Perception method and system for analyzing the working status of intelligent terminals through current
Through the current analysis and perception method, Kalman filtering and perceptron classification are used to identify the current state of the smart terminal, solving the problems of leakage and peak current interference, realizing automated detection and accurate prediction, and improving the detection efficiency and product quality of the terminal.
Patent Information
- Application Number
- CN202111247743.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-10-26
AI Technical Summary
During the assembly process of the smart terminal, components breakdown or welding short circuits lead to leakage, resulting in too high system working current, battery consumption is too fast, and may even cause fires, which are difficult to effectively detect and prevent in the existing technology.
Through the current analysis and perception method, floating-point data with preset accuracy is used to obtain the current value of the program-controlled power supply in real time, forming a time series data stream, performing Kalman filtering and perceptron classification, identifying and predicting the three states of the terminal, and calculating the average and standard current to determine leakage conditions.
Implement automated analysis, reduces manual judgment errors, reduces workers' labor and operation complexity, improves production line efficiency and product quality, effectively solves the problem of peak current interference, and accurately predicts the shutdown, power-on and standby status of the terminal. .
Smart Images

Figure CN113986972B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of smart terminal manufacturing, and in particular to a sensing method and system for analyzing the working status of a smart terminal through electric current. Background Art
[0002] With the development of technology, smart devices, especially smartphones, have become a must-have product for everyone, facilitating communication and enriching people's daily lives. Because smart devices are battery-powered, improvements in reducing power consumption and extending battery life are extremely important.
[0003] Today's tablets, mobile phones, and other handheld devices are mostly developed based on Android and Linux systems. Because they are battery-powered and require recharging only once a day or even several days, power consumption management is extremely stringent. If a component breaks down or a soldering short occurs, leakage can occur, leading to excessive current and rapid battery drain. In even greater danger, a short circuit can occur between the positive and negative terminals, causing high temperatures that can lead to battery swelling or even fire, damaging the device or posing a fire hazard.
[0004] Therefore, in order to be able to pick out machines with abnormal working current during the terminal manufacturing process (especially during the whole machine assembly process), the present invention proposes a sensing method and system for analyzing the working status of smart terminals through current. Summary of the Invention
[0005] In response to the problems in the related art, the present invention proposes a sensing method and system for analyzing the working status of a smart terminal through current, so as to overcome the above-mentioned technical problems existing in the existing related art.
[0006] To this end, the specific technical solutions adopted in the present invention are as follows:
[0007] According to one aspect of the present invention, a sensing method for analyzing the working status of a smart terminal through current is provided, and the sensing method includes the following steps:
[0008] S1. Using floating-point data with preset precision to represent the current value of the programmable power supply obtained in real time;
[0009] S2, forming a time series data stream from the floating point data according to a preset time series rule;
[0010] S3, performing Kalman filtering on the time series data stream;
[0011] S4. Use the perceptron classification method to identify the current and predict the three states of the terminal under test;
[0012] S5. Calculate the average current and standard current in each state and compare them to determine the leakage conditions in the three states.
[0013] Furthermore, the current value of the programmable power supply in S1 is obtained every 50 milliseconds, and the preset accuracy is 0.001 mA.
[0014] Furthermore, the Kalman filtering process performed on the time series data stream in S3 includes the following steps:
[0015] S31, using a collection Represents the time series data stream, let Represents the vector at time t, and uses F t Represents the prediction matrix, and the current estimate at time t is calculated as:
[0016]
[0017] S32, add the total current consumption of each component in the terminal under test, add them up and draw power from the power line, based on the fact that the power consumption of each component changes dynamically at all times, and get the increment at time t as vector Then the estimated current at time t is:
[0018]
[0019] Among them, B t Represents the dynamic current change increment matrix;
[0020] S33, the measured peak current and the measured interference of the programmable power supply are collectively referred to as interference, and the covariance matrix Q t Represents the interference current noise, then the total covariance matrix P t Expressed as:
[0021]
[0022] S34, combined set Using matrix H t As each t moment takes out the set The conversion matrix is used to calculate the estimated distribution of the current consumption of the terminal under test at time t:
[0023]
[0024]
[0025] Among them, H t is the identity matrix, H t T H tThe transposed matrix of , ∑ represents the covariance of the predicted Gaussian distribution;
[0026] S35, combined with the interference covariance matrix fusion part, calculate and obtain:
[0027]
[0028] in, represents the predicted current value at time t, Indicates that at time t iterations, represents the distribution mean, K represents the Kalman gain matrix, K′ represents the Kalman gain iteration, R t represents the noise covariance, P t ′ represents the iteration result, and P t ' will be P t Substitute into the formula to generate a new P t ′.
