CN117828499B

CN117828499B - PCBA abnormal part determination method, system, storage medium and electronic equipment

Info

Publication number: CN117828499B
Application number: CN202410241041.3A
Authority: CN
Inventors: 陈茂; 黄勇庆; 谢永智
Original assignee: Shenzhen Tianyi Electronics Co ltd
Current assignee: Shenzhen Tianyi Electronics Co ltd
Priority date: 2024-03-04
Filing date: 2024-03-04
Publication date: 2024-05-28
Anticipated expiration: 2044-03-04
Also published as: CN117828499A

Abstract

A PCBA abnormal part determining method, a system, a storage medium and electronic equipment relate to the technical field of electronics. The method comprises the following steps: acquiring specific conditions when the target client applies the PCBA, and sampling and detecting at least one target PCBA in the batch PCBA according to the specific conditions; when determining that the abnormal part exists in the target PCBA, determining a second production condition corresponding to the specific condition in the first production conditions of the batch PCBA; extracting features of the first production condition and the second production condition to obtain target features; and inputting the target features into the recognition model after training, and outputting the target abnormal parts in the batch PCBA. By adopting the method, the accuracy of identifying the abnormal PCBA parts can be improved.

Description

PCBA abnormal part determination method, system, storage medium and electronic equipment

Technical Field

The application relates to the technical field of electronics, in particular to a method and a system for determining abnormal PCBA parts, a storage medium and electronic equipment.

Background

The printed circuit board (printed circuit board assembly, PCBA) is a circuit board formed by mounting electronic components on a printed circuit board, and is a core component of an electronic product. The quality of PCBA directly affects the performance and reliability of electronic products. However, various problems, such as device missing, position dislocation, direction error, poor welding, etc., are inevitably caused in the PCBA manufacturing process, and the functional defects or potential safety hazards of the finished product are caused.

PCBA is widely applied to various electronic products, and PCBA produced in the same batch can be transported to different clients for application in different scenes. Different application environments can have different degrees of impact on the PCBA. The detection of the PCBA is usually carried out only in a standard environment, and the complex application condition of the PCBA in actual working is difficult to simulate. Therefore, it is difficult to comprehensively and accurately judge PCBA anomalies that may occur under specific conditions in a batch of PCBAs only by detection under standard environments.

Disclosure of Invention

The application provides a method, a system, a storage medium and electronic equipment for determining abnormal PCBA parts, which can improve the accuracy of identifying the abnormal PCBA parts.

In a first aspect of the present application, the present application provides a PCBA abnormal piece determination method, comprising:

Acquiring specific conditions when the target client applies the PCBA, and sampling and detecting at least one target PCBA in the batch PCBA according to the specific conditions;

When determining that the abnormal part exists in the target PCBA, determining a second production condition corresponding to the specific condition in the first production conditions of the batch PCBA;

Extracting features of the first production condition and the second production condition to obtain target features;

and inputting the target features into the recognition model after training, and outputting the target abnormal parts in the batch PCBA.

By adopting the technical scheme, the target PCBA is sampled and detected based on the specific conditions by acquiring the specific conditions of the target customer application PCBA, so that personalized and accurate anomaly detection aiming at different customer demands is realized, the detection result is more in line with the actual use environment of the target PCBA, and the omission caused by simple universal standard conditions is avoided. After the abnormal part is determined, the first standard production condition and the second corresponding abnormal condition are further analyzed, the source of the abnormal occurrence is found, the first condition can be improved in a targeted manner, the production flow is optimized, and the problem is prevented from repeatedly occurring. And extracting various characteristics of the first condition and the second condition, marking and fusing, and obtaining an accurate model for distinguishing the normal and the abnormal through sample training. The model comprehensively utilizes different characteristics, and realizes the function of accurately distinguishing abnormal parts. The method can improve the accuracy of identifying the abnormal PCBA.

Optionally, the feature extracting the first production condition and the second production condition to obtain the target feature includes:

determining a condition difference value between the specific condition and a standard test condition;

Based on the condition difference value, adjusting the weight ratio of each production condition in the first production conditions to obtain adjusted first production conditions;

extracting the characteristics of the adjusted first production condition to obtain first characteristics;

And extracting the characteristics of the second production condition to obtain second characteristics.

By adopting the technical scheme, the difference value represents the importance of the parameter to the abnormality identification according to the difference between the specific condition and the standard condition. And carrying out difference analysis and adjusting the weight of the first condition feature according to the difference analysis, so that the extracted feature is focused on the production parameter which is more critical to the abnormal identification. By extracting the first condition features after weight adjustment and extracting the second abnormal corresponding condition features, two types of features which can represent different production links are obtained. These features may more fully reflect the cause of the anomaly generation. The first characteristic and the second characteristic are extracted respectively and then fused, and the parameter information of different production conditions is comprehensively utilized, so that the model is trained more fully, and the abnormality judgment can be performed more accurately.

Optionally, the target feature includes a first feature corresponding to a first production condition and a second feature corresponding to a second production condition, the inputting the target feature into the trained recognition model, outputting a target anomaly in the batch PCBA, including:

Inputting the first characteristic into a recognition model after training is completed, and outputting a first abnormal part;

inputting the second characteristic into the recognition model after training is completed, and outputting a second abnormal part;

and determining the common abnormal piece in the first abnormal piece and the second abnormal piece as the target abnormal piece.

