allocation:

[0073] Taking into account the sensitivity attenuation and drift of sensors after long-term use, the edge computing gateway monitors the signal stability, signal-to-noise ratio and historical drift trend of each sensor in real time, and combines the sensor self-diagnosis information to evaluate the real-time health status of each sensor.

[0074] For sensor data with low health, a lower weight is assigned or marked as a secondary reference signal. The data contribution is dynamically adjusted in the subsequent feature fusion and solder joint quality prediction process to ensure the reliability and robustness of the overall analysis results.

[0075] S206: Feature normalization and multimodal data fusion preparation:

[0076] To eliminate the impact of differences in sensor ranges on subsequent feature analysis, the edge computing gateway normalizes multi-physical quantity data and image features, and forms a multimodal feature set for a single solder joint according to solder joint number, time sequence, and process stage, providing a unified standardized input data set for subsequent quality prediction model calls.

[0077] Step S3: Call the pre-trained welding quality prediction model in the edge computing gateway, take the multi-physical quantity sensor data, process parameters and solder joint image features as input, and output the solder joint quality score and defect category;

[0078] The edge computing gateway calls the pre-trained welding quality prediction model to perform a comprehensive analysis of the multi-physical quantity data, process parameters, and microscopic image features acquired and pre-processed in steps S1 and S2, and outputs the solder joint quality score and defect category, specifically including the following:

[0079] S31: Overall architecture of the model:

[0080] The welding quality prediction model adopts a multimodal deep learning framework and includes the following key components:

[0081] CNN image feature extraction network, responsible for performing layer-by-layer convolution and feature extraction on microscopic images to extract key visual features such as solder joint contours, fusion area morphology, and surface roughness;

[0082] LSTM time series analysis network, which performs time series modeling on multi-physical quantity sensor signals such as pressure, temperature, ultrasonic power, current, and vibration, and extracts dynamic change trends, small fluctuation characteristics, and abnormal mutation characteristics during the welding process;

[0083] The feature fusion layer aligns and deeply fuses the image features output by CNN and the temporal features output by LSTM. It also combines the real-time process parameters during welding (such as the current welding force and ultrasonic power setting) to form a multimodal fusion feature vector, providing complete information input for the final quality scoring and defect identification.

[0084] The scoring and classification layer outputs the quality score (numerical index) and defect category (normal, cold solder joint, over-solder joint, solder joint offset) of a single solder joint based on multimodal fusion features.

[0085] S32: Multimodal feature fusion method:

[0086] The feature fusion layer adopts cross-modal feature alignment and weighted fusion strategy:

[0087] Through timestamp synchronization, it is ensured that the image features and multi-physical quantity features belong to the same welding cycle;

[0088] Normalize the image features, time series features and process parameter features respectively;

[0089] Combined with the feature contribution analysis of historical samples, dynamic weights are assigned to different feature channels;

[0090] Through a multi-layer fully connected network, multimodal features are mapped layer by layer and fused with high-dimensional features, retaining the key information of each modality while eliminating the interference of low-correlation features, thereby enhancing the robustness of defect recognition.

[0091] S33 prediction model training process:

[0092] The welding quality prediction model is pre-trained based on massive historical production data and labeled defect samples. The training process includes:

[0093] Data preparation: Collect historical welding process data, microscopic image data, process parameter data and corresponding solder joint quality inspection results;

[0094] Data annotation: Combined with the results of microscopic inspection, mechanical testing and rework analysis, the defect type (cold solder joint, over-solder joint, offset) and quality score are annotated for historical samples;

[0095] Data enhancement: To address the imbalance in the number of defect category samples, we use multi-physical quantity time series enhancement and solder joint image simulation generation technology to supplement scarce defect samples;

[0096] Model training: A step-by-step pre-training + joint fine-tuning strategy is adopted. First, the CNN image feature extraction network and the LSTM time series analysis network are trained separately. Then, process parameter features are introduced into the feature fusion layer and the overall model is trained jointly.

[0097] Model evaluation: K-fold cross-validation is used to evaluate the model's adaptability to different materials, equipment states, and environmental conditions, and feature importance analysis is used to evaluate the contribution of multimodal information to the prediction results.

