Cloud manufacturing quality control and defect prediction method and system based on deep learning

Abstract

The present application relates to the technical field of cloud manufacturing, and particularly relates to a cloud manufacturing quality control and defect prediction method and system based on deep learning. The method comprises the following steps: acquiring production site data in a manufacturing process through a data acquisition device and uploading the data to the cloud for next step analysis; extracting information features of the production site data and performing feature fusion to obtain an information data feature set; constructing a joint deep learning prediction model, training the joint deep learning prediction model using the information data feature set; and completing cloud manufacturing quality control and defect prediction through the trained joint deep learning prediction model. The present application extracts various features of production site data, trains a joint deep learning prediction model, and then completes cloud manufacturing quality control and defect prediction, thereby solving the problems of quality and defect prediction in the manufacturing process, effectively improving production efficiency, optimizing product quality, and promoting the intelligentization and automation upgrade of the manufacturing process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cloud manufacturing, and particularly relates to a cloud manufacturing quality control and defect prediction method and system based on deep learning. BACKGROUND

[0002] Cloud manufacturing is a production mode that promotes the collaborative work and resource sharing of multiple links and participants through information technology, Internet of Things, big data and other means. Due to the complexity of the manufacturing process involved in cloud manufacturing and the influence of various factors, quality problems often occur. The dynamic nature and complexity of the production process make quality control and defect prediction have very high real-time requirements, and traditional quality control methods cannot meet this demand.

[0003] In the manufacturing process, various types and forms of defects emerge in an endless stream. The characteristics and patterns of these defects are often very complex, and traditional quality control methods are difficult to effectively identify and predict these defects. Existing quality control technologies are mostly based on experience and rules, relying on manual monitoring and periodic inspection, and it is difficult to fully utilize large-scale production data for intelligent decision-making. Therefore, traditional methods cannot accurately and timely perform quality control and cannot provide comprehensive optimization support for the manufacturing process. SUMMARY

[0004] In view of the deficiencies of existing methods and the needs of actual applications, in order to effectively improve production efficiency, optimize product quality, and promote the intelligentization and automation upgrade of the manufacturing process, solve the quality problem and defect prediction problem in the manufacturing process. On the one hand, the present application provides a cloud manufacturing quality control and defect prediction method based on deep learning, comprising the following steps:

[0005] Through a data acquisition device, production site data in the manufacturing process is obtained and uploaded to the cloud for further analysis; information features of the production site data are extracted and fused to obtain an information data feature set; a joint deep learning prediction model is constructed, and the joint deep learning prediction model is trained using the information data feature set; cloud manufacturing quality control and defect prediction are completed through the trained joint deep learning prediction model. The present application extracts various features of production site data, trains a joint deep learning prediction model, and then completes cloud manufacturing quality control and defect prediction, solves the quality problem and defect prediction problem in the manufacturing process, effectively improves production efficiency, optimizes product quality, and promotes the intelligentization and automation upgrade of the manufacturing process.

[0006] Optionally, the cloud manufacturing quality control and defect prediction method based on deep learning further comprises standardizing the production site data;

[0007] The standardization of the production site data satisfies the following formula:

[0008]

[0009]

[0010] wherein, denotes the value of the normalized numerical data, denotes the numerical data, denotes the mean value, denotes the standard deviation, denotes the normalized value of the normalized image, denotes the pixel value of the original image. The production site data is normalized to eliminate the influence of the numerical dimension on the prediction result, which is beneficial to improve the prediction efficiency of the present application.

[0011] Optionally, the production site data includes numerical production site data and image production site data.

[0012] The information features of the production site data are extracted, including the following steps:

[0013] The information features of the numerical production site data are extracted by using a statistical feature extraction method, and the information features are processed by dimension reduction by using a principal component analysis method; the local features in the image production site data are extracted by using a convolutional neural network, and high-level features are extracted layer by layer by using multiple convolutional layers and pooling layers. The present application uses different feature extraction methods for different types of production site data, which is beneficial to reasonably extract effective features of the data and further beneficial to improve the accuracy of the present application.

[0014] Optionally, the information features are fused to obtain an information data feature set, which satisfies the following formula:

[0015]

[0016] wherein, denotes the information data feature set, , denotes the weight coefficient, denotes the information feature of the numerical production site data, denotes the information feature of the image production site data. The present application sets the weight of the information feature according to the actual situation, which is further beneficial to identify the features of different types of production site data.

