TWM601387U - Intelligent system for testing sensed data - Google Patents

Abstract

An intelligent system for testing sensed data is provided. The system includes a computer system having a memory for storing instructions and algorithmic programs, and a database for storing sensed data obtained from an object to be tested. The sensed data is divided into a training set and a testing set. A deep learning algorithm is performed on the training set for retrieving features of the object so as to establish a deep learning based testing model. The testing set is used to test the testing model, and a machine learning algorithm is performed on the data that fails to pass the test. Historical parameter data is further introduced to train the testing model being established by machine learning, and a model optimization algorithm is applied to optimize the testing model for obtaining optimized control parameters of the object.

Description

Sensing data intelligent detection system

The specification discloses a technology for detecting sensing data, in particular a sensing data intelligent detection system that uses deep learning and machine learning methods to establish a detection model from the sensing data.

When detecting whether a product, a system, or a field is abnormal, the common way is to use a specific sensor to sense the data, such as sensing sound, shooting images, etc., and then use analysis tools to analyze the sensed data Information to determine whether the object under test is abnormal.

For example, when it is necessary to determine whether an object surface is flawed, the surface of the object can be photographed with a camera, and the surface image of the object can be compared with the sample image to determine whether there is any abnormality. Taking a motor as an example, the general method is to record the audio generated when the motor is running by a sound sensor. After comparing the sound model, it can be judged whether there is abnormal operation, which can be used as a basis for future improvement.

The specification discloses an intelligent detection system for sensing data. The system is like a computer system in which a detection method for sensing data is run. The computer system is provided with a memory in which programs and algorithms for executing the intelligent detection method for sensing data are stored. There is a database in which the sensing data obtained from an object to be detected is stored.

In an embodiment, the sensed data is data obtained by sensing the object to be detected by the sensor. In the intelligent detection method for sensed data executed by the system, the sensed data is first divided into training data and test data. Perform a deep learning algorithm on the training data, obtain the characteristics of the object to be detected from the training data, establish a detection model for detecting the object to be detected, and then use the test data to test the detection model. The result is the failed test data, the machine learning algorithm is executed, the detection model is trained with historical parameter data, and the detection model is optimized with a model optimization algorithm to obtain the optimized control parameters of the object to be tested.

Further, verification data is also obtained from the training data. When the detection model is formed from the training data through the deep learning algorithm in the above steps, the verification data can be used to verify the detection model to obtain a parameter model for generating control parameters . The control parameters are parameters that drive the operation of the object to be detected, and generating optimized control parameters is one of the main purposes of the method.

Further, the model optimization algorithm can be used to optimize the parameter model, and then generate optimized control parameters, and continue to drive the object to be detected. Repeat the above steps to obtain the sensor data again, and optimize the sensor data by repeating the sensor data intelligent detection method. Control parameters of the detection object.

Further, in one embodiment, a K-fold cross-validation method can be used to evaluate multiple models generated from multiple deep learning algorithms and machine learning methods using verification data, and the evaluation factors are, for example, generated from multiple deep learning algorithms The accuracy, precision and recall rate of each model are obtained from the multiple models, and then the detection model is selected based on the score.

In order to further understand the features and technical content of the present invention, please refer to the following detailed descriptions and drawings about the present invention. However, the drawings provided are only for reference and explanation and are not used to limit the present invention.

The following are specific specific examples to illustrate the implementation of this creation, and those skilled in the art can understand the advantages and effects of this creation from the content disclosed in this specification. This creation can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the concept of this creation. In addition, the drawings in this creation are merely schematic illustrations, and are not depicted in actual size, and are stated in advance. The following embodiments will further describe the related technical content of this creation in detail, but the disclosed content is not intended to limit the protection scope of this creation.

It should be understood that although terms such as “first”, “second”, and “third” may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one element from another, or one signal from another signal. In addition, the term "or" used in this document may include any one or a combination of more of the associated listed items depending on the actual situation.