[0098] Model deployment: After training is completed, the model is converted into a lightweight inference format and deployed to the edge computing gateway to support real-time prediction on site.

[0099] S34: Prediction result output and feedback logic:

[0100] The welding quality prediction model performs real-time inference after each welding cycle and outputs the following results:

[0101] Solder joint quality score: quantifies the overall quality of the solder joint in the form of 0-100 points, with a higher score indicating better solder joint quality;

[0102] Defect category: Classify solder joints into normal, cold solder joints, over-solder joints, solder joint offset, and surface contamination;

[0103] Risk level: Combined with the score and defect category, the system automatically assesses the risk level of the current solder joint (low, medium, high). High-risk solder joints trigger subsequent process parameter adjustments and abnormal alarms.

[0104] The prediction results are transmitted to the parameter dynamic adjustment module of the edge gateway in real time, forming a closed-loop control logic of "real-time prediction-parameter adjustment-quality reassessment", which significantly improves the stability of the welding process and the ability to intervene in defects early.

[0105] Step S4: adjusting process parameters in real time according to the solder joint quality score and defect category, and sending adjustment instructions to the device controller via the edge gateway;

[0106] After receiving the solder joint quality score and defect category output by the prediction model, the edge computing gateway automatically calls the preset parameter adjustment strategy that matches the current defect characteristics based on the risk level classification and defect type characteristics.

[0107] The parameter adjustment strategy is stored in the edge parameter adjustment strategy library. This strategy library combines a large amount of historical production data, expert experience rules and equipment status feedback information to formulate in advance the adjustment range and adjustment amplitude of key parameters such as pressure, ultrasonic energy, welding time, etc. under different defect types and different scoring levels, ensuring that each defect has targeted compensation measures.

[0108] When calling a specific adjustment strategy, the edge computing gateway will also combine the current batch material characteristics, environmental parameters (such as temperature and humidity), and equipment operating status (such as the aging trend of the ultrasonic transducer) to perform real-time fine-tuning of the preset parameters through an adaptive compensation model to ensure that the parameter adjustment is consistent with historical experience and fits the current working conditions.

[0109] The adjusted parameters are sent to the equipment controller via EtherCAT or TSN high-speed channels, and the process parameters of the next solder joint are revised in real time, thus forming a dynamic closed-loop control of "real-time quality assessment - defect identification - parameter adaptive adjustment", significantly improving welding stability and early defect intervention effects.

[0110] Step S5: Upload the welding process data and welding results to the cloud platform. The cloud platform builds a long-term optimization model based on big data analysis, historical process parameters and expert rules, and periodically updates it to the edge device.

[0111] The method further includes equipment health prediction and fault warning steps, specifically: the edge computing gateway performs full-cycle operating status monitoring on key components of the copper wire bonding equipment (including welding heads, ultrasonic transducers, and transmission mechanisms) to obtain multi-dimensional health indicators such as vibration characteristics, ultrasonic system power curves, and temperature rise trends of key components.

[0112] The health monitoring data is combined with the equipment's historical operating data and historical fault records to construct an equipment health scoring model based on multi-feature trend analysis, abnormal pattern recognition and key parameter attenuation model, to conduct real-time assessment of the equipment's current health status and predict the remaining life (RUL) of key components.

[0113] The edge computing gateway uploads real-time health scores, remaining life prediction results, and abnormal feature data to the cloud platform. The cloud platform further confirms potential failure modes through big data comparative analysis, and combines the preset expert knowledge rule base to conduct secondary verification and risk classification of the prediction results to form an equipment health report.

[0114] The cloud platform also supports a remote expert diagnosis collaboration mechanism, where experts can view equipment health data and prediction results in real time. Based on on-site production conditions, equipment maintenance records and other information, they can make targeted maintenance suggestions or remotely issue equipment adjustment, maintenance or shutdown warning instructions, thus realizing a full life cycle health management closed loop of equipment health prediction, remote diagnosis and real-time intervention.