[0017] Optionally, the joint deep learning prediction model includes a defect category prediction sub-model and a defect occurrence probability prediction sub-model.

[0018] The output layer activation function of the defect category prediction sub-model satisfies the following formula:

[0019]

[0020] in, Indicates the output result. This represents the weight matrix of the fully connected layer. This represents the vectorized result of the information data feature set. The term "bias" is used; the output layer activation function of the defect occurrence probability prediction sub-model employs a linear activation function. This invention constructs activation functions for defect categories and defect occurrence probabilities separately, which facilitates accurate identification and prediction of defect categories and probabilities, further improving the prediction accuracy of this invention.

[0021] Optionally, training the joint deep learning prediction model using the information data feature set includes the following steps:

[0022] A defect category loss function and a defect occurrence probability loss function are constructed; the joint deep learning prediction model is trained using the loss functions and the information data feature set. This invention constructs loss functions for defect categories and defect occurrence probabilities separately, which is beneficial for accurately identifying and predicting defect categories and probability of occurrence, further improving the prediction accuracy of this invention.

[0023] Optionally, the defect category loss function satisfies the following formula:

[0024]

[0025] in, This represents the value of the loss function. Indicates the number of samples. Indicates the first The true label of each sample Indicates the first The probability that a sample belongs to the defect category.

[0026] Optionally, the defect occurrence probability loss function satisfies the following formula:

[0027]

[0028] in, This represents the value of the loss function. Indicates the number of samples. Indicates the first The true value of each sample Indicates the first The predicted value for each sample.

[0029] Optionally, the joint deep learning prediction model is trained using the loss function and the information data feature set, satisfying the following formula:

[0030]

[0031] wherein, denotes the iteratively updated model parameters, denotes the model parameters before iteration update, denotes the learning rate, denotes the gradient of the loss function. By calculating the loss, the model parameters are quickly adjusted, which is conducive to improving the training speed of the model.

[0032] In a second aspect, in order to efficiently execute the cloud manufacturing quality control and defect prediction method based on deep learning provided by the present application, the present application further provides a cloud manufacturing quality control and defect prediction system based on deep learning, comprising a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory are connected with each other, wherein the memory is used for storing a computer program, the computer program contains program instructions, the processor is configured to call the program instructions, and execute the cloud manufacturing quality control and defect prediction method based on deep learning as described in the first aspect of the present application. The cloud manufacturing quality control and defect prediction system based on deep learning provided by the present application has compact structure and stable performance, can stably execute the cloud manufacturing quality control and defect prediction method based on deep learning provided by the present application, and further improves the overall applicability and practical application ability of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A cloud manufacturing quality control and defect prediction method based on deep learning provided by an embodiment of the present application is shown in the flowchart;

[0034] Figure 2 A cloud manufacturing quality control and defect prediction system based on deep learning provided by an embodiment of the present application is shown in the block diagram;

[0035] Figure 3 A cloud manufacturing quality control and defect prediction device structure schematic diagram provided by an embodiment of the present application is shown in the block diagram. DETAILED DESCRIPTION

[0036] The specific embodiments of the present application will be described in detail below. It should be noted that the embodiments described herein are only used for illustration and do not limit the present application. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present application. However, it is obvious to those skilled in the art that the present application does not necessarily have to be implemented by using these specific details. In other examples, in order to avoid obscuring the present application, well-known circuits, software or methods are not specifically described.

[0037] Throughout this specification, references have been made to "one embodiment," "an embodiment," "one example," or "an example" meant to encompass the particular feature, structure, or characteristic following the embodiment or example. Thus, appearances of the phrases "in one embodiment," "in an embodiment," "one example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics can be combined in any suitable manner in one or more embodiments or examples. Moreover, those skilled in the art will appreciate that the diagrams provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0038] Referring to Figure 1 , in order to effectively improve production efficiency, optimize product quality, and promote the intelligentization and automation upgrade of the manufacturing process, the quality problem and defect prediction problem in the manufacturing process are solved. The application provides a cloud manufacturing quality control and defect prediction method based on deep learning, as shown in Figure 1 , in one embodiment, the method comprises the following steps:

[0039] S1, through the data acquisition equipment, the production site data in the manufacturing process is acquired and uploaded to the cloud for next step analysis.

