Intelligent production line fault prediction and adaptive adjustment system and method based on artificial intelligence

Abstract

The invention discloses an intelligent production line fault prediction and adaptive adjustment system and method based on artificial intelligence. In the modern manufacturing industry, a traditional production line maintenance mode is difficult to find potential faults in advance, so that shutdown loss is huge, and the problems of insufficient data utilization, poor prediction and adjustment effects and the like exist in the prior art. The system comprises a data acquisition module, a data processing module, a fault prediction module and a self-adaptive adjustment module. The data acquisition module acquires multi-modal data through various sensors and carries out edge calculation preprocessing on the multi-modal data. The data processing module cleans data by using deep learning, extracts features, and stores the features by using a block chain technology; the fault prediction module fuses multiple technologies to improve the prediction precision; and the adaptive adjustment module generates and optimizes an adjustment strategy by using reinforcement learning. The method comprises the steps of data acquisition, processing, fault prediction and adaptive adjustment. According to the method, accurate fault prediction and efficient adaptive adjustment are realized, the intelligent level of a production line is improved, the fault rate and the production cost are reduced, the enterprise competitiveness is enhanced, and the method has a wide application prospect in the field of intelligent manufacturing.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and particularly relates to a system and method for realizing fault prediction and adaptive adjustment of an intelligent production line by using artificial intelligence, big data, sensor technology, etc., aiming to improve the operation efficiency, stability and reliability of the production line. Background Art

[0002] In the current environment of increasingly fierce global manufacturing competition, the efficient and stable operation of the production line is one of the core elements for an enterprise to maintain its competitiveness. As the manufacturing industry continues to move towards intelligence and automation, the complexity of the production line has increased sharply, the correlation and coordination between equipment have become closer, and any fault in any link may trigger a chain reaction, resulting in the paralysis of the entire production line.

[0003] Traditional production line maintenance methods mainly rely on regular inspections and after-fact repairs. The regular inspection mode checks equipment at fixed time intervals, and this method has obvious blindness. Since the operating state of the equipment cannot be understood in real time, potential fault hazards may have occurred in the equipment between two inspections but have not been discovered and processed in time. For example, in an electronic chip manufacturing production line, the wear of key components of some high-precision equipment may develop rapidly in a short time, and it is difficult for regular inspections to capture such subtle changes. Once a component is damaged, it will not only cause the equipment to stop running, but also may make the chips being produced scrapped, resulting in huge economic losses.

[0004] After-fact repair is carried out only after the equipment fails and the production line is forced to stop. This method will not only cause production stagnation, delay the order delivery time, but also may seriously affect the product quality due to the abnormal operation of the equipment during the fault period. According to statistics, in some large automobile manufacturing enterprises, when the production line stops due to a fault for one day, the direct economic loss can reach millions or even tens of millions of yuan, not including the potential loss of business reputation caused by the delay of customer orders.

[0005] With the continuous development of emerging technologies such as the Internet of Things, big data, and artificial intelligence, the manufacturing industry's demand for intelligent management of production lines has become increasingly urgent. Although there are already some fault warning systems based on monitoring technologies in the market, most of them have problems such as incomplete data collection and inaccurate analysis algorithms. Some systems only rely on a single type of sensor data and cannot comprehensively reflect the operating status of equipment; in terms of data analysis, traditional algorithms are difficult to process complex industrial data, resulting in a low accuracy rate of fault prediction, and false alarms and missed alarms occur frequently. At the same time, even if a fault is predicted, the existing adjustment means often lack flexibility and real-time performance and cannot quickly make the optimal adjustment decision according to the actual situation of the production line. Therefore, how to use advanced technical means to build an intelligent production line management system that can monitor in real time, accurately predict faults, and achieve adaptive adjustment has become an important issue that the current manufacturing industry urgently needs to solve. Summary of the Invention

[0006] The core objective of the present invention is to provide a highly intelligent production line fault prediction and adaptive adjustment system and method. Through comprehensive data collection, in-depth data mining and analysis, and intelligent adjustment strategies, it realizes early warning and accurate prediction of potential faults in the production line, and automatically adjusts production line parameters before or when a fault occurs to ensure the continuous, stable and efficient operation of the production line, thereby reducing the production cost of enterprises and improving production efficiency and product quality.