The disclosure book discloses a sensory data intelligent detection system, in which the sensory data intelligent detection method is run. One of the main purposes is to use the deep learning algorithm to train and build the model, and to cooperate with the machine learning algorithm to perform later testing and training. A detection model for detecting a specific object to be detected is trained from the measurement data. For example, the object to be inspected can be a product, a system, or a field. The sensor is used to sense specific information of the object to be inspected, such as images, sounds, and vibrations, and the resulting sensing data such as shooting the object to be inspected The obtained image, the recorded sound produced by the object to be detected and converted spectrum, or the information of the vibration generated when the object to be detected is detected in operation, these sensed data can be used as information for detecting various objects to be detected for optimization Control parameters that drive the operation of the object to be detected.

For an embodiment of the intelligent detection system for sensing data, refer to the system architecture embodiment diagram shown in FIG. 1.

The figure shows an object 10 to be detected. The object to be detected is driven by control parameters. The sensing data intelligent detection system can optimize its own control parameters by learning the sensing data, or optimize the system that generates the object to be detected Control parameters. For example, the object to be detected 10 is sensed by one or more sensors (sensor one 101, sensor two 102, and sensor three 103) to obtain sensing data. Taking the image generated by the object to be detected 10 as an example, the first sensor 101, the second sensor 102, and the third sensor 103 can be cameras that shoot at multiple angles; if it is used to sense the environment of a certain field, the first sensor 101. The second sensor 102 and the third sensor 103 may be sensors located at different positions in the field. The sensing data generated by sensor one 101, sensor two 102, and sensor three 103 can be stored in the sensing data processing host 12, and then converted and preliminarily processed into data in the database 145 in the computer system 14.

The computer system 14 is provided with a processor 141, a memory 143, and a database 145. The memory 143 is used to store programs and algorithms for performing the intelligent detection method of sensing data, and the database 145 is used to store one or more sensors. The sensing data obtained after measuring the object 10 to be inspected, and then the processor 141 of the computer system 14 executes the intelligent sensing data detection method.

Using the above system architecture, FIG. 2 shows an embodiment diagram of the main flow of a method for intelligent detection of sensor data executed by the system.

At the beginning of the system, in step S201, the sensing data obtained from the object to be detected is obtained, and then in step S203, the sensing data is divided into training data and testing data. In a specific embodiment , And then the validation data can be obtained from the training data. In step S205, a deep-learning algorithm is executed on the training data, and the characteristics of the object to be detected can be obtained from the training data. These features reflect the information about the operation of the object to be detected or the operation of related systems. It learns from it to establish a detection model to detect the object to be detected, which is a detection model based on deep learning.

After that, in step S207, the detection model obtained in the above steps is tested with the test data selected in the sensing data, and the detection result includes the part that passed the test and the part that failed the test. The tested data indicates that the detection model meets expectations, and that the model parameters derived from the current detection model are appropriate and do not need to be adjusted. However, if the detection result includes the part that failed the test, in the intelligent detection method for sensing data, as in step S209, the machine learning algorithm can be executed for the failed test data, and the above steps can be trained with historical parameter data The obtained detection model can be trained with historical parameter data to obtain a parameter model, and then a model optimization algorithm is used to optimize the detection model. This is a detection model based on machine learning. In step S211, the optimized object to be detected can be obtained Control parameters.

Machine learning (machine learning) methods cover many, such as support vector machine (Support vector machine), random forest (Random Forest), simple Bayes (Naïve Bayes) and deep learning algorithms and so on. The above-mentioned deep learning algorithm is a kind of machine learning method. It uses the computing power of the processor in the computer system to pass the large amount of sensing data obtained through linear or non-linear in multiple processing layers. Linear transformation (linear or non-linear transformation) uses a feature extraction step to obtain features representing the characteristics of the sensing data of the object to be detected.

Common deep learning algorithms use convolutional neural networks (Convolution Neural Networks, CNN) or recurrent neural network (Recurrent neural network, RNN) two methods, convolutional neural network can be in its convolutional layer (convolutional layer) Use the convolution operation to filter the sensed data, and use the computing power of the computer processor to gradually extract the features in order to finally establish a model (such as the above-mentioned detection model) that can be used to detect the sensed data generated by the system.