[0040] In an embodiment, the data acquisition equipment includes a temperature sensor, a humidity sensor, a pressure sensor, and a high-definition camera. Through these data acquisition equipment, the environment data and production data related to the manufacturing process are collected, and other related data (such as text information) are combined to form a data stream uploaded to the cloud database.

[0041] The data content of each sensor is marked as , wherein , represents the maximum value of the sensor content , is the corresponding sensor value.

[0042] The high-definition camera is used to collect high-quality image data, has high sensitivity, high frame rate, and accurate color restoration capability, and can provide clear video stream data. The image data is marked as , wherein , is the maximum value of the image data mark.

[0043] The data stream in the uploading process is as follows:

[0044]

[0045] wherein, represents the data stream, a function representing data upload, a function representing sensor data, a function representing image data, a function representing other text or metadata.

[0046] S2, extracting information features of the production site data and performing feature fusion to obtain an information data feature set.

[0047] In one embodiment, before extracting the information features of the production site data, the production site data is also subjected to standardization processing;

[0048] The standardization processing of the production site data satisfies the following formula:

[0049]

[0050]

[0051] wherein, a value after standardization processing of numerical data, numerical data, a mean value, a standard deviation, a normalized value of the standardized image, a pixel value of the original image. Through standardization processing, the numerical production site data obtained by the sensor can be converted into standardized data with zero mean value and unit variance.

[0052] Further, the production site data includes numerical production site data and image production site data; the extraction of the information features of the production site data includes the following steps:

[0053] S21, using a statistical feature extraction method to extract information features of the numerical production site data, and using principal component analysis to perform dimension reduction processing on the information features.

[0054] From the data after standardization processing, key information is extracted for feature selection, and features that have important influence on defect prediction are selected to reduce data dimension and improve processing efficiency.

[0055] Using a statistical feature extraction method, such as mean, variance, maximum value, minimum value, etc., to extract information features of the numerical production site data, and then using principal component analysis for dimension reduction processing to obtain representative features.

[0056] S22, extracting local features in the image production site data through a convolutional neural network, using multiple convolutional layers and pooling layers to extract high-level features layer by layer.

[0057] In some embodiments, the output of a convolutional layer or a pooling layer can be selected as features, which are usually high-dimensional local feature representations. In some other embodiments, the convolutional neural network can be trained from historical production data with labels, and then features can be extracted.

[0058] Further, the data of different modalities are input into respective fully connected layers for processing to obtain output representations of each modality, and then the output results of all modalities are weighted and summed or spliced to obtain final fusion data for training of a subsequent deep learning model.

[0059] Specifically, the information features are fused to obtain an information data feature set, which satisfies the following formula:

[0060]

[0061] wherein, represents the information data feature set, , represents a weight coefficient, represents an information feature of a numerical value type production site data, represents an information feature of an image type production site data.

[0062] Further, after obtaining the information data feature set, the information data feature set needs to be vectorized.

[0063] S3, constructing a joint deep learning prediction model, and training the joint deep learning prediction model by using the information data feature set.

[0064] In some embodiments, the joint deep learning prediction model includes a defect category prediction sub-model and a defect occurrence probability prediction sub-model.

[0065] Specifically, the defect category prediction sub-model adopts a convolutional neural network, which is suitable for learning spatial features of image data and extracting local features through a convolutional layer, and is suitable for a classification task of defect patterns; the defect occurrence probability prediction sub-model adopts a recurrent neural network or a long short-term memory network, which is suitable for processing of time series data and can capture time-dependent relationships and sequence patterns in data, and can obtain trend changes of time series data.

[0066] Further, the output layer activation function of the defect category prediction sub-model satisfies the following formula:

[0067]

[0068] wherein, represents an output result, represents a weight matrix of a fully connected layer, indicates the information data feature set vectorization result, indicates the bias term;

[0069] The output layer activation function of the defect occurrence probability prediction sub-model adopts a linear activation function.

[0070] Further, the training of the joint deep learning prediction model by using the information data feature set comprises the following steps:

[0071] S31, construct a defect category loss function and a defect occurrence probability loss function.