[0007] In the first aspect, an embodiment of the present invention provides an intelligent production line fault prediction and adaptive adjustment system based on artificial intelligence, including:

[0008] A data collection module, which is used to deploy various types of sensors at key parts of the production line, including temperature sensors, pressure sensors, vibration sensors, current sensors, displacement sensors, vision sensors, and acoustic sensors, to collect physical parameters of equipment operation, material state information, and production environment data in real time. The data acquisition card converts the analog signals collected by the sensors into digital signals and transmits them to the data processing module, and a redundant transmission link design is adopted to ensure data transmission stability;

[0009] A data processing module, including a data cleaning unit, which uses algorithms based on statistics and machine learning to perform deduplication, noise reduction, outlier processing, and missing data filling on the original data; a feature extraction unit, which uses a model combining a convolutional neural network (CNN) and a recurrent neural network (RNN) to extract deep-level and high-dimensional fault features from the cleaned data; a data storage unit, which selects a distributed database system combined with data warehouse technology to classify and store and manage the processed data and feature parameters;

[0010] The fault prediction module constructs a multi-layer fault prediction model based on a deep neural network (DNN), combines a long short-term memory network (LSTM) and an attention mechanism, performs supervised learning using a large amount of historical fault data and normal operation data, and introduces transfer learning technology to improve the generalization ability of the model, outputting information such as fault types, fault occurrence probabilities, and estimated occurrence times;

[0011] The adaptive adjustment module formulates an adjustment strategy according to the output result of the fault prediction module through a combination of an expert system and a reinforcement learning algorithm, controls the execution unit to convert the adjustment strategy into a control instruction and send it to the production line controller, and adopts a redundant control and verification mechanism to ensure the accurate execution of the control instruction.

[0012] In some embodiments of the present invention, the edge computing device in the above data acquisition module performs preliminary screening and preprocessing on the collected data, reducing the data transmission volume and improving the data processing efficiency and real-time performance.

[0013] In some embodiments of the present invention, the above data cleaning unit adopts an abnormal data detection model based on deep learning, such as a generative adversarial network (GAN) combined with an isolation forest algorithm, to accurately identify and remove abnormal data.

[0014] In some embodiments of the present invention, the above data storage unit adopts a distributed database, such as the Hadoop Distributed File System (HDFS) combined with a NoSQL database (such as Cassandra), and uses blockchain technology to ensure the security, immutability, and traceability of the data.

[0015] In some embodiments of the present invention, during the training process of the above fault prediction module, the fault data of other similar production lines is used to pre-train the model to optimize the model parameters when the data of this production line is limited.

[0016] In some embodiments of the present invention, a large amount of production experience and fault handling rules are pre-stored in the expert system of the above adaptive adjustment module, and the reinforcement learning algorithm optimizes the adjustment strategy according to the feedback result after adjustment by interacting with the production line environment.

[0017] Furthermore, the system of the present invention is composed as follows:

[0018] In some embodiments of the present invention, the above data acquisition module: a variety of types of sensors are widely deployed at various key parts of the production line, such as key rotating components of equipment, material conveying nodes, environmental control areas, etc., including temperature sensors, pressure sensors, vibration sensors, current sensors, displacement sensors, vision sensors, and acoustic sensors, etc. These sensors collect physical parameters, material state information, and production environment data during the operation of the equipment in real time, realizing all-round monitoring of the production line.

[0019] In some embodiments of the present invention, it further includes a data acquisition card: responsible for quickly and accurately converting the analog signals collected by the sensors into digital signals, and transmitting the data to the data processing module through a high-speed data transmission bus or wireless communication technology. To ensure the stability and reliability of data transmission, a redundant transmission link design is adopted. When the main transmission link fails, the backup link can automatically take over the data transmission work to avoid data loss.