By adopting the technical scheme, the influence of the specific condition on the product quality can be reflected, and the weight of each parameter in the first condition is adjusted accordingly. The first condition after adjustment can extract the characteristic which has important meaning on the abnormal recognition more accurately. Simultaneously, two types of features corresponding to the first condition and the second condition are respectively extracted and used as the first feature and the second feature. And respectively inputting different features into the recognition model, carrying out two-stage recognition, and taking the intersection to judge the final abnormal part. The method for staged identification and multi-feature fusion can fully utilize the feature information extracted from different angles of the first condition and the second condition, and improves the identification accuracy.

Optionally, before the specific condition when the PCBA is applied to the target client is obtained, the method further includes:

Acquiring normal production conditions of normal PCBA and abnormal production conditions of abnormal PCBA;

Marking the normal production condition and the abnormal production condition to obtain a marked condition;

Performing characteristic extraction operation on the normal production condition and the abnormal production condition respectively to obtain different sample characteristics;

training each decision tree in the initial recognition model by adopting the different sample characteristics to obtain a corresponding training result;

and combining the training results to obtain the recognition model after training.

By adopting the technical scheme, the method carries out labeling and feature extraction by acquiring sufficient data containing normal and abnormal samples, and provides a proper sample set for model training. The decision trees are trained by adopting different sample characteristics, so that each decision tree can learn rules for accurately distinguishing a certain type of sample, and the distinguishing capability of the model is improved. According to the scheme, accurate and comprehensive model training flow is realized through technical means such as obtaining of marking data, feature training and result fusion. The obtained training completion model can accurately and stably judge the abnormality, improves the judging effect, provides a reliable basis for subsequent abnormality identification, and has stronger adaptability and generalization capability.

Optionally, the sample features include: material batch information, production process information, staff information, tool equipment information used and production environment information.

By adopting the technical scheme, the sample characteristics provided by the scheme comprise information of a plurality of aspects such as material batch, process flow, personnel operation, equipment state, environmental conditions and the like. These features comprehensively reflect various influencing factors from raw materials to operation flow, to the whole production process of equipment, environment and the like. The sample data containing the multidimensional features can enable the model to fully learn the comprehensive influence of different factors on the product quality.

Optionally, the training results include a training decision tree, and combining the training results to obtain a training feature recognition model, including:

fitting each trained decision tree through a soft voting mechanism to obtain the trained recognition model.

By adopting the technical scheme, a weighted fusion method of soft voting is adopted, and a plurality of trained decision tree models are combined. The soft voting can reasonably preset the weight according to the verification performance of the single model. And carrying out weighted accumulation judgment during prediction. The knowledge learned by each decision tree is exerted, and the action degree of different models is controlled through weights.

Optionally, the sample features further include an out-bag sample, and after training each decision tree in the initial recognition model by using the different sample features to obtain a corresponding training result, the method further includes:

Determining the accuracy of the current identification model through the sample set outside the bag;

determining the recall rate of the training result according to the correct sample number of the training result;

and adjusting parameters of the identification model according to the accuracy rate and the recall rate.

By adopting the technical scheme, a new sample outside the bag is introduced, the model is predicted, and the accuracy is calculated. The accuracy can evaluate the prediction effect of the model on new data, and the over-fitting problem is found. And meanwhile, the recall rate is calculated, so that the model can effectively identify the marked positive sample. The accuracy and recall rate together judge the prediction quality of the model. And the model is subjected to multi-round parameter adjustment according to the two indexes, so that accurate judgment is ensured, and the report missing rate is controlled. The model performance is evaluated more comprehensively and accurately.

In a second aspect of the present application, there is provided a PCBA anomaly determination system, the system comprising:

The specific condition determining module is used for acquiring specific conditions when the PCBA is applied to the target clients and carrying out sampling detection on at least one target PCBA in the batch PCBA according to the specific conditions;

The production condition determining module is used for determining a second production condition corresponding to the specific condition in the first production conditions of the batch PCBA when determining that the target PCBA has the abnormal component;

the target feature determining module is used for extracting features of the first production condition and the second production condition to obtain target features;

And the target abnormal part output module is used for inputting the target characteristics into the recognition model after training is completed and outputting the target abnormal parts in the batch PCBA.

In a third aspect the application provides a computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the above-described method steps.

In a fourth aspect of the application there is provided an electronic device comprising: a processor, a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the above-mentioned method steps.

In summary, one or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:

By adopting the technical scheme of the application, the target PCBA is sampled and detected based on the specific conditions by acquiring the specific conditions of the target customer application PCBA, so that personalized and accurate anomaly detection aiming at different customer demands is realized, the detection result better accords with the actual use environment of the target PCBA, and the omission caused by simple universal standard conditions is avoided. After the abnormal part is determined, the first standard production condition and the second corresponding abnormal condition are further analyzed, the source of the abnormal occurrence is found, the first condition can be improved in a targeted manner, the production flow is optimized, and the problem is prevented from repeatedly occurring. And extracting various characteristics of the first condition and the second condition, marking and fusing, and obtaining an accurate model for distinguishing the normal and the abnormal through sample training. The model comprehensively utilizes different characteristics, and realizes the function of accurately distinguishing abnormal parts. The method can improve the accuracy of identifying the abnormal PCBA.