For example, the sensing data obtained from the object to be detected, such as image data, sound data, vibration data, etc., in the method, the system can set a part (such as 90%) of the sensed data Training data. The training data uses deep learning algorithms (such as CNN, RNN) to establish a detection model, and the remaining part of the sensing data (such as 10%) is set as testing data to test the system The established detection model.

Figure 3 shows an embodiment of the process of a method for intelligently detecting system operation sensing data.

At the beginning of the process, the sensing data (301) can be generated from the object to be detected, and the sensing data (303) of the intelligent detection establishment model of the sensing data can be formed, the sensing data or its part (such as the training obtained from the sensing data) Data) is used to execute the deep learning algorithm (305) to establish a detection model based on deep learning, and then test the detection model (307) with partial data (such as the test data obtained from the sensor data), and target the Part, that is, the process without subsequent parameter optimization, ends the process (309); for the part that fails the test, it is used to build a parameter model (311).

In the process of establishing the parameter model (311), machine learning is performed again on the data that failed the test, historical parameter data is obtained from the historical parameter database (313) in the system, the parameters are modeled, and the parameter model is established. A model optimization algorithm (315) optimizes the model parameters of the detection model to form a detection model based on machine learning, and finally generates optimized control parameters (317), which can be re-injected into the object to be detected and continue to generate sensing Data (301).

It is mentioned here that the control parameters are parameters that drive the operation of the object to be inspected or the operating parameters of a system that generates the object to be inspected. In particular, in one embodiment, validation data can also be derived from the training data. When the training data is used to form a detection model through a deep learning algorithm, the validation data can be used to validate the detection model to obtain A parameter model is used to generate the control parameters, and then the parameter model is optimized by the model optimization algorithm (315), and then the optimized control parameters are output.

In this way, the optimized control parameters (317) drive the object to be detected, and the sensing data (301) is obtained again, and the control parameters of the object to be detected are continuously optimized by repeating the intelligent detection method of the sensor data until the system presets A state of expectation.

Figure 4 then shows another example of the process of system operation. This example shows that the continuously generated sensing data (303) stored in the system will become the historical sensing database (400), and these historical sensing databases 400 continue to perform deep learning ( 305), using the computing power of the computer system to continuously train data to optimize the detection data. On the other hand, after testing the detection model, the parameters in the data that pass the test can become part of the historical parameter database (313), and continue to perform machine learning repeatedly to optimize the parameter model.

According to an embodiment, when the model optimization algorithm is executed, a variety of parameter models may be used, such as genetic algorithms (Genetic Algorithms), particle swarm optimization (Particle Swarm Optimization), ant colony optimization (Ant Colony Optimization), One of Simulated Annealing, Tabu Search, Seagull Optimization Algorithm, and Bayesian Optimization. The optimized parameter models such as gradient boosting decision tree model (Gradient Boosting Decision Tree), extreme gradient boosting model (Extreme Gradient Boosting), category boosting model (Categorical boosting), light GBM model (Light GBM), random forest model (Random Forest), Support Vector Machine model (Support Vector Machine), Relevance vector machine model (Relevance vector machine), simple Bayes classification model (Naïve Bayes), K nearest neighbor calculus model (K nearest neighbor), CNN model or RNN model . In a preferred embodiment, the model optimization algorithm used by the system can use Bayesian Optimization, the deep learning algorithm can use the CNN model, and the parameter model can use the extreme gradient boosting model (Extreme Gradient Boosting).

According to one of the embodiments, a K-fold cross-validation method may be used in the intelligent detection method for sensing data, and this verification method may be to divide the training data into multiple groups (K) data, One set of data is used as the data for the verification model, and the other sets of data (K−1) are used as the training data. After that, K sets of data are used for K verifications, and finally the parameters of the evaluation model can be obtained. Among them, the K-fold cross-validation method evaluates multiple models generated from multiple deep learning algorithms from the verification data, and then selects detection models from them, and one of the ways to evaluate multiple models is to obtain an accuracy of each model Factors such as accuracy, precision, and recall, including the evaluation equations derived from these factors.