[0020] In some embodiments of the present invention, the above data processing module: the data cleaning unit uses algorithms based on statistics and machine learning to perform operations such as deduplication, noise reduction, and outlier processing on the collected raw data. By setting reasonable data thresholds and data distribution models, it identifies and removes abnormal data points caused by sensor failures, electromagnetic interference, etc., and at the same time intelligently fills in missing data to ensure the integrity and accuracy of the data.

[0021] In some embodiments of the present invention, the above feature extraction unit: adopts a model that combines convolutional neural network (CNN) and recurrent neural network (RNN) in deep learning to automatically extract deep-level and high-dimensional fault features from the cleaned data. CNN is good at processing images and spatial data and can effectively extract local features in sensor data; RNN has strong processing ability for time series data and can capture the dynamic change trends and long-term dependence relationships of the data. The combination of the two can more accurately extract fault-related feature parameters, such as frequency features of vibration signals, trend features of temperature changes, etc.

[0022] In some embodiments of the present invention, the above data storage unit: selects a distributed database system, such as Hadoop Distributed File System (HDFS) combined with a NoSQL database (such as Cassandra), to classify and store the processed data and the extracted feature parameters. This storage method can not only meet the storage requirements of massive industrial data but also has high scalability and high fault tolerance. At the same time, data warehouse technology is used to archive and manage historical data for convenient subsequent data mining and analysis.

[0023] In some embodiments of the present invention, the above-mentioned fault prediction module: constructs a multi-layer fault prediction model based on a deep neural network (DNN), combining a long short-term memory network (LSTM) and an attention mechanism. The LSTM can effectively handle the long-term dependence problem in time series data, avoid the phenomena of gradient disappearance and gradient explosion, and better learn the development trend of faults; the attention mechanism allows the model to pay more attention to key information when processing data, improving the accuracy of prediction. In the model training stage, a large amount of historical fault data and normal operation data are used for supervised learning, and at the same time, transfer learning technology is introduced to draw on the fault data and model training experience of other similar production lines, so as to quickly improve the generalization ability of the model in the case of limited data on this production line.

[0024] In some embodiments of the present invention, the above-mentioned adaptive adjustment module: when the fault prediction module outputs a fault prediction result, the adaptive adjustment module is immediately activated. According to the fault type, severity, and the real-time operating state of the production line, an adjustment strategy is formulated by combining an expert system and a reinforcement learning algorithm. A large amount of production experience and fault handling rules are pre-stored in the expert system, which can quickly give a preliminary adjustment plan; the reinforcement learning algorithm continuously interacts with the production line environment, optimizes the adjustment strategy according to the feedback results after adjustment, and finds the optimal adjustment plan.

[0025] In some embodiments of the present invention, the above-mentioned control execution unit: converts the adjustment strategy into specific control instructions and sends them to the controller of the production line. To ensure the accurate execution of the control instructions, a redundant control and verification mechanism is adopted. Before sending the control instructions, the instructions are verified and error-correction encoded multiple times; during the execution process, the execution results are monitored in real time and compared with the expected results, and once a deviation is found, adjustments and corrections are immediately made.

[0026] In a second aspect, an embodiment of the present application provides an intelligent production line fault prediction and adaptive adjustment method based on artificial intelligence, including the following steps:

[0027] Data acquisition step: various sensors in the data acquisition module collect production line operation data according to a preset sampling frequency. After the data acquisition card converts the analog signal into a digital signal, it is transmitted to the data processing module through a redundant transmission link;

[0028] Data processing step: after the data processing module receives the data, the data cleaning unit cleans and preprocesses the data, removes abnormal data and noise, the feature extraction unit extracts data features using a CNN-RNN model, and the data storage unit stores the processed data and feature parameters in a distributed database;

[0029] Fault prediction steps: The fault prediction module obtains real-time characteristic parameters from the database and inputs them into the fault prediction model based on DNN-LSTM-Attention for analysis and prediction. The model is continuously optimized by combining transfer learning technology during the training process, and outputs information such as fault types, fault occurrence probabilities, and expected occurrence times.