Drawings

FIG. 1 is a schematic flow chart of a method for determining abnormal PCBA parts provided by an embodiment of the application;

FIG. 2 is a schematic diagram of a PCBA anomaly determination system provided by an embodiment of the application;

fig. 3 is a schematic structural diagram of an electronic device according to the disclosure.

Reference numerals illustrate: 300. an electronic device; 301. a processor; 302. a communication bus; 303. a user interface; 304. a network interface; 305. a memory.

Detailed Description

In order that those skilled in the art will better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments.

In describing embodiments of the present application, words such as "for example" or "for example" are used to mean serving as examples, illustrations, or descriptions. Any embodiment or design described herein as "such as" or "for example" in embodiments of the application should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "or" for example "is intended to present related concepts in a concrete fashion.

In the description of embodiments of the application, the term "plurality" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating an indicated technical feature. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.

With miniaturization and functional complexity of electronic products, the mounting density of components on a Printed Circuit Board (PCB) is increasing, and the manufacturing difficulty of the PCBA process is greatly increased. Problems that readily occur during PCBA manufacturing include: misalignment, missing, cold solder joint, excessive solder paste, etc., of the device, which if not effectively detected and controlled, can directly lead to product failures and differences. However, the traditional PCBA testing method is limited to detection under a standard environment, and is difficult to adapt to complex working conditions under different client application environments, so that the reliability of the PCBA is comprehensively evaluated.

Specifically, the application environments of different customers may have differences in temperature, humidity, vibration, etc., which may have different effects on the quality of the PCBA. For example, the aging process of PCBA devices and welding spots can be accelerated under high temperature conditions, short circuit faults can be caused by an excessively humid environment, and joint loosening is easy to cause due to strong vibration. But standard static detection makes it difficult to verify these potential failure modes. In order to reduce product rework rate and quality risk, a PCBA anomaly detection method that can take into account specific customer application conditions is needed.

Based on this, an embodiment of the present application provides a method for determining abnormal PCBA, in an embodiment, please refer to fig. 1, fig. 1 is a flow chart of the method for determining abnormal PCBA provided by the embodiment of the present application, and the method may be implemented by a computer program, may be implemented by a single chip microcomputer, or may be run on a PCBA abnormal determining system based on von neumann system. The computer program may be integrated in the application or may run as a stand-alone tool class application. Specifically, the method may include the steps of:

Step 101: and acquiring specific conditions when the PCBA is applied by the target client, and sampling and detecting at least one target PCBA in the batch PCBA according to the specific conditions.

The specific conditions refer to parameter requirements of target clients on the actual use environment of the PCBA, and mainly comprise temperature conditions, humidity conditions, mechanical conditions and the like.

Temperature conditions: the temperature environment range which the PCBA needs to adapt to is reflected by parameters such as the highest use temperature, the lowest use temperature and the like.

Humidity conditions: the environment parameters of the humidity which the PCBA needs to endure are reflected by parameters such as the normal working humidity, the short-time tolerance humidity and the like.

Mechanical conditions: the PCBA is composed of vibration parameters, impact parameters, free falling parameters and the like, and reflects the mechanical environmental effect required to be born by the PCBA in the transportation and use processes.

The specific conditions are determined by the target client according to the actual use scene of the product, and compared with the general product quality standard, the specific conditions can simulate the actual working environment of the PCBA more accurately. The potential problems of the PCBA in the use environment can be effectively discovered by acquiring and detecting according to the specific conditions.

Specifically, because of the differences of specific conditions such as temperature, humidity and the like in the PCBA application environments of different clients, specific condition parameters of target clients need to be acquired in order to accurately evaluate the reliability and fault risk of the PCBA under the actual use scene. The specific method is that a quality engineer knows the application scene of the client product through visit investigation and obtains the specific condition of the client product.

After acquiring specific conditions, the engineer inputs the parameters into the detection device to configure a matched test scheme. In the PCBA batch production process, sample plates are extracted from a conveying belt and loaded into test equipment, and the PCBA is dynamically detected by automatically simulating specific conditions of a customer by adopting means of high-low temperature cycle test, temperature-humidity combined test and the like. The test mode which is matched with the actual use scene can efficiently find out the fault points in the product, such as the problems of welding spot cracking, loose line and the like, and the effective rate can be improved by 30%. After the detection is finished, a detection report is generated, the problem plate is marked, and the production process is fed back, so that defect batch is avoided.

Step 102: when it is determined that the target PCBA exists in the abnormal part, determining a second production condition corresponding to the specific condition in the first production conditions of the batch PCBA.

Wherein, the first production condition refers to standard production parameter conditions in the PCBA manufacturing process. Illustratively, the first production condition may include: the processing technology conditions are as follows: SMT process parameters, component placement parameters, reflow parameters, etc.; equipment conditions: performance state of SMT mounter, performance state of reflow oven, etc.; operator conditions: operator skill level, degree of operational specification, etc.; factory environmental conditions: antistatic conditions, cleanliness conditions, etc. The first manufacturing conditions determine the normal manufacturing process flow and control criteria for the PCBA.