[0029] Furthermore, in S31, x t-1 、x t and x t+1 All conform to the Gaussian distribution, that is, each component has a mean μ and variance δ 2 , indicating uncertainty; the interference current noise Q in S33 t It also conforms to the Gaussian distribution.
[0030] Furthermore, the method of using a perceptron classification method to identify the current in S4 and predicting the three states of the terminal under test includes the following steps:
[0031] S41, adopt collection Represents the time series data stream, let Represents the vector at time t, let vector is the weight at time t, then we get;
[0032]
[0033] Where b represents the bias;
[0034] S42, add bias b to vector Get And in the vector Add 1 to get The above calculation formula Converted to vector form, we get:
[0035]
[0036] S43, with the formula It is a decision boundary model that predicts a binary classification problem. The calculation formula of the current prediction value is as follows:
[0037]
[0038] S44, marking the current data in the power-on state, the power-off state and the standby state respectively;
[0039] S45, using the above three sets of current data sequences to train the power-on state A and other states B, And in state B, train the system to distinguish between shutdown state and standby state.
[0040] Furthermore, the current data marked in S44 includes a current measurement value in a power-on state, a current measurement value in a power-off state, and a current measurement value in a standby state.
[0041] Furthermore, in S45, the three sets of current data are used to train the power-on state A and other states B, and the And in state B, train the system to distinguish between shutdown state and standby state. The following steps are also included:
[0042] S451, read training samples The superscript t represents the tth group of vectors, and the prediction is performed
[0043] S452, if represents the predicted current value, If the standard answer is marked, the parameter is updated
[0044] S453, repeat S452 until the gradient drops below 0.0001, and finally obtain the
[0045] Furthermore, the comparison formula of the average current and the standard current in each state in S5 is as follows:
[0046]
[0047] in, It represents the average current in the case of no leakage, and |δ| represents the threshold radius of n leakage-free machines, or tolerance.
[0048] According to another aspect of the present invention, a sensing system for analyzing the operating status of a smart terminal through current is provided. The system includes an application software layer, a hardware layer, and a semi-finished device layer to be tested. The application software layer includes an identification program. The hardware layer includes a computer and a programmable power supply, and the identification program runs on the computer. The semi-finished device layer to be tested includes a plurality of terminals to be tested.
[0049] The identification program is composed of Kalman filters with different step sizes and multi-classification perceptron learning algorithms and predictions, which are used to predict the three basic states of the terminal under test: shutdown current and state, startup current and state, and standby current and state.
[0050] When the identification program is executed, the following steps are implemented:
[0051] S1. Using floating-point data with preset precision to represent the current value of the programmable power supply obtained in real time;
[0052] S2, forming a time series data stream from the floating point data according to a preset time series rule;
[0053] S3, performing Kalman filtering on the time series data stream;
[0054] S4. Use the perceptron classification method to identify the current and predict the three states of the terminal under test;
[0055] S5. Calculate the average current and standard current in each state and compare them to determine the leakage conditions in the three states.
[0056] Furthermore, the recognition program also includes, before prediction, training using several groups of manually labeled current measurement values under three states to obtain a prediction model.