[0030] Adaptive adjustment steps: The adaptive adjustment module formulates an adjustment strategy according to the fault prediction results through an expert system and a reinforcement learning algorithm, and controls the execution unit to convert the adjustment strategy into control instructions and send them to the production line controller. Redundant control and verification are performed during the execution process, and the adjustment strategy is dynamically optimized according to real-time feedback.

[0031] In some embodiments of the present invention, in the above data acquisition step, the edge computing device preliminarily screens and preprocesses the acquired data to reduce the data transmission volume.

[0032] In some embodiments of the present invention, in the above data processing step, the data cleaning unit uses a deep learning-based abnormal data detection model for data cleaning, and the data storage unit uses blockchain technology to ensure data security and traceability.

[0033] In some embodiments of the present invention, in the above fault prediction step, the model is pre-trained with the fault data of other similar production lines to optimize the performance of the model in the case of limited data on this production line; in the above adaptive adjustment step, the reinforcement learning algorithm continuously optimizes the adjustment strategy according to the real-time feedback of the production line.

[0034] The embodiments of the present invention have at least the following advantages or beneficial effects:

[0035] The present invention deploys various types of sensors at key parts of the production line, which can comprehensively and real-time obtain the physical parameters of equipment operation, material state information, and production environment data, avoiding inaccurate judgment of equipment status due to the limitations of a single data type. The redundant transmission link design ensures the stability and reliability of data transmission, prevents data loss, and provides a solid data foundation for subsequent fault prediction and adjustment.

[0036] Each unit of the data processing module of the present invention works in cooperation. The data cleaning unit ensures the quality and accuracy of the data and removes interference information; the feature extraction unit uses an advanced neural network model to extract deep-level and high-dimensional fault features, which helps to more accurately identify potential fault hazards; the distributed database system combines data warehouse technology to efficiently store and manage massive data, facilitating subsequent data mining and analysis.

[0037] The fault prediction module of the present invention adopts a model that combines a deep neural network with LSTM and an attention mechanism, which can fully consider the development trend and key information of faults, and combines supervised learning and transfer learning technologies, improving the accuracy and lead time of fault prediction, enabling enterprises to take countermeasures in advance and reducing the losses caused by faults.

[0038] In the adaptive adjustment module of the present invention, an adjustment strategy is formulated by combining an expert system and a reinforcement learning algorithm, which not only utilizes rich production experience but also can dynamically optimize the strategy according to real-time feedback. The redundant control and verification mechanism ensure the accurate execution of control instructions, guaranteeing that the production line can quickly and accurately respond to the fault prediction results and maintain stable operation.

[0039] The present invention conducts preliminary screening and preprocessing of data at the data collection site through edge computing devices, reducing the amount of data that needs to be transmitted to the data processing module, reducing the network transmission pressure, and at the same time improving the real-time performance of data processing, enabling the system to respond more quickly to the operating state of the production line.

[0040] The abnormal data detection model based on deep learning of the present invention, such as GAN combined with the isolation forest algorithm, can more accurately identify and remove abnormal data. Compared with traditional data cleaning methods, this model can learn the complex distribution and characteristics of data, effectively avoiding incorrect analysis and prediction caused by abnormal data and improving the accuracy of subsequent fault prediction.

[0041] The present invention provides a storage solution with high scalability and high fault tolerance through a distributed database such as HDFS combined with a NoSQL database (such as Cassandra), which can meet the storage requirements of massive industrial data. The application of blockchain technology ensures the security, immutability, and traceability of data, enabling data to be reliably protected throughout its life cycle and providing a credible data basis for enterprise decision-making.

[0042] During the training process of the fault prediction module of the present invention, the fault data of other similar production lines are used for pre-training. Even when the data of this production line is limited, the model can learn more extensive fault patterns and characteristics, thereby enhancing the generalization ability of the model and enabling the model to maintain a high prediction accuracy when facing different production environments and equipment states.