The second production condition refers to a corresponding relation between the abnormality found according to the specific condition detection and the specific production parameter based on the first production condition, and comprises the following steps: process related parameters: the baking temperature is higher, the reflow time is insufficient, and the like; device-related parameters: the conveyor belt has too high speed, too high press-fitting force and the like; operator related parameters: operational non-norms result in component position offset, etc. The second production condition is a direct cause condition that causes PCBA abnormality. By determining the second condition, the source of the abnormal generation can be found out, and the targeted improvement can be performed.

When an abnormality is found to exist in the target PCBA by specific condition detection, the cause of the abnormality generation needs to be further analyzed for subsequent improvement. For this reason, the whole production process of the batch of PCBA needs to be traced back, and the first standard production condition parameters are determined.

The first production condition is used as a standard parameter for normal production of PCBA, and covers the whole manufacturing flow. Taking an SMT process as an example, the first condition includes parameters such as temperature, time, pressure, etc. of each process of the SMT, and information such as a performance state of the device and a skill level of an operator. These parameters determine the product quality level.

And then, according to the detected abnormal condition, judging a key second production condition causing the abnormality by combining the corresponding relation between the first condition and the specific condition. For example, a shorter reflow time may lead to a cold solder joint phenomenon, which corresponds exactly to the risk of solder joint burn-in high temperature environments in certain conditions.

The purpose of determining the second condition is to find the root cause of the anomaly generation for targeted improvement. In the above example, after determining that the reflow time is shorter is the second condition that directly leads to the dummy solder fault, the reflow oven time can be increased, and the high-temperature environment fault rate can be reduced.

By analyzing the second condition, the first condition can be continuously optimized to adapt to the requirement of the specific condition, and the PCBA product quality is improved. Meanwhile, the second condition provides a basis for analysis of reasons, so that the reworking cost is reduced.

Step 103: and extracting the characteristics of the first production condition and the second production condition to obtain target characteristics.

After the first standard production condition and the second abnormal corresponding condition are determined, the characteristic information of the first standard production condition and the second abnormal corresponding condition needs to be extracted to establish an abnormal identification model.

The feature extraction is to select key features related to abnormality recognition from a large number of first conditions and second conditions, and construct effective model input. For example, the characteristics of temperature and humidity parameters, equipment debugging states and the like of the SMT process can be extracted from the first condition, and the characteristics of insufficient reflow time and the like corresponding to the cold joint can be extracted from the second condition.

Feature extraction is realized by a program, the input is detailed parameter detection data of the first condition and the second condition, and the output is a feature vector of the first condition and the second condition. Wherein, a certain characteristic is given different weights to characterize the importance of the characteristic to abnormality recognition. There are various feature processing methods, for example, a principal component analysis method can automatically learn feature weights.

After feature vectors of two types of conditions are obtained, the feature vectors are combined into target features. The target feature fully represents the associated information of the first condition and the second condition and comprises key indexes for distinguishing the abnormality. The automatic identification of the anomalies in the batch PCBA can be realized by inputting the target features. Feature extraction is the basis for ensuring recognition effect.

Based on the above embodiment, as a possible implementation manner, in step 103: the step of extracting features of the first production condition and the second production condition to obtain the target feature may specifically further include the following steps:

step 201: a condition difference is determined for the particular condition from the standard test condition.

Specifically, in extracting the features of the first condition and the second condition, it is also necessary to consider the difference of the specific condition from the standard condition.

The standard condition is a universal PCBA test standard, and is not specific to a certain use environment. While specific conditions are focused on the specific usage scenario of the customer. There is a certain difference in conditions between the two.

To reflect the effect of this difference on PCBA quality, specific numerical differences in temperature, humidity, etc. of the two are determined in step 201. This is achieved by comparing the two types of conditions item by item. The purpose of determining the difference in conditions is to subsequently adjust the weights of the parameters in the first condition. Because the specific condition is the actual usage requirement of the customer, its difference from the standard condition represents which parameters are more important. For example, if the highest temperature in a particular condition is 10 degrees higher than the standard condition, the weight of the baking temperature parameter in the first condition may be increased. This means that the temperature has a greater impact on the fault and should be taken as a more important feature.

Step 202: and adjusting the weight ratio of each production condition in the first production condition based on the condition difference value to obtain the adjusted first production condition.

After determining the difference value of the two types of conditions, the weight of each parameter in the first condition needs to be adjusted based on the difference value, so as to obtain the adjusted first condition characteristic. The weight adjustment is required because the magnitude of the difference represents the importance of the parameter to fault identification. If the temperature difference is large, the weight of the temperature characteristic should be enhanced.

Specifically, a correspondence model of the difference and the weight may be established. The larger the difference, the more its weight is lifted, the higher the duty cycle in the first conditional feature vector. Through weight adjustment, the extracted first condition features are focused on important parameters, and fault modes corresponding to specific conditions can be identified in a targeted manner.

Step 203: and extracting the characteristics of the adjusted first production conditions to obtain first characteristics.

Step 204: and extracting the characteristics of the second production condition to obtain second characteristics.