[0072] Specifically, the defect category loss function satisfies the following formula:

[0073]

[0074] Wherein, indicates the loss function value, indicates the number of samples, indicates the true label of the i-th sample, indicates the probability that the i-th sample is a defect category. The defect occurrence probability loss function satisfies the following formula:

[0075]

[0076]

[0077] Wherein, indicates the loss function value, indicates the number of samples, indicates the true value of the i-th sample, indicates the predicted value of the i-th sample.

[0078] S32, training the joint deep learning prediction model by using the loss function and the information data feature set.

[0079] In the embodiment, the training of the joint deep learning prediction model by using the loss function and the information data feature set satisfies the following formula:

[0080]

[0081] Wherein, indicates the model parameter updated by iteration, indicates the model parameter before iteration, indicates the learning rate, indicates the gradient of the loss function.

[0082] ​​​​Further, training the joint deep learning prediction model with the information data feature set requires dividing the training set and test set in a certain proportion. After training the model with the training set, the test set is used to calculate the accuracy, recall rate, F1 score and other indicators to verify and evaluate the prediction performance of the model.

[0083] S4, completing cloud manufacturing quality control and defect prediction by the trained joint deep learning prediction model.

[0084] In one embodiment, the trained joint deep learning prediction model is used to monitor the data changes in the production process in real time, especially the sensor data and camera image data. Real-time analysis and prediction of the data are performed according to the prediction results to trigger a real-time alarm mechanism and feedback to ensure that defects can be discovered in time and appropriate measures can be taken.

[0085] Specifically, one or more threshold values are set to determine the severity or probability of occurrence of defects. When the predicted defect probability exceeds the preset threshold value , an alarm will be triggered. The alarm can be sent in multiple ways, such as SMS, email, App push notification or system internal message, etc. The alarm content includes the defect type, defect location, predicted probability and suggested control measures.

[0086] Further, once the alarm is triggered, the production system will automatically execute the corresponding feedback measures according to the defect type and prediction results. The feedback measures include but are not limited to adjusting production parameters, enabling backup production lines, starting device self-checking programs, etc.

[0087] In the embodiment, a rule-based control algorithm or reinforcement learning algorithm is used to automatically adjust the production parameters according to the prediction results. For example, if it is predicted that high temperature leads to defects, the system automatically reduces the temperature setting value.

[0088] For multiple possible defects, the production system prioritizes the defects according to their severity and impact range to ensure that the most critical defects can be handled first. The alarm information and feedback measures are also displayed to the operator through the user interface, and the operator can take further manual intervention measures according to the suggestions. All alarms and feedback operations are recorded in the cloud manufacturing database for subsequent data analysis and model optimization, forming a closed-loop feedback mechanism. Further, specific repair suggestions can be generated according to the defect type and prediction results through the built-in knowledge base and rule engine, for example, different repair measures are recommended for different types of defects.

[0089] In yet another embodiment, the cloud database is also responsible for storing all data during quality control and defect prediction processes, including sensor data, image data, prediction results, etc. The database needs to update these data in real time to ensure the timeliness of the data, and manage the data to ensure its integrity and security. The specific steps include:

[0090] Data storage: all collected data, model prediction results, alarm records and feedback measures are stored in the cloud manufacturing database.

[0091] Data update: real-time reception and storage of new production data to ensure that the data in the database is up-to-date, and version control and data backup to prevent data loss and damage.

[0092] Data management: through the data management module, the stored data is organized, retrieved, backed up and restored to ensure efficient access and security of the data. Data management algorithm model:

[0093] Data indexing and retrieval: efficient database indexing algorithms such as B-tree or hash index are used to improve data retrieval speed.

[0094] Data backup and recovery: regular data backup and redundant storage technology are used to ensure quick recovery of data in case of failure.

[0095] Data security: encryption algorithms such as AES are used to encrypt stored data to ensure data confidentiality and integrity.

[0096] Model retraining: use the latest production data and feedback records to retrain the joint deep learning prediction model regularly to improve the prediction accuracy and adaptability of the model.

[0097] In an embodiment, the present application also provides a cloud manufacturing quality control and defect prediction device based on deep learning, which can stably execute the present application of a cloud manufacturing quality control and defect prediction method based on deep learning, including the following modules:

[0098] Data acquisition module: this module integrates sensors, cameras and other devices to collect various data in the manufacturing process in real time. Sensors collect environmental data such as temperature, humidity and pressure, and cameras capture image and video data during the production process. All collected data is transmitted to the cloud database processing submodule through the data upload submodule.