[0043] The present invention can quickly give a preliminary adjustment plan through a large amount of production experience and fault handling rules pre-stored in the expert system, providing a reasonable starting point for the reinforcement learning algorithm. The reinforcement learning algorithm continuously optimizes the adjustment strategy according to the feedback results after adjustment through interaction with the production line environment, finds the optimal solution, and improves the production line's ability to respond to faults and adaptive ability.

[0044] The method of the present invention covers the complete processes of data acquisition, processing, fault prediction, and adaptive adjustment. Each step is closely connected and works collaboratively to form a closed-loop intelligent management system. From the real-time data acquisition to the execution of the final adjustment strategy, it ensures that the operating status of the production line can be comprehensively and timely monitored and optimized, improving the overall operating efficiency and stability of the production line.

[0045] In the data acquisition step of the present invention, an edge computing device is introduced for preliminary screening and preprocessing, reducing the data transmission volume, decreasing the occupancy of network bandwidth, and at the same time accelerating the data processing speed, enabling the system to more efficiently process and analyze data and timely detect potential fault hazards.

[0046] In the data cleaning unit of the present invention, a deep learning-based abnormal data detection model is used for data cleaning, improving the quality and reliability of the data. The data storage unit utilizes blockchain technology to ensure data security and traceability, enabling enterprises to safely use and manage data, providing a solid foundation for fault prediction and decision-making.

[0047] In the fault prediction step of the present invention, the fault data of other similar production lines are used for pre-training, optimizing the performance of the model in the case of limited data of this production line and improving the accuracy of fault prediction. In the adaptive adjustment step, the reinforcement learning algorithm continuously optimizes the adjustment strategy according to the real-time feedback of the production line, enabling the production line to better adapt to different fault situations, achieving more accurate and efficient adaptive adjustment, and further reducing the impact of faults on production. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 is the principle block diagram of the present invention;

[0050] Figure 2 is the schematic diagram of the process steps of the present invention;

[0051] Figure 3 is the structural block diagram of an electronic device provided by an embodiment of the present invention.

[0052] Explanation of reference numerals: 101, memory; 102, processor; 103, communication interface. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the drawings herein can be arranged and designed in a variety of different configurations.

[0054] Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of the present invention.

[0055] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not require further definition and explanation in subsequent drawings.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0057] In the description of the embodiments of the present invention, "a plurality of" represents at least two.

[0058] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly defined and limited, if the terms "arranged", "installed", "connected", and "coupled" are used, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0059] Such as Figures 1-3 , in the first aspect, an artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system provided by an embodiment of the present invention includes:

[0060] A data acquisition module is used to deploy various types of sensors at key parts of the production line, including temperature sensors, pressure sensors, vibration sensors, current sensors, displacement sensors, vision sensors, and acoustic sensors, to collect physical parameters of equipment operation, material status information, and production environment data in real time. The data acquisition card converts the analog signals collected by the sensors into digital signals and transmits them to the data processing module, and a redundant transmission link design is adopted to ensure data transmission stability;

[0061] A data processing module includes a data cleaning unit that uses algorithms based on statistics and machine learning to remove duplicates, reduce noise, handle outliers, and fill in missing data from the original data; a feature extraction unit that uses a model combining convolutional neural network (CNN) and recurrent neural network (RNN) to extract deep-level and high-dimensional fault features from the cleaned data; a data storage unit that selects a distributed database system combined with data warehouse technology to classify and store and manage the processed data and feature parameters;

[0062] A fault prediction module constructs a multi-layer fault prediction model based on a deep neural network (DNN), combines long short-term memory network (LSTM) and attention mechanism, uses a large amount of historical fault data and normal operation data for supervised learning, and introduces transfer learning technology to improve the generalization ability of the model, and outputs information such as fault type, fault occurrence probability, and estimated occurrence time;

[0063] An adaptive adjustment module formulates an adjustment strategy by combining an expert system and a reinforcement learning algorithm according to the output result of the fault prediction module, controls the execution unit to convert the adjustment strategy into a control instruction and send it to the production line controller, and adopts a redundant control and verification mechanism to ensure the accurate execution of the control instruction.