[0115] Step S6: Combining real-time closed-loop control, edge intelligent prediction, and long-term cloud optimization, a three-level collaborative intelligent control system is formed to dynamically adapt to changes in materials, equipment status, and the environment, thereby improving welding stability and finished product quality.

[0116] The method supports remote expert intervention and parameter optimization collaborative mode. Experts remotely access the real-time operating status of the equipment, welding process data, weld quality scores and defect classification results through the cloud platform, and grasp the equipment operating conditions and product quality trends during the welding process in real time.

[0117] The cloud platform not only displays real-time data, but also summarizes the equipment's historical process parameters, defect distribution statistics, solder joint quality score trends, and equipment health assessment results. Combined with the cloud-based expert knowledge base, it conducts a comprehensive analysis and trend prediction of current welding anomalies, forms a problem diagnosis report, and provides all-round support for expert decision-making.

[0118] When abnormal trends are obvious or high-risk defects occur frequently, remote experts can combine historical data comparison, abnormal feature analysis and expert rule reasoning to remotely adjust process parameters directly on the cloud platform, such as fine-tuning key parameters such as pressure, welding time, and ultrasonic energy;

[0119] At the same time, the cloud platform supports experts to remotely issue temporary control strategies, such as adjusting the limit of equipment operating parameters, calling special parameters for specific batches of materials, or switching special process flows, to ensure rapid response in abnormal situations, prevent the expansion of defects or equipment damage, and realize remote expert-equipment linkage collaboration and abnormal emergency response.

[0120] Remote intervention operations are fully recorded and synchronously fed back to the edge computing gateway and device controller, forming a full-process traceability management mechanism of real-time control-remote intervention-historical record tracing, ensuring that key decisions are transparent and controllable, and providing data basis for subsequent process optimization and abnormality review.

[0121] The edge computing gateway and the device controller achieve high real-time data interaction through the EtherCAT or TSN protocol; the edge computing gateway and the cloud platform transmit historical data through the MQTT protocol, or transmit real-time status data through the OPC-UA protocol, to achieve hierarchical uploading and distributed collaborative control of historical data and real-time data.

[0122] Example 2, a method for controlling an intelligent copper wire bonding device for high-density semiconductor packaging based on the Internet of Things, for achieving high-density welding of semiconductors, comprising the following steps:

[0123] Step A1: Deploy a high-sensitivity pressure sensor, ultrasonic power sensor, current sensor, temperature sensor, vibration sensor, and high-resolution microscopic image acquisition device on a copper wire bonding device used for high-density semiconductor packaging to perform high-frequency synchronous acquisition of multi-physical quantity data and solder joint images during the formation of tiny solder joints.

[0124] The sensor and microscopic image acquisition device optimize the layout and sampling frequency while meeting the requirements of high-density packaging with small solder joint spacing, tiny solder joint size, and extremely narrow welding window, thereby achieving high-sensitivity, low-interference real-time perception during the welding process.

[0125] Step A2: Receive multi-physics data through the edge computing gateway and perform data preprocessing specifically designed for high-density solder joints, including time synchronization, signal denoising, feature extraction, and abnormal data identification and elimination;

[0126] In view of the characteristics of increased micro-vibration interference and overlapping electromagnetic coupling signals during high-density welding, a multi-stage filtering + adaptive signal separation algorithm is used, combined with the welding process window characteristics, to accurately distinguish interference signals from valid signals;

[0127] A sub-pixel edge recognition algorithm is used on microscopic images to accurately segment the boundaries of high-density solder joints, ensuring the fidelity of the features of tiny solder joint morphology.

[0128] Step A3: The edge computing gateway calls a solder joint quality prediction model trained specifically for high-density package welding. It combines multi-physical sensor data, solder joint microscopic image features, and current welding process parameters to output a solder joint quality score and defect category.

[0129] The prediction model enhances the ability to identify features specific to high-density packaging, such as the thermal impact of adjacent solder joints, stress concentration in micro-solder joints, and solder joint micro-molding quality, thereby improving the accuracy of identifying high-density soldering defects.