For example, the K-fold cross-validation method uses 10-fold cross-validation as an example. During the calculation, the acquired sensing data is divided into 10 equal parts, and the first point data can be used as a test For the test data used by the model, the remaining 9 points can be used for training data. Then, in the next round of the process, the second point data will be used as the data for testing the detection model, and the remaining 9 points will also be used for training data. By analogy, a total of 10 rounds of the process can be executed to obtain an accuracy rate of 10 and an average value to obtain a more objective accuracy rate.

In one embodiment, the relationship between whether the result predicted by each detection model is yes or no (YES/NO) and whether the actual detection result obtained by actually measuring the object to be detected is yes or no (YES/NO) The "accuracy", "precision" and "recall rate" are derived based on the characteristics, which are used as the basis for evaluating the detection model.

For example, the data whose model prediction result is YES and the actual detection result is YES is set to "TP"; the model prediction result is YES and the actual detection result is NO (predicted The data of wrong) is set to "FP"; the result of model prediction is no (NO) and the actual test result is (YES) (indicating that the prediction is wrong) data is set to "FN"; the model predicted is no (NO) and the actual The data whose test is also No (NO) is set to "TN".

According to the above settings, the "accuracy" indicates the correct ratio of model prediction, the mathematical formula is: "Accuracy=(TP+TN)/N _totoal ", that is: the model prediction result is yes and the actual detection result is also yes The data ("TP") plus the data that the model predicts to be no and the actual test is also no ("TN") is divided by the total data (N _total ).

The "accuracy" is one of the main goals for the entire system to optimize the detection model. The mathematical formula is: "Precision=TP/(TP+FP)", that is, the actual detection result is yes and the model prediction is also yes ("TP") is the proportion of the total number predicted by the model (ie, "TP" + "FP").

The "recall rate" represents the error rate of the model, and the mathematical formula is: "Recall=TP/(TP+FN)", that is, the data that the model prediction result is yes and the actual detection result is also yes ("TP") It accounts for the percentage of all data actually detected as yes (ie "TP" + "FN").

Furthermore, other calculations (such as F1 Score=2/((1/Precision)+(1/Recall))) of the calculation results of the above "accuracy", "precision" and "recall rate" can be used as evaluation The basis for the pros and cons of the model.

FIG. 5 shows a schematic diagram of an embodiment of the operation of training data, verification data, and test data in the method for intelligent detection of sensor data. The following description is provided in conjunction with the flowchart of the embodiment of the method for intelligent detection of sensor data shown in FIG. 6.

The database 50 in the system is provided with sensing data and parameter data (step S601). The sensing data can be mainly divided into training data 501 and test data 505, and part of the training data 501 can be used as verification data 503 (step S603). Based on the training data 501, a deep learning algorithm is used to train the model (52) to obtain a detection model (step S605). Perform a test model (56) with the test data 505 (step S607) to obtain a parameter model for generating control parameters. At this time, the control parameters can be generated by the machine learning model (step S609). On the other hand, the verification data 503 is to perform model verification to obtain a parametric model (step S611).

According to an embodiment, the parameter model can be optimized by the model optimization algorithm (53) (step S613), and the parameter model can be continuously tested with test data (step S615), and then the detection model (54) can be formed. After repeating the above process, different deep learning algorithms can be introduced, and multiple detection models can be obtained through repeated optimization of training data 501, verification data 503, and test data 505 to generate control parameters (step S617). After comparing the model 57, the model (58) suitable for the system is determined according to the requirements, and the intelligent detection process of the sensing data (59) is completed.

For the above method of optimizing the model, please refer to the example flow chart of optimization in the method for intelligent detection of sensing data shown in FIG.