Specifically, the principle of the feature extraction method is the same as that of the feature extraction method in step 103, and the specific process can refer to step 103, and will not be described in detail herein.

Step 104: and inputting the target characteristics into the trained recognition model, and outputting target abnormal parts in the PCBA in batches.

Specifically, after the first condition feature and the second condition feature are acquired and combined into the target feature, the abnormal part identification in the batch PCBA needs to be realized based on the target feature.

Specifically, a trained recognition model is already obtained at this time. The model is obtained by marking a large amount of normal and abnormal condition data in advance, extracting characteristics and performing repeated training and optimization.

Then, only the target features obtained in the previous step need be directly fed as input into the recognition model after training. The recognition model automatically analyzes the features, judges the input target features according to the key production condition information reflected in the features and the knowledge learned by the model, and judges whether the corresponding PCBA is abnormal or not.

The final model outputs an anomaly prediction result to identify the target PCBA which is judged to be an anomaly. Thus, the abnormal identification and screening of the whole batch of products are completed.

Based on the above embodiment, as an alternative embodiment, in step 104: inputting the target features into the trained recognition model, outputting target anomalies in the batch PCBA, comprising:

Step 301: and inputting the first characteristic into the recognition model after training, and outputting the first abnormal part.

Step 302: and inputting the second characteristic into the recognition model after training, and outputting a second abnormal part.

Step 303: and determining the common abnormal piece in the first abnormal piece and the second abnormal piece as a target abnormal piece.

First, the first condition features which are extracted and obtained, namely the production condition features reflecting the specific environment of the client, are input into the recognition model which is trained and then predicted. After the model analyzes these features, the PCBA judged to be abnormal is output as the first abnormal piece. This step is referred to as "first feature recognition". The first feature represents the specific environmental conditions specified by the target customer, and the input of these features can find abnormal items that do not meet the customer's requirements.

And then, the extracted second condition features, namely the condition features reflecting the general production process, are input into the same identification model to obtain a second abnormal part judged by the model. The second feature represents the general production conditions, and the model can identify anomalies that do not conform to the standard process flow by inputting these features.

And finally, taking an intersection of the identification results of the two steps, namely, jointly judging the PCBA as abnormal in two times in sequence, and taking the PCBA as a final target abnormal part.

The purpose of feature fusion is that different features represent different aspects of conditions, and commonly judged abnormal parts are most likely to have problems and need to be processed preferentially. By using the first and second characteristics to identify the abnormality and then taking the intersection as the target abnormality, the implementation scheme can reasonably utilize the information of different characteristics, so that the abnormality identification is more accurate, and the problem product which needs to be optimized is found, thereby continuously improving the production process and improving the product quality and the production efficiency.

The foregoing embodiment describes a method for determining abnormal PCBA parts, and based on the foregoing embodiment, a training process of an identification model will be described, where the process may include the following steps:

step 401: and obtaining the normal production conditions of the normal PCBA and the abnormal production conditions of the abnormal PCBA.

In order to train a model that effectively identifies PCBA anomalies, a dataset containing normal and abnormal samples needs to be prepared.

Specifically, first, various standard production condition data during normal PCBA production needs to be collected as a normal sample, which is called "obtaining normal production conditions". Obtaining sufficient normal samples is necessary because they contain the characteristics of the product under quality-controlled conditions, and the model needs to build normal discriminant criteria by learning these characteristics.

Meanwhile, production conditions of the PCBA with known faults also need to be collected as abnormal samples, which is referred to as "obtaining abnormal production conditions". Abnormal samples are also important because the model must learn the data laws of these conditions in order to subsequently identify different fault conditions.

After the sample data of normal and abnormal production conditions are obtained, a training set containing two types of sample characteristics is constructed. This lays a data foundation for subsequent model training. By obtaining sample data which is sufficient and labeled with categories, the model can learn the knowledge and the discriminant rules required for identifying abnormal conditions.

Step 402: and (3) marking the normal production condition and the abnormal production condition to obtain a marked condition.

After the data sets of both normal and abnormal PCBA production conditions are obtained, these samples need to be labeled for model learning.

Specifically, the collected normal production condition samples are first labeled, and their "normal" category labels are designated. This portion of the data can be modeled to represent normal product production conditions. Clear category labeling is a precondition for successful training of subsequent models.

At the same time, it is also necessary to add an "abnormal" category label to the sample of abnormal production conditions. The model can be made to make sure that these data correspond to fault conditions and understand different condition patterns.

The labeled dataset provides the basis for training of the model. The model can learn data distribution and characteristics of different categories through labeling information. Finally, model training is carried out based on the marked data set, so that a classifier capable of effectively judging normal and abnormal conditions can be obtained, and recognition and prediction of PCBA abnormality can be completed.

Step 403: and respectively carrying out feature extraction operation on the normal production condition and the abnormal production condition to obtain different sample features.

After labeling the collected normal and abnormal production condition sample data sets, further extraction of data features of the two types of samples is needed to facilitate model training.

Specifically, first, feature engineering is performed on the marked normal production conditions, and feature vectors capable of representing normal sample distribution are obtained by using a feature extraction method. The feature extraction of the normal sample can enable the model to learn key features which can score normal conditions most, and a basis for judging the normal is established.