[0130] Step A4: Based on the solder joint quality score and defect category, the process parameters of the corresponding solder joint are adjusted in real time, and high-precision fine-tuning instructions are sent to the equipment controller through the edge computing gateway;

[0131] The edge computing gateway dynamically calculates the compensation amounts for process parameters such as pressure, ultrasonic energy, and welding time based on the solder joint spacing, solder joint arrangement density, and influencing factors of adjacent solder joints. It also combines historical welding data with an expert rule library to form a dynamic parameter compensation model dedicated to high-density solder joints, and adaptively adjusts point by point.

[0132] Step A5: Upload the high-density welding process data and welding results to the cloud platform. Based on big data analysis, historical process parameters, material batch characteristics, and high-density packaging expert rules, the cloud platform builds a long-term parameter optimization model for high-density packaging and periodically updates it to the edge device.

[0133] The cloud platform also combines the long-term reliability evaluation results of high-density packaging solder joints, and continuously iterates the temperature rise control strategy, ultrasonic energy distribution strategy and multi-point stress balance strategy during the welding process to provide dynamic parameter references for the edge.

[0134] Step A6: Combining real-time closed-loop control, edge intelligent prediction, and long-term cloud optimization, a three-level collaborative control system is formed for high-density packaging. This system dynamically adapts to small fluctuations in wire material, changes in thermal impact of adjacent solder joints, and equipment state evolution, improving the consistency and long-term reliability of high-density packaging welding.

[0135] The method supports remote expert intervention and specialized diagnosis of high-density solder joints. Experts remotely access high-density soldering process data, single-point quality scores, and defect distribution maps through a cloud platform. Combining batch process history data with expert experience rules, they remotely issue point-by-point fine-tuning instructions or global compensation strategies, enabling expert-equipment joint decision-making and ensuring stable batch quality under complex packaging structures.

[0136] The edge computing gateway and the device controller achieve high real-time data interaction through EtherCAT or TSN protocol;

[0137] The edge computing gateway and the cloud platform transmit historical data via the MQTT protocol, or transmit real-time status data via the OPC-UA protocol, to achieve hierarchical uploading and distributed collaborative control of historical and real-time data.

[0138] Through dynamic sharing of multi-source data, real-time prediction and point-by-point compensation control, a closed loop of intelligent monitoring of the entire process and dynamic quality optimization is formed for high-density packaging welding.

[0139] Example 3, an intelligent copper wire bonding equipment control system based on the Internet of Things, see Figure 1 Shown, including:

[0140] 1. Copper wire bonding equipment, the copper wire bonding equipment is configured with the following units:

[0141] The pressure monitoring unit is installed on the welding head or bonding arm to collect pressure data in real time during the welding process;

[0142] Ultrasonic power monitoring unit, which obtains ultrasonic energy output power and energy fluctuation information in real time;

[0143] Temperature monitoring unit, monitoring the instantaneous temperature and temperature rise curve of the welding area;

[0144] A current monitoring unit, which obtains the operating current characteristics of the ultrasonic transducer;

[0145] Vibration monitoring unit, which monitors the micro-vibration characteristics of the bond arm and welding head during the welding process;

[0146] Microscopic imaging unit, which obtains microscopic topography images of solder joint surfaces and forms high-resolution image sequences;

[0147] The data acquisition and synchronization unit synchronizes the time of the above multi-sensor and imaging data and packages them into a single weld point multimodal data set;

[0148] The equipment control unit receives real-time parameter adjustment instructions from the edge computing gateway and drives the equipment to dynamically adjust process parameters such as pressure, ultrasonic power, and welding time.

[0149] 2. Edge computing gateway, the edge computing gateway includes the following units:

[0150] The data receiving unit receives multi-physical quantity data and microscopic image data from the copper wire bonding equipment via the EtherCAT or TSN interface;

[0151] The time synchronization and preprocessing unit performs time alignment, filtering and denoising, feature extraction, and abnormal data removal on the received data;

[0152] The AI inference unit calls the pre-trained welding quality prediction model to perform quality scoring and defect classification on the multimodal data of a single solder joint;

[0153] The parameter dynamic adjustment unit calls the parameter adjustment strategy library based on the AI inference results and scoring levels, and generates real-time parameter adjustment instructions in combination with the adaptive compensation model;