[0057] The beneficial effects of the present invention are as follows: by adopting Kalman filtering with different step lengths and multi-classification perceptron learning algorithms and predictions, it is possible to predict the three basic states of the shutdown current and state, the startup current and state, and the standby current and state of the terminal under test, realize automated analysis, and reduce the occurrence of manual judgment, thereby effectively solving the errors in manual test current data, and at the same time greatly reducing the workload and operation complexity of production line workers, thereby effectively improving production line efficiency, improving accuracy, and improving product quality; in addition, compared with traditional machine recognition algorithms, the present invention adopts Kalman filtering, which effectively solves the problem of filtering peak current during current testing, solves the interference of peak current on the detection of the three states of the terminal, and adopts a multi-classification perceptron algorithm to identify current, thereby predicting the three states of shutdown, startup and standby of the terminal under test. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0059] Figure 1 is a flow chart of a sensing method for analyzing the working status of a smart terminal through current according to an embodiment of the present invention;
[0060] Figure 2 2 is a schematic diagram showing the principle of a sensing method for analyzing the working status of a smart terminal through current according to an embodiment of the present invention;
[0061] Figure 3 3 is a schematic diagram of the principle of a sensing system for analyzing the working status of a smart terminal through current according to an embodiment of the present invention. DETAILED DESCRIPTION
[0062] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0063] According to an embodiment of the present invention, a sensing method and system for analyzing the working status of a smart terminal through current are provided.
[0064] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1-2 As shown, according to one embodiment of the present invention, a sensing method for analyzing the working status of a smart terminal through current is provided, and the sensing method includes the following steps:
[0065] S1. Using floating-point data with preset precision to represent the current value of the programmable power supply obtained in real time;
[0066] The S1 reads the current of the programmable power supply in real time through a program, and reads the current every 50 milliseconds. The current value is represented by floating-point data with an accuracy of 0.001, and the unit is milliampere, such as 0.078mA, 3.303mA, 689.035mA, etc.
[0067] S2, forming a time series data stream from the floating point data according to a preset time series rule;
[0068] Since a floating-point data is generated every 50 milliseconds, 20 floating-point data are generated in 1 second, and N*20 floating-point data are generated in N seconds. Therefore, arranging these floating-point data in a column according to the time series forms a time-serialized data stream.
[0069] S3, performing Kalman filtering on the time series data stream;
[0070] The step of performing Kalman filtering on the time series data stream in S3 includes the following steps:
[0071] S31, using a collection Represents the time series data stream, let represents the vector at time t (i.e., the two related current data before and after time t are incorporated into the Kalman filter system), x t-1 、x t and x t+1 All conform to the Gaussian distribution, that is, each component has a mean μ and variance δ 2 , indicating uncertainty; using F t Represents the prediction matrix, and the current estimate at time t is calculated as:
[0072]
[0073] S32, add the total current consumption of each component in the terminal under test, add them up and draw power from the power line (the sum of the power consumption from the power line will dynamically affect the current measurement sequence). Compared with time t-1, the increment at time t is vector Then the estimated current at time t is:
[0074]
[0075] Among them, B t Represents the dynamic current change increment matrix;
[0076] S33, the measured peak current and the measured interference of the programmable power supply are collectively referred to as interference, and the covariance matrix Q t represents the interference current noise, and the interference current noise Q t It also conforms to the Gaussian distribution, then the total covariance matrix P t Expressed as:
[0077]
[0078] S34. Synthesize the serialized data stream collection before Use matrix Ht as the set to be taken out at each time t The conversion matrix is used to calculate the estimated distribution of the current consumption of the terminal under test at time t:
[0079]
[0080]
[0081] Among them, H t is the identity matrix, H t T H t The transposed matrix of , ∑ represents the covariance of the predicted Gaussian distribution. There is also an interference covariance matrix, which also conforms to the Gaussian distribution. Therefore, the fusion distribution of these two Gaussian distributions is calculated as follows:
[0082] S35, combined with the interference covariance matrix fusion part, calculate and obtain:
[0083]
[0084] in, represents the predicted current value at time t, Indicates that at time t iterations, represents the distribution mean, K represents the Kalman gain matrix, K′ represents the Kalman gain iteration, R t represents the noise covariance, P t ′ represents the iteration result, and P t ' will be P t Substitute into the formula to generate a new P t ′.
[0085] After applying the Kalman filter algorithm, the current curve is significantly improved. The green line is the noise, and the red line is the filtered curve.
[0086] Since there are K possibilities, there are three possibilities for this application, namely, shutdown state, power on state, and standby state, so K=3, and the Secondary classification method, each time distinguishing two types C i and C j , in each case there is only one category C k Win.