Meanwhile, feature extraction is required to be performed on a sample data set marked as an abnormal category to obtain feature vectors representing abnormal distribution. The feature extraction of the abnormal sample aims to provide feature input required by model learning to judge abnormal conditions.

Through the steps, two groups of characteristics can be obtained, one group represents normal production conditions, and the other group represents abnormal conditions, so that a characteristic set required by model training is formed. Subsequently, the model may determine whether the input data is normal or abnormal based on differences in these features. Compared with the original condition data, the feature vector constructed through feature extraction can be better distinguished by the model from the intrinsic modes of different conditions. Therefore, the feature extraction is respectively carried out on the data sets of the two types of samples, so that more representative and distinguishing features can be obtained, and the discrimination capability and accuracy of the follow-up model are improved.

In one possible embodiment, the sample characteristics include material lot information, production process information, staff information, tool equipment used information, and production environment information.

Extracting multifaceted information capable of reflecting the whole production process as characteristics may specifically include: material batch information reflecting performance parameters of used components, production process information reflecting condition parameters of each manufacturing step, staff information reflecting operation normalization of operators, using tool equipment information reflecting equipment states and performances, production environment information reflecting environmental conditions such as static electricity, temperature and humidity and the like.

By integrating all the information as the characteristics, the model can fully learn the comprehensive influence of different factors on normal and abnormal conditions, and the recognition accuracy is improved.

Step 404: and training each decision tree in the initial recognition model by adopting different sample characteristics to obtain a corresponding training result.

First, an integrated model containing a plurality of decision trees is constructed as an initial recognition model. The decision tree is chosen because it can represent the feature distribution of the sample and also facilitate interpretation of the reasons for model judgment. Then, using the extracted characteristics of the normal production conditions to train each decision tree in the initial model respectively to obtain a set of training results under the normal conditions, and each decision tree can learn the discriminant rule describing the normal condition.

Simultaneously, the characteristics of the abnormal samples are used for respectively training each decision tree in the model, and a training result under a group of abnormal conditions is obtained. This may enable each decision tree to learn to judge anomalies based on features.

After each decision tree is trained for each feature, a set of normal discrimination results and a set of abnormal discrimination results can be obtained. The results comprise knowledge of each decision tree based on corresponding characteristics, and a foundation is laid for a subsequent integrated model. Through training respectively, the model can fully learn the discrimination rules under different characteristic conditions, and the comprehensive capability of discriminating normal and abnormal conditions is improved.

On the basis of the above embodiment, as an alternative embodiment, the sample feature further includes an out-of-bag sample, in step 404: training each decision tree in the initial recognition model by adopting different sample characteristics, and after obtaining a corresponding training result, further comprising the following steps:

step 501: and determining the accuracy of the current identification model through the sample set outside the bag.

First, a certain amount of new normal and abnormal condition samples are prepared as an out-of-bag sample set. The new sample is introduced to more objectively evaluate the discrimination effect of the model in practical application.

And then, sequentially inputting the data in the sample set outside the bag into a currently trained recognition model to judge the model. And calculating the accuracy according to the judging result of the model and the actual labeling of the sample. For example, the accuracy is obtained by determining the ratio of the number of correct samples to the total number.

The accuracy can evaluate how the discrimination performance of the current model is on the new sample, and whether the over-fitting phenomenon exists or not. By introducing a new sample set test and calculating an accuracy index, the effect of the model in practical application can be more objectively and comprehensively judged, and the optimization model is adjusted in a targeted manner, so that the generalization capability of the model is improved.

Step 502: and determining the recall rate of the training result according to the correct sample number of the training result.

First, the number of abnormal conditions in the training process, all marked as positive samples, is counted. Then, when the model predicts a new sample set, the number of positive sample anomalies predicted is counted. The model recall is then obtained by dividing the number of positive samples predicted by the total number of training samples. The recall reflects the proportion of positive samples that the model can correctly identify to all positive samples.

The purpose of calculating the recall rate is to judge whether the model can effectively judge the marked abnormality correctly, so as to avoid the condition of missing report. The comprehensive performance of the model can be comprehensively evaluated by combining two indexes of accuracy and recall. If the recall rate is too low, the model needs to be adjusted to prevent the abnormal condition of too many missed reports. The two indexes are integrated, so that the model has stronger reproduction capability and provides more reliable abnormality detection results on the premise of ensuring accurate judgment.

Step 503: and adjusting parameters of the identification model according to the accuracy and the recall rate.

First, it is determined whether the values of the two indexes meet the requirements. E.g., whether the accuracy has reached a desired threshold, and whether the recall is too low. And then, according to the analysis result, determining parameter items and adjustment directions which need to be adjusted. For example, the complexity of the model and the number of training rounds are increased to promote two indexes. Then, according to the content determined in the previous step, relevant parameters in the identification model are updated. And finally, retraining the model by using the adjusted parameters and re-evaluating the indexes.

If the index does not reach the standard, the analysis and adjustment are needed to be continued until the comprehensive performance of the model meets the requirements. Through repeated iterative optimization in the process, the performance of the model on the overall accuracy and recall rate can be continuously improved. Therefore, the final model can ensure higher accurate discrimination and can effectively avoid the abnormal condition of missing report.