[0064] In a second aspect, an embodiment of the present invention provides an intelligent production line fault prediction and adaptive adjustment method based on artificial intelligence, including the following steps:

[0065] A data acquisition step: various sensors in the data acquisition module collect production line operation data according to a preset sampling frequency. After the data acquisition card converts the analog signal into a digital signal, it is transmitted to the data processing module through a redundant transmission link;

[0066] A data processing step: after the data processing module receives the data, the data cleaning unit cleans and preprocesses the data, removes abnormal data and noise, the feature extraction unit uses the CNN-RNN model to extract data features, and the data storage unit stores the processed data and feature parameters in a distributed database;

[0067] Fault prediction steps: The fault prediction module obtains real-time characteristic parameters from the database and inputs them into the fault prediction model based on DNN-LSTM-Attention for analysis and prediction. During the training process, the model is continuously optimized by combining transfer learning techniques, and outputs information such as fault types, fault occurrence probabilities, and estimated occurrence times.

[0068] Adaptive adjustment steps: The adaptive adjustment module formulates adjustment strategies according to the fault prediction results through an expert system and a reinforcement learning algorithm, and controls the execution unit to convert the adjustment strategies into control instructions and send them to the production line controller. Redundancy control and verification are carried out during the execution process, and the adjustment strategies are dynamically optimized according to real-time feedback.

[0069] Embodiment 1:

[0070] Application of the single-crystalline silicon solar cell production line based on the system and method of the present invention

[0071] Application scenario: In the production line of a single-crystalline silicon solar cell manufacturing enterprise, the intelligent production line fault prediction and adaptive adjustment system and method based on artificial intelligence of the present invention are introduced. The production line includes multiple key processes such as silicon wafer cutting, surface texturing, diffusion junction formation, etching, coating, and printing sintering.

[0072] Specific implementation

[0073] Data acquisition: Vibration sensors, temperature sensors, and current sensors are installed on the silicon wafer cutting equipment to monitor the vibration of the cutting tool, the equipment temperature, and the motor current in real time; temperature sensors, pressure sensors, and gas flow sensors are installed on the diffusion furnace to monitor the temperature, pressure, and gas flow during the diffusion process. The data acquisition card converts these analog signals into digital signals and transmits them to the data processing module through industrial Ethernet.

[0074] Data processing: The data cleaning unit removes noise and outliers from the collected data. The feature extraction unit uses a CNN-RNN model to extract features related to equipment faults, such as the change features of the vibration frequency of the cutting tool and the trend features of the temperature fluctuation of the diffusion furnace, and stores the processed data in a distributed database.

[0075] Fault prediction: The fault prediction module predicts the wear condition of the cutting tool of the silicon wafer cutting equipment and the temperature control fault of the diffusion furnace according to historical data and real-time characteristic parameters. For example, when it is detected that the vibration frequency of the cutting tool abnormally increases and the current increases, it is predicted that the tool is about to wear and an early warning is issued.

[0076] Adaptive adjustment: The adaptive adjustment module adjusts the cutting speed and pressure of the silicon wafer cutting equipment according to the fault prediction results, and arranges a tool replacement plan at the same time; for the diffusion furnace, adjusts the heating power and gas flow to ensure the stability of the diffusion process.

[0077] Effect evaluation: By implementing the system and method of the present invention, the fault downtime of the single-crystalline silicon solar cell production line has been reduced by 35%, the product yield has been increased by 8%, and the production efficiency has been improved by 20%.

[0078] Comparative example 1:

[0079] Traditional regular inspection and after-fact maintenance mode

[0080] Application scenario: It is also a production line of a single-crystalline silicon solar cell manufacturing enterprise, adopting the traditional regular inspection and after-fact maintenance mode. The inspection personnel check the equipment at fixed time intervals (once a week), and only perform maintenance when the equipment fails.

[0081] Specific implementation

[0082] The inspection personnel use simple measuring tools (such as multimeters, thermometers, etc.) to check the key parameters of the equipment and record the operating status of the equipment.

[0083] When the equipment fails and causes the production line to stop, the maintenance personnel will conduct fault troubleshooting and maintenance on the equipment.