[0099] Multi-modal data fusion module: This module is responsible for fusing multi-modal data from different data sources (such as sensor data, image data, etc.), through standardization and normalization processing, so that data of different modalities can be analyzed under a unified standard, and their respective features can be extracted. Different modalities of data will be integrated in this module to provide more comprehensive and accurate data input.

[0100] Joint deep learning module: The joint deep learning module is the core of the device, which uses deep learning algorithms such as deep neural networks (such as convolutional neural networks CNN and recurrent neural networks RNN) to analyze and process multi-modal data. The platform can automatically learn and identify defect patterns in the data, and make defect predictions and quality control optimizations based on historical data, supporting real-time decision-making and adjustments during the manufacturing process.

[0101] Real-time alarm module: Based on the prediction results and real-time data generated by the deep learning analysis platform, the real-time alarm module is responsible for monitoring abnormal conditions during the manufacturing process. When the system detects quality problems or defect trends, the alarm module will trigger an alarm in a timely manner to remind operators to take necessary measures. Alarm information is transmitted to relevant personnel through visual, audio, or remote notification, ensuring that problems can be solved as soon as possible.

[0102] Cloud manufacturing database module: The cloud database is used to store and manage all data related to quality control and defect prediction, including sensor data, image data, prediction results, model training data, etc. The database is also responsible for regularly updating the stored data to ensure its timeliness, and ensuring the integrity, security, and reliability of the data.

[0103] The present application not only improves the accuracy of defect prediction, but also optimizes the quality control process through real-time early warning and automatic feedback mechanism, improving product quality and production efficiency.

[0104] Please refer to Figure 2 In order to efficiently execute the cloud manufacturing quality control and defect prediction method based on deep learning provided by the present application, the present application also provides a cloud manufacturing quality control and defect prediction system based on deep learning, which includes input device, output device, processor, memory, the input device, output device, processor, memory are connected with each other, the memory contains program instructions, the program instructions are used for the steps of the cloud manufacturing quality control and defect prediction method based on deep learning. The cloud manufacturing quality control and defect prediction system based on deep learning of the present application has compact structure and stable performance, can stably execute the cloud manufacturing quality control and defect prediction method based on deep learning of the present application, further improves the overall applicability and practical application ability of the present application.

[0105] In embodiments, the processor can be a central processing unit (CPU), the processor can also be other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The input device can be used to obtain data information. The output device can be used to output the results obtained by the program instructions contained in the computer program stored in the memory provided by the application. The memory can include read-only memory and random access memory, and provide instructions and data to the processor. Part of the memory can also include non-volatile random access memory.

[0106] In yet another optional embodiment, referring to Figure 3 In order to efficiently implement the cloud manufacturing quality control and defect prediction method provided by the application, the embodiment also provides a cloud manufacturing quality control and defect prediction device based on deep learning, as shown in Figure 3 The device comprises:

[0107] The memory 10 is used to store a computer program; the processor 20 is used to execute the computer program to realize the cloud manufacturing quality control and defect prediction method based on deep learning. The memory 10, the processor 20, the communication interface 31 and the communication bus 32. The memory 10, the processor 20, the communication interface 31 all complete the communication among each other through the communication bus 32.

[0108] In embodiments, the memory 10 is used to store one or more program instructions, and the memory 10 can store program instructions for realizing the following functions:

[0109] Through the data acquisition device, the production site data in the manufacturing process is obtained and uploaded to the cloud for further analysis; the information features of the production site data are extracted and fused to obtain an information data feature set; a joint deep learning prediction model is constructed, and the joint deep learning prediction model is trained using the information data feature set; and the cloud manufacturing quality control and defect prediction are completed through the trained joint deep learning prediction model.

[0110] In a possible implementation, the memory 10 can include a program storage area and a data storage area, where the program storage area can store an operating system and applications required by at least one function and the like, and the data storage area can store data created during use. In addition, the memory 10 can include a read-only memory and a random access memory, and provide instructions and data for the processor. A part of the memory can also include an NVRAM. The memory stores an operating system and operation instructions, executable modules or data structures, or a subset thereof, or an extended set thereof, where the operation instructions can include various operation instructions for implementing various operations. The operating system can include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0111] The processor 20 can be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic device, which can be a microprocessor or any conventional processor. The processor 20 can invoke a program stored in the memory 10. The communication interface 31 can be an interface of a communication module for connecting with other devices or systems.