According to the above embodiment, in the process of using the model optimization algorithm to optimize the model to define the optimal detection model, historical parameter data can be imported (step S701), and the machine learning algorithm is used to train these data to establish a parameter model (step S703). Then use the parameter model to obtain the control parameters, and use the K-fold cross-validation method one by one (step S705) to test multiple models generated from multiple deep learning algorithms with the verification data (step S707), and then evaluate them (if using The above-mentioned factors such as "accuracy", "precision" and "recall rate") and selecting the most suitable detection model (step S709). Then, a model parameter optimization algorithm is performed on the selected model (step S711) to obtain an optimized detection model (step S713), and the detection model can be used to generate optimal control parameters.

For an example, refer to the flowchart shown in FIG. 8. This example shows an intelligent method for establishing a detection model for detecting a motor. The object to be detected is a motor.

Obtain the vibration sound of the motor through the sound sensor (step S801), convert the audio into a frequency spectrum, create data in the sensing database (step S803), and then train the sensing data with a deep learning algorithm (such as the training obtained therein) Data), extract the sound characteristics of the motor operation status in the data, obtain the correlation of various audio frequencies on the abnormal operation of the motor and its control parameters, and establish a detection model for detecting the motor (step S805). After that, the system uses the verification data obtained from the sensing data to verify the above-mentioned step training to obtain a detection model (step S807), and determine the detection model after verification (step S809).

The system uses this detection model to generate control parameters, drive the motor, and at the same time record sound through the sensor to form the audio sensing data, and then detect these sensing data (such as the imaged spectrum) (step S811), from the sensing Has the data passed the test? (Step S813), if the test is passed, it means that the control parameters determined by the current detection model are used to drive the motor to meet the requirements of the system, and the process ends (Step S815).

If it fails the test (No), it means that it is necessary to continue to optimize the parameters with the parameter model (step S817), and then drive the motor to operate with the generated control parameters (step S819), and continue to use the sensor to obtain the vibration sound. , Judging whether to pass the test from the test data of the test model, and repeat the above steps to continuously optimize the parameter model to obtain the optimal control parameters.

Further, according to another embodiment of the application, the disclosed intelligent detection method for sensing data is applied to the detection screen, the object to be detected is the screen, and the screen is driven by display parameters to display the content. In the above process, image sensing can be used first The device obtains a large number of screen image data, which becomes the sensing data in the system’s database, from which the features in the image data are obtained by the deep learning algorithm, the correlation between the image and the screen display parameters is established, and the detection model of the detection screen can be established, including The detection model is determined by the verification data verification, and then the detection model is used to generate the display parameters, and then the image data of the screen is generated, tested, and the subsequent machine learning algorithm is executed to optimize the detection model to obtain the optimized display parameters of the screen.

To sum up, according to the sensory data intelligent detection system described in the above embodiments, in particular, when the training data is used to build a model, in addition to part of the sensor data as training data (training data), part of it is also used as a test Model data (testing data), and then obtain a part from the training data as verification data (validation data) to verify the detection model generated by the system, which is an internal optimization step. What's more, when training a model, a variety of deep learning and machine learning algorithms can also be used to generate a variety of models. These models can be optimized in the same way, and then the model can be compared and evaluated before one of the models is selected for practical application.

The content disclosed above is only a preferred and feasible embodiment of the present model, and does not therefore limit the scope of the patent application of the present model. Therefore, all equivalent technical changes made using the description and schematic content of the present model are included in the application of the present model. Within the scope of the patent.