[0084] Effect evaluation: The fault downtime of the production line is relatively long, reaching more than 10 hours per month on average, the product yield is relatively low, only about 85%, and the production efficiency improvement is slow. Compared with the enterprise implementing the present invention, the production efficiency is more than 30% lower.

[0085] Example 2:

[0086] Optimized application of the polycrystalline silicon solar cell production line

[0087] Application scenario: The production line of a polycrystalline silicon solar cell manufacturing enterprise has problems such as equipment aging and frequent failures. The production line includes links such as polycrystalline silicon ingot casting, silicon wafer cutting, and cell manufacturing.

[0088] Data acquisition: Install multiple temperature sensors, liquid level sensors, and stress sensors on the polycrystalline silicon ingot casting furnace to monitor the temperature distribution, silicon liquid level, and ingot casting stress during the ingot casting process in real time; install vision sensors and pressure sensors on the cell printing equipment to monitor the quality of the printed pattern and the printing pressure. The data is transmitted to the data processing module through wireless communication technology.

[0089] Data processing: Use an abnormal data detection model based on deep learning to clean the data and remove abnormal values caused by sensor failures or environmental interference. The feature extraction unit uses a CNN-RNN model to extract key features, such as the temperature gradient feature of the ingot casting furnace and the edge feature of the cell printed pattern, and stores them in a distributed database.

[0090] Fault prediction: The fault prediction module combines transfer learning technology and uses the fault data of other similar production lines to pre-train the model, improving the accuracy of prediction. For example, predicting the thermal field fault of the ingot casting furnace and the nozzle blockage fault of the cell printing equipment.

[0091] Adaptive adjustment: The adaptive adjustment module adjusts the heating strategy and cooling speed of the ingot casting furnace according to the prediction results, avoiding ingot casting defects caused by uneven thermal field; for the cell printing equipment, adjusts the printing pressure and nozzle cleaning cycle to ensure printing quality.

[0092] Effect evaluation: After implementation, the equipment failure rate of the polysilicon solar cell production line has been reduced by 40%, the conversion efficiency of the products has been increased by 5% on average, and the production cost has been reduced by 15%.

[0093] Comparative example 2:

[0094] This single sensor monitoring and simple data analysis mode

[0095] Application scenario: In the production line of a polysilicon solar cell manufacturing enterprise, only a single type of sensor (such as a temperature sensor) is installed on some key equipment, and simple data analysis methods (such as threshold judgment) are used for fault monitoring.

[0096] Specific implementation

[0097] Install a temperature sensor on the polysilicon ingot casting furnace, set a temperature threshold, and send an alarm when the temperature exceeds the threshold.

[0098] Data analysts judge whether the equipment may have a fault according to the simple statistical analysis results of the temperature data.

[0099] Effect evaluation: Due to incomplete data collection, the true operating state of the equipment cannot be accurately reflected, the accuracy rate of fault prediction is low, and false alarms and missed alarms occur frequently. The equipment failure rate of the production line is high, the improvement of the conversion efficiency of the products is not obvious, and the production cost remains high.

[0100] Example 3:

[0101] Such as Figure 3, embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor 102, it implements the system according to any one of the above first aspects. If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0102] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention. For those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present application is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. An intelligent production line fault prediction and adaptive adjustment system based on artificial intelligence, characterized in that: include: The data acquisition module is used to deploy various types of sensors at key locations of the production line, including temperature sensors, pressure sensors, vibration sensors, current sensors, displacement sensors, visual sensors, and acoustic sensors, to collect real-time physical parameters of equipment operation, material status information, and production environment data. The data acquisition card converts the analog signals collected by the sensors into digital signals and transmits them to the data processing module, and adopts a redundant transmission link design to ensure data transmission stability. The data processing module, including the data cleaning unit, uses statistics and machine learning algorithms to remove duplication, reduce noise, process outliers, and fill in missing data for the original data; Features The extraction unit uses a model combining convolutional neural network (CNN) and recurrent neural network (RNN) to extract deep and high-dimensional fault features from the cleaned data; Data storage unit, using a distributed database system combined with data warehouse technology to classify, store and manage processed data and feature parameters; The fault prediction module builds a multi-layer fault prediction model based on a deep neural network (DNN), combines the long short-term memory network (LSTM) and the attention mechanism, uses a large amount of historical fault data and normal operation data for supervised learning, and introduces transfer learning technology to improve the generalization ability of the model, outputting information such as fault type, fault probability, and expected time of occurrence; The adaptive adjustment module formulates an adjustment strategy based on the output results of the fault prediction module by combining the expert system and the reinforcement learning algorithm. The control execution unit converts the adjustment strategy into control instructions and sends them to the production line controller, and adopts redundant control and verification mechanisms to ensure the accurate execution of control instructions.