[0112] Of course, it should be noted that, Figure 3 The structures shown do not constitute a limitation on the cloud manufacturing quality control and defect prediction device based on deep learning in the embodiment, and in actual applications, the cloud manufacturing quality control and defect prediction device based on deep learning can include more or fewer components than Figure 3 those shown, or combine certain components.

[0113] The embodiment also provides a storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the steps of the cloud manufacturing quality control and defect prediction method based on deep learning.

[0114] The storage medium can include a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0115] To sum up, the present application extracts various features of production site data, re-trains a joint deep learning prediction model, and then completes cloud manufacturing quality control and defect prediction, solves the quality problem and defect prediction problem in the manufacturing process, effectively improves the production efficiency, optimizes the product quality, and promotes the intelligentization and automation upgrade of the manufacturing process.

[0116] Therefore, the application effectively overcomes the shortcomings in the prior art and has high industrial utilization value.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the application, and are not intended to limit the application; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the application, and they should be covered in the scope recorded in the application.

Claims

1. A cloud manufacturing quality control and defect prediction method based on deep learning, characterized in that, The method comprises the following steps: Through a data acquisition device, production site data in a manufacturing process are acquired and uploaded to a cloud for next step analysis; Information features of the production site data are extracted and fused to obtain an information data feature set; A joint deep learning prediction model is constructed, and the joint deep learning prediction model is trained by using the information data feature set; Cloud manufacturing quality control and defect prediction are completed by using the trained joint deep learning prediction model; The production site data comprise numerical production site data and image production site data; The information features of the production site data are extracted by the following steps: Statistical feature extraction is used to extract information features of the numerical production site data, and principal component analysis is used to reduce the dimension of the information features; Local features in the image production site data are extracted by using a convolutional neural network, and high-level features are extracted layer by layer by using multiple convolutional layers and pooling layers; The information features are fused to obtain the information data feature set, and the following formula is satisfied: wherein, representing a set of information data features, , representing a weight coefficient, representing an information feature of a numerical value type production site data, representing an information feature of an image type production site data; The joint deep learning prediction model comprises a defect category prediction submodel and a defect occurrence probability prediction submodel; An output layer activation function of the defect category prediction submodel satisfies the following formula: wherein, represents an output result, represents a weight matrix of a full connection layer, represents a feature set vectorization result of information data, represents a bias term; An output layer activation function of the defect occurrence probability prediction submodel adopts a linear activation function; The joint deep learning prediction model is trained by using the information data feature set by the following steps: A defect category loss function and a defect occurrence probability loss function are constructed; The joint deep learning prediction model is trained by using the loss functions and the information data feature set; The defect category loss function satisfies the following formula: in, This represents the value of the loss function. Indicates the number of samples. Indicates the first The true label of each sample Indicates the first The probability that each sample belongs to the defect category; The defect occurrence probability loss function satisfies the following formula: in, This represents the value of the loss function. Indicates the number of samples. Indicates the first The true value of each sample Indicates the first Predicted values ​​for each sample; The joint deep learning prediction model is trained by using the loss functions and the information data feature set, and the following formula is satisfied: wherein, denotes the iteratively updated model parameters, denotes the model parameters before the iteration update, denotes the learning rate, denotes the gradient of the loss function. 2.The cloud manufacturing quality control and defect prediction method based on deep learning according to claim 1, wherein, The production site data are also standardized; The production site data are standardized, and the following formula is satisfied: wherein, represents a value after standardization processing of numerical data, represents numerical data, represents a mean value, represents a standard deviation, represents a normalized value of a standardized image, represents a pixel value of an original image.

3. A cloud manufacturing quality control and defect prediction system based on deep learning, characterized in that, The cloud manufacturing quality control and defect prediction system based on deep learning comprises an input device, an output device, a processor and a memory, the input device, the output device, the processor and the memory are connected to each other, the memory comprises program instructions, and the program instructions are used to execute the cloud manufacturing quality control and defect prediction method based on deep learning in any one of claims 1-2.

Images