Step 405: and combining the training results to obtain the recognition model after training.

In one possible implementation, each trained decision tree may be fitted by a mechanism of soft voting to obtain a trained recognition model.

Specifically, first, initializing weight setting is performed on a plurality of decision tree models after training is completed. The weights may be predetermined based on the verification performance of the model, with better performing models setting higher weights. And then, inputting the new sample characteristics into each decision tree model in turn, distinguishing each model, and outputting a classification result and confidence coefficient. And then, carrying out weighted accumulation on the classification results of all the decision trees according to preset weights, and outputting a final judgment result through setting a threshold value. And finally, continuously adjusting the weight and the fusion threshold value of the single model, and re-evaluating the effect of the fusion model on the new sample.

The weight parameters are repeatedly adjusted, so that the optimal setting of the combined model can be obtained, and the aim of integrating the advantages of a plurality of decision trees is fulfilled. The soft voting mode can fully utilize the knowledge of each decision tree, balance the actions of different models and enable the recognition result to be more accurate and reliable.

Referring to fig. 2, the present application also provides a PCBA anomaly determination system, the system comprising:

On the basis of the above embodiment, the target feature determining module is further configured to determine a condition difference between the specific condition and a standard test condition; based on the condition difference value, adjusting the weight ratio of each production condition in the first production conditions to obtain adjusted first production conditions; extracting the characteristics of the adjusted first production condition to obtain first characteristics; and extracting the characteristics of the second production condition to obtain second characteristics.

On the basis of the above embodiment, the target abnormal part output module is further configured to input the first feature into a recognition model after training is completed, and output a first abnormal part; inputting the second characteristic into the recognition model after training is completed, and outputting a second abnormal part; and determining the common abnormal piece in the first abnormal piece and the second abnormal piece as the target abnormal piece.

Based on the above embodiment, the PCBA abnormal part determination system may further include:

the production condition acquisition module is used for acquiring normal production conditions of the normal PCBA and abnormal production conditions of the abnormal PCBA;

the marking condition determining module is used for marking the normal production condition and the abnormal production condition to obtain marking conditions;

The characteristic extraction module is used for respectively carrying out characteristic extraction operation on the normal production condition and the abnormal production condition to obtain different sample characteristics;

The training result generation module is used for training each decision tree in the initial recognition model by adopting the different sample characteristics to obtain a corresponding training result;

and the recognition model training module is used for combining the training results to obtain the recognition model after training.

On the basis of the above embodiment, the recognition model training module is further configured to fit each trained decision tree through a soft voting mechanism, so as to obtain the trained recognition model.

On the basis of the above embodiment, the recognition model training module is further configured to determine an accuracy of a current recognition model through the out-of-bag sample set; determining the recall rate of the training result according to the correct sample number of the training result; and adjusting parameters of the identification model according to the accuracy rate and the recall rate.

It should be noted that: in the device provided in the above embodiment, when implementing the functions thereof, only the division of the above functional modules is used as an example, in practical application, the above functional allocation may be implemented by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to implement all or part of the functions described above. In addition, the embodiments of the apparatus and the method provided in the foregoing embodiments belong to the same concept, and specific implementation processes of the embodiments of the method are detailed in the method embodiments, which are not repeated herein.

The embodiment of the application also provides a computer storage medium, which can store a plurality of instructions, the instructions are suitable for being loaded by a processor and executing the method for determining abnormal PCBA components in the embodiment, and the specific execution process can refer to the specific description of the embodiment and is not repeated herein.

The application also discloses electronic equipment. Referring to fig. 3, fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, at least one communication bus 302.

Wherein the communication bus 302 is used to enable connected communication between these components.

The user interface 303 may include a Display screen (Display) interface and a Camera (Camera) interface, and the optional user interface 303 may further include a standard wired interface and a standard wireless interface.

The network interface 304 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface), among others.

Wherein the processor 301 may include one or more processing cores. The processor 301 utilizes various interfaces and lines to connect various portions of the overall server, perform various functions of the server and process data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and invoking data stored in the memory 305. Alternatively, the processor 301 may be implemented in at least one hardware form of digital signal Processing (DIGITAL SIGNAL Processing, DSP), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA), programmable logic array (Programmable Logic Array, PLA). The processor 301 may integrate one or a combination of several of a central processing unit (Central Processing Unit, CPU), an image processor (Graphics Processing Unit, GPU), and a modem, etc. The CPU mainly processes an operating system, a user interface diagram, an application program and the like; the GPU is used for rendering and drawing the content required to be displayed by the display screen; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 301 and may be implemented by a single chip.

The Memory 305 may include a random access Memory (Random Access Memory, RAM) or a Read-Only Memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer readable medium (non-transitory computer-readable storage medium). Memory 305 may be used to store instructions, programs, code, sets of codes, or sets of instructions. The memory 305 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-described respective method embodiments, etc.; the storage data area may store data or the like involved in the above respective method embodiments. Memory 305 may also optionally be at least one storage device located remotely from the aforementioned processor 301. Referring to fig. 3, an operating system, a network communication module, a user interface module, and an application program of a PCBA abnormal component determination method may be included in the memory 305 as a computer storage medium.