2. The artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system according to claim 1 is characterized in that: The edge computing device in the data acquisition module performs preliminary screening and preprocessing on the collected data, thereby reducing the amount of data transmission and improving data processing efficiency and real-time performance.

3. The artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system according to claim 1 is characterized in that: The data cleaning unit adopts an abnormal data detection model based on deep learning, such as a generative adversarial network (GAN) combined with an isolation forest algorithm, to accurately identify and remove abnormal data.

4. The artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system according to claim 1 is characterized in that: The data storage unit adopts a distributed database, such as Hadoop Distributed File System (HDFS) combined with a NoSQL database (such as Cassandra), and uses blockchain technology to ensure the security, non-tamperability and traceability of the data.

5. The artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system according to claim 1 is characterized in that: During the training process, the fault prediction module uses the fault data of other similar production lines to pre-train the model and optimizes the model parameters when the data of the current production line is limited.

6. The artificial intelligence-based intelligent production line fault prediction and adaptive adjustment system according to claim 1 is characterized in that: A large amount of production experience and fault handling rules are pre-stored in the expert system of the adaptive adjustment module, and the reinforcement learning algorithm optimizes the adjustment strategy according to the adjusted feedback results by interacting with the production line environment.

7. An artificial intelligence-based intelligent production line fault prediction and adaptive adjustment method, characterized in that: The following steps are involved: Data collection steps: Various sensors in the data collection module collect production line operation data according to the preset sampling frequency. The data collection card converts the analog signal into a digital signal and transmits it to the data processing module through a redundant transmission link; Data processing step: After the data processing module receives the data, the data cleaning unit cleans and preprocesses the data to remove abnormal data and noise. The feature extraction unit uses the CNN-RNN model to extract data features. The data storage unit stores the processed data and feature parameters in a distributed database. Fault prediction step: The fault prediction module obtains real-time feature parameters from the database and inputs them into the fault prediction model based on DNN-LSTM-Attention for analysis and prediction. The model is continuously optimized in the training process by combining transfer learning technology, and outputs information such as fault type, fault probability, and expected time of occurrence. Adaptive adjustment steps: The adaptive adjustment module formulates an adjustment strategy based on the fault prediction results through the expert system and reinforcement learning algorithm. The control execution unit converts the adjustment strategy into control instructions and sends them to the production line controller. Redundant control and verification are performed during the execution process, and the adjustment strategy is dynamically optimized based on real-time feedback.

8. The method for fault prediction and adaptive adjustment of intelligent production line based on artificial intelligence according to claim 7 is characterized in that: In the data collection step, the edge computing device performs preliminary screening and preprocessing on the collected data to reduce the amount of data transmission.

9. The method for fault prediction and adaptive adjustment of intelligent production line based on artificial intelligence according to claim 7, characterized in that: In the data processing step, the data cleaning unit uses an abnormal data detection model based on deep learning to perform data cleaning, and the data storage unit uses blockchain technology to ensure data security and traceability.

10. The method for fault prediction and adaptive adjustment of intelligent production line based on artificial intelligence according to claim 7, characterized in that: In the fault prediction step, the model is pre-trained using fault data from other similar production lines to optimize the performance of the model when the data of this production line is limited; in the adaptive adjustment step, the reinforcement learning algorithm continuously optimizes the adjustment strategy based on real-time feedback from the production line.

Images