10: Object to be detected 101: Sensor One 102: Sensor Two 103: Sensor Three 12: Sensing data processing host 14: Computer system 141: Processor 143: Memory 145: Database 301: Generate sensing data 303: Sensed Data 305: Deep Learning Algorithm 307: test detection model 309: End 311: Parametric Model 313: Historical parameter database 315: Model optimization algorithm 317: Generate control parameters 400: Historical sensor database 50: database 501: training data 503: verify data 505: test data 52: training model 53: Model optimization algorithm 54: Detection model 56: Test model 57: compare models 58: Judgment Model 59: Done Steps S201～S211: the main flow of the intelligent detection method of sensing data Steps S601～S617: the flow of the sensing data intelligent detection method Steps S701～S713: Optimization process in the intelligent detection method of sensing data Steps S801～S819: Process example of the intelligent detection method of sensing data

Figure 1 shows an embodiment diagram of the system architecture of a sensing data intelligent detection system;

Figure 2 shows an embodiment diagram of the main flow of a method for intelligent detection of sensed data;

Figure 3 shows one of the process embodiments of a method for intelligent detection of system operation sensing data;

Figure 4 shows the second embodiment of the process of a method for intelligent detection of system operation sensing data;

FIG. 5 shows a schematic diagram of an operation embodiment of training data, verification data, and test data in the intelligent detection method for sensing data;

Fig. 6 shows a flowchart of an embodiment of a method for intelligent detection of sensor data;

Figure 7 shows an exemplary flow chart of optimization in the intelligent detection method of sensor data; and

Fig. 8 shows an exemplary flow chart of an intelligent detection method using sensing data.

301: Generate sensing data

303: Sensed Data

305: Deep Learning Algorithm

307: test detection model

309: End

311: Parametric Model

313: Historical parameter database

315: Model optimization algorithm

317: Generate control parameters

Claims (10)

An intelligent detection system for sensing data includes: a computer system provided with a memory in which programs and algorithms for executing a sensing data intelligent detection method are stored, and the computer system is also provided with a database, which is stored from a to-be-tested The sensing data obtained by the subject; wherein a processor of the computer system is used to execute the sensing data intelligent detection method, including: acquiring the sensing data, distinguishing the sensing data into training data and test data; The data executes a deep learning algorithm, obtains the characteristics of the object to be detected from the training data, and establishes a detection model for detecting the object to be detected; the detection model is tested with the test data; the test result is failed The machine learning algorithm is executed on the test data, and the detection model is trained with historical parameter data, and the detection model is optimized with a model optimization algorithm to obtain the optimized control parameters of the object to be detected. The sensing data intelligent detection system of claim 1, wherein the object to be detected is sensed by one or more sensors to obtain sensing data, and the generated sensing data is stored in a sensing data processing host , And then converted and initially processed to become the sensing data in the database of the computer system. The sensory data intelligent detection system of claim 2, wherein the one or more sensors are a sound sensor, the object to be detected is a motor, and the vibration sound is obtained through the sound sensor, and the The audio is converted to frequency spectrum, and the sensing data in the database is established. The sensory data intelligent detection system of claim 2, wherein the one or more sensors is an image sensor, the object to be detected is a screen, and the image The image sensor obtains the image data of the screen and becomes the sensing data in the database. The sensing data intelligent detection system according to claim 1, wherein verification data is obtained from the training data, and when the training data forms the detection model through the deep learning algorithm, the verification data is used to verify the detection model, To obtain a parameter model for generating the control parameter. The sensor data intelligent detection system of claim 5, wherein the parameter model is optimized by the model optimization algorithm, and then the optimized control parameter is generated. The sensory data intelligent detection system according to claim 6, wherein, in the sensory data intelligent detection method, the optimized control parameter is then driven to the object to be detected to obtain the sensory data again, and by repeating all The intelligent detection method for sensing data described above optimizes the control parameters of the object to be detected. The sensory data intelligent detection system according to claim 7, further comprising a historical sensor database for storing the sensory data continuously generated by the sensory data intelligent detection system, and proceed with the data in the historical sensor database Deep learning, continuous training data to optimize the parameter model. The sensing data intelligent detection system according to any one of claim 5 to 8, wherein the verification data is used to evaluate a plurality of models generated from a plurality of deep learning algorithms by a K-fold cross-validation method, and then the Detection model. The sensing data intelligent detection system according to claim 9, wherein the basis for evaluating the detection model includes obtaining an accuracy and an accuracy of each model from the plurality of models generated by the plurality of deep learning algorithms Degree and a recall rate.

Images