[0154] The control instruction issuing unit sends the dynamically adjusted welding pressure, ultrasonic power, welding time and other parameters to the equipment control unit through the real-time communication interface;

[0155] Equipment health assessment unit, which analyzes equipment operating status, key component operating parameters, and historical health data in real time to generate health scores and remaining life prediction results;

[0156] The remote collaborative interface unit receives expert fine-tuning instructions or temporary control strategies remotely issued by the cloud platform and executes them in conjunction with the local control strategy;

[0157] The data upload unit uploads real-time process data, prediction results, parameter adjustment records, equipment health status and other data to the cloud platform according to the preset classification strategy.

[0158] 3. Cloud platform, which includes the following units:

[0159] Data receiving and storage unit, which receives and hierarchically stores real-time data, historical data, and prediction results uploaded by the edge computing gateway;

[0160] Remote monitoring unit provides a visual interface to display the real-time operating status of the equipment, solder joint quality score trends, defect distribution statistics and equipment health status;

[0161] Big data analysis unit, which conducts in-depth mining of historical data across the entire lifecycle, multiple scenarios, and multiple devices to analyze differences in material batches, equipment aging trends, and causes of defects;

[0162] Expert knowledge base unit, which stores welding process standards, expert experience rules and defect classification and disposal strategies, providing a rule basis for parameter optimization and remote intervention;

[0163] The long-term parameter optimization unit builds a multi-scenario dynamic parameter optimization model based on historical data, expert knowledge, and reinforcement learning algorithms, and regularly sends the latest parameter recommendation strategies to the edge computing gateway;

[0164] The equipment health prediction unit generates equipment health assessment reports based on uploaded real-time health scores, fault trend data, and historical maintenance records, and combines expert opinions to form remote warnings or maintenance recommendations;

[0165] The remote expert collaboration unit allows authorized experts to remotely view device status, historical data, and real-time prediction results, and send fine-tuning instructions or emergency strategies to the edge computing gateway through the remote collaboration interface.

[0166] 4. A communication module, comprising:

[0167] Device-edge communication unit, using EtherCAT or TSN protocol, enables high-real-time data transmission and control command interaction between copper wire bonding equipment and edge computing gateway;

[0168] The edge-to-cloud communication unit uses the MQTT protocol to transmit historical data and the OPC-UA protocol to transmit real-time status data, supporting hierarchical data transmission and remote monitoring;

[0169] Remote expert access unit, which supports remote expert login to the cloud platform via VPN or dedicated industrial network to participate in real-time monitoring, data analysis and remote intervention;

[0170] The security encryption and permission management unit implements end-to-end encryption for all uplink and downlink data, and ensures data security and traceability of remote control operation permissions through hierarchical permission control.

[0171] The system forms a three-level collaborative intelligent control system that combines real-time control on the device side, intelligent prediction and adaptive parameter adjustment on the edge, and long-term parameter optimization and equipment health prediction on the cloud side.

[0172] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling an intelligent copper bonding wire device based on the Internet of Things, characterized in that: include: Pressure, temperature, ultrasonic power, current, vibration and microscopic image acquisition devices are installed on the copper wire bonding equipment to synchronously collect multi-physical quantity data and solder joint images during the welding process; Receive multi-physical quantity data through the edge computing gateway and perform time alignment, filtering and denoising, feature extraction, and abnormal data removal; The pre-trained welding quality prediction model is called in the edge computing gateway, taking multi-physical sensor data, process parameters, and solder joint image features as input, and outputting solder joint quality scores and defect categories. Adjust process parameters in real time based on the solder joint quality score and defect category, and send adjustment instructions to the equipment controller via the edge gateway; Upload welding process data and welding results to the cloud platform, which builds a long-term optimization model based on big data analysis, historical process parameters and expert rules, and periodically updates it to the edge device; Combining real-time closed-loop control, edge intelligent prediction and long-term cloud optimization, a three-level collaborative intelligent control system is formed, which dynamically adapts to changes in materials, equipment status and environment, and improves welding stability and finished product quality.

2. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The sensor comprises: Pressure sensors, ultrasonic energy sensors, temperature sensors, current sensors and vibration sensors installed on the bonding head or bonding arm; A microscopic imaging device is arranged in the welding area, which obtains the microscopic morphology of the welding point and forms a multimodal data source by combining multi-physical quantity data.

3. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The preprocessing of the edge computing gateway includes: Synchronize and align multi-physical quantity data based on timestamps; De-noising of pressure, ultrasonic energy and current data by Kalman filtering or wavelet transform; Perform edge enhancement, contrast enhancement, and key area segmentation of solder joints on microscopic images; Combine data integrity detection with sensor health assessment to eliminate or compensate for abnormal and missing data.

4. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The welding quality prediction model includes: CNN image feature extraction network to obtain solder joint geometry, surface roughness and fusion area features; LSTM time series analysis network, which analyzes the temporal trends and mutation characteristics of multi-physical quantity data; The feature fusion layer combines the output features of CNN and LSTM, and integrates the process parameters to form the solder joint quality score and defect classification results.

5. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: After receiving the defect category and quality score output by the prediction model, the edge computing gateway automatically calls the preset parameter adjustment strategy based on the defect type and risk level, and adjusts the pressure, ultrasonic energy or welding time in real time; among them, the parameter adjustment rules are dynamically optimized by combining the historical welding parameter library, expert process rule library and adaptive compensation model.

6. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The cloud platform periodically receives welding process data, solder joint quality scores, and defect classification results uploaded by the edge computing gateway, and builds a long-term parameter optimization model based on big data analysis, expert knowledge, and reinforcement learning algorithms; The optimization model combines the equipment operating status, material batch differences and environmental parameter changes to form a dynamic parameter recommendation strategy for the entire life cycle, and regularly updates it to the edge computing gateway through the parameter delivery interface.

7. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The method further includes equipment health prediction and fault warning steps, specifically: Real-time monitoring of equipment operation vibration characteristics, ultrasonic system power curves, and temperature rise changes of key components through the edge computing gateway; Combining historical fault records and expert rules, we build equipment health score and remaining life prediction models; The prediction results are uploaded to the cloud platform in real time and combined with expert remote diagnosis opinions to form a closed loop of equipment health management.

8. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The edge computing gateway and the device controller achieve high real-time data interaction through EtherCAT or TSN protocol; The edge computing gateway and the cloud platform transmit historical data through the MQTT protocol, or transmit real-time status data through the OPC-UA protocol, realizing hierarchical uploading and distributed collaborative control of historical data and real-time data.

9. The method for controlling an intelligent copper bonding wire device based on the Internet of Things according to claim 1, wherein: The method supports remote expert intervention and parameter optimization collaborative mode. Experts remotely access the equipment operating status, real-time quality scores and defect classification results through the cloud platform. By combining historical welding process data with expert rules, they remotely issue parameter fine-tuning instructions or temporary control strategies to achieve expert-equipment collaborative decision-making and emergency response.

10. An intelligent copper wire bonding equipment control system based on the Internet of Things, characterized in that: The system applies any one of claims 1 to 9 above, a method for controlling an intelligent copper bonding wire device based on the Internet of Things, comprising: Copper wire bonding equipment, equipped with multiple sensors and microscopic imaging modules, is used to collect multi-physical quantity signals and solder joint images during the welding process in real time; Edge computing gateways receive device sensor data, perform data preprocessing, AI reasoning analysis, and dynamic adjustment of process parameters, and issue control instructions to device controllers. The cloud platform receives data uploaded by devices and edge gateways, performs remote monitoring, big data analysis, long-term optimization of process parameters and equipment health prediction, and periodically issues optimization strategies; Communication module: devices communicate with edge gateways via EtherCAT or TSN, and edge gateways communicate with cloud platforms via MQTT or OPC-UA, enabling hierarchical data transmission and remote collaborative control. The system forms a three-level collaborative intelligent control system that combines real-time control on the device side, intelligent prediction and adaptive parameter adjustment on the edge, and long-term parameter optimization and equipment health prediction on the cloud side.

Images