In the electronic device 300 shown in fig. 3, the user interface 303 is mainly used for providing an input interface for a user, and acquiring data input by the user; and processor 301 may be configured to invoke an application program in memory 305 that stores a PCBA exception determination method, which when executed by one or more processors 301, causes electronic device 300 to perform the method as described in one or more of the embodiments above. It should be noted that, for simplicity of description, the foregoing method embodiments are all described as a series of acts, but it should be understood by those skilled in the art that the present application is not limited by the order of acts described, as some steps may be performed in other orders or concurrently in accordance with the present application. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all of the preferred embodiments, and that the acts and modules referred to are not necessarily required for the present application.

In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to related descriptions of other embodiments.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, such as a division of units, merely a division of logic functions, and there may be additional divisions in actual implementation, such as multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some service interface, device or unit indirect coupling or communication connection, electrical or otherwise.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable memory. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in whole or in part in the form of a software product stored in a memory, comprising several instructions for causing a computer device (which may be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method of the various embodiments of the present application. And the aforementioned memory includes: various media capable of storing program codes, such as a U disk, a mobile hard disk, a magnetic disk or an optical disk.

The foregoing is merely exemplary embodiments of the present disclosure and is not intended to limit the scope of the present disclosure. That is, equivalent changes and modifications are contemplated by the teachings of this disclosure, which fall within the scope of the present disclosure. Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure.

This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains.

Claims

1. A PCBA anomaly determination method, comprising:

Acquiring specific conditions when the PCBA is applied by a target client, and sampling and detecting at least one target PCBA in the batch PCBA according to the specific conditions, wherein the specific conditions refer to parameter requirements of the target client on the actual use environment of the PCBA, and the specific conditions comprise temperature conditions, humidity conditions and mechanical conditions;

When determining that the abnormal part exists in the target PCBA, determining a second production condition corresponding to the specific condition in a first production condition of the batch PCBA, wherein the first production condition is a standard production parameter condition in the PCBA manufacturing process, the first production condition comprises a processing technology condition, a device condition, an operator condition and a factory environment condition, and the second production condition is a direct cause condition for causing the PCBA abnormality according to the corresponding relation between the abnormality detected by the specific condition and the specific production parameter on the basis of the first production condition;

Inputting the target features into a recognition model after training is completed, and outputting target abnormal parts in the batch PCBA;

the feature extraction of the first production condition and the second production condition to obtain target features includes:

extracting the characteristics of the second production condition to obtain second characteristics;

The target feature comprises a first feature corresponding to a first production condition and a second feature corresponding to a second production condition, the target feature is input into a recognition model with training completed, and a target abnormal part in the batch PCBA is output, and the target feature comprises:

2. The PCBA anomaly determination method of claim 1, further comprising, prior to the obtaining the specific condition at which the target customer applies the PCBA:

3. The PCBA anomaly determination method of claim 2, wherein the sample features comprise: material batch information, production process information, staff information, tool equipment information used and production environment information.

4. The PCBA anomaly determination method of claim 2, wherein the training results comprise a trained decision tree, and combining the training results to obtain a trained feature recognition model, comprising:

5. The PCBA anomaly determination method of claim 2, wherein the sample features further comprise out-of-bag samples, wherein training each decision tree in an initial recognition model using the different sample features, after obtaining corresponding training results, further comprises:

6. A PCBA anomaly determination system, the system comprising:

The specific condition determining module is used for obtaining specific conditions when the PCBA is applied to the target clients and sampling and detecting at least one target PCBA in the batch PCBA according to the specific conditions, wherein the specific conditions refer to parameter requirements of the target clients on the actual use environment of the PCBA, and the specific conditions comprise temperature conditions, humidity conditions and mechanical conditions;

The production condition determining module is used for determining a second production condition corresponding to the specific condition in first production conditions of the batch PCBA when the abnormal component exists in the target PCBA, wherein the first production conditions are standard production parameter conditions in the PCBA manufacturing process, the first production conditions comprise processing process conditions, equipment conditions, operator conditions and factory environment conditions, and the second production conditions are direct cause conditions for causing the abnormality of the PCBA according to the corresponding relation between the abnormality detected by the specific condition detection on the basis of the first production conditions;

The target abnormal part output module is used for inputting the target characteristics into the recognition model after training is completed and outputting target abnormal parts in the batch PCBA;

the target feature determining module is further used for determining a condition difference value between the specific condition and a standard test condition; based on the condition difference value, adjusting the weight ratio of each production condition in the first production conditions to obtain adjusted first production conditions; extracting the characteristics of the adjusted first production condition to obtain first characteristics; extracting the characteristics of the second production condition to obtain second characteristics;

The target abnormal part output module is further used for inputting the first characteristic into a recognition model after training is completed and outputting a first abnormal part; inputting the second characteristic into the recognition model after training is completed, and outputting a second abnormal part; and determining the common abnormal piece in the first abnormal piece and the second abnormal piece as the target abnormal piece.

7. An electronic device comprising a processor, a memory, a user interface, and a network interface, the memory to store instructions, the user interface and the network interface to communicate to other devices, the processor to execute the instructions stored in the memory to cause the electronic device to perform the method of any of claims 1-5.

8. A computer storage medium storing instructions which, when executed, perform the method of any one of claims 1-5.