[0087] S4. Use the perceptron classification method to identify the current and predict the three states of the terminal under test;
[0088] The method of using a perceptron classification method to identify the current in S4 and predicting the three states of the terminal under test includes the following steps:
[0089] S41, adopt collection Represents the time series data stream, let Represents the vector at time t, let vector is the weight at time t, then we get;
[0090]
[0091] Where b represents the bias;
[0092] S42, add bias b to vector Get And in the vector Add 1 to get The above calculation formula Converted to vector form, we get:
[0093]
[0094] S43, with the formula It is a decision boundary model that predicts a binary classification problem. The calculation formula of the current prediction value is as follows:
[0095]
[0096] S44. Mark the current data in the power-on state, power-off state and standby state respectively; the details are as follows:
[0097] 236.345mA, 302.023mA, 689.035mA, 565.224mA, 784.454mA... are the current measurement values when the power is on;
[0098] 0.045mA, 0.079mA, 0.095mA, 0.084mA, 0.094mA... are the current measurement values in the off state;
[0099] 4.345mA, 3.023mA, 4.035mA, 4.224mA, 3.454mA... are the current measurement values in the standby state.
[0100] S45, using the above three sets of current data sequences to train the power-on state A and other states B, And in state B, train the system to distinguish between shutdown state and standby state.
[0101] Specifically, in S45, the three sets of current data are used to train the power-on state A and other states B, and the And in state B, train the system to distinguish between shutdown state and standby state. The following steps are also included:
[0102] S451, read training samples The superscript t represents the tth group of vectors, and the prediction is performed
[0103] S452, if represents the predicted current value, If the standard answer is marked, the parameter is updated
[0104] S453, repeat S452 until the gradient drops below 0.0001, and finally obtain the
[0105] The motherboard of the terminal under test generally has two types of problems: leakage and short circuit. For example, a voltage short to ground may occur due to a breakdown of the resistor, capacitor, or inductor or a patch error, or a small resistor may be patched instead of a capacitor, causing leakage. In other words, the abnormal power supply of the motherboard is either a high current caused by a short circuit or a weak leakage caused by the load. Therefore, it is necessary to accurately identify whether these two conditions exist in three states. The details are as follows:
[0106] S5. Calculate the average current and standard current in each state and compare them to determine the leakage conditions in the three states.
[0107] The comparison formula of the average current and the standard current in each state in S5 is as follows:
[0108]
[0109] in, It represents the average current in the case of no leakage, and |δ| represents the threshold radius of n leakage-free machines, or tolerance.
[0110] According to another embodiment of the present invention, Figure 3 As shown, a sensing system for analyzing the working status of smart terminals through current is provided. The system includes an application software layer, a hardware layer, and a semi-finished machine layer to be tested. The application software layer includes an identification program, the hardware layer includes a computer and a programmable power supply, and the identification program runs on the computer. The semi-finished machine layer to be tested includes several terminals to be tested.
[0111] Among them, the recognition program is composed of Kalman filters with different step sizes and multi-classification perceptron learning algorithms and predictions, which are used to predict the three basic states of the terminal under test: shutdown current and state, startup current and state, and standby current and state; before prediction, dozens or even hundreds of sets of manually labeled golden data (current measurement values under three states) need to be trained to obtain a prediction model.
[0112] In specific applications, the computer is physically connected to a programmable power supply, and a recognition program runs on the computer. This program can control the programmable power supply in real time and read the voltage and current of the programmable power supply. The program then uses an algorithm to predict and identify whether the current of the terminal under test is abnormal. When a worker directly plugs the power cord from the programmable power supply into the assembled semi-finished terminal, the recognition program begins to measure the terminal's shutdown current. After the recognition program obtains the shutdown current in real time, the machine learning algorithm can predict that the terminal under test is currently in shutdown mode. When the worker presses and holds the terminal's power button, the terminal starts up and consumes more current. The recognition program obtains the terminal's startup current in real time during the startup process, and the machine learning algorithm can predict that the terminal under test is currently in power-on mode. When the worker briefly presses the terminal's power button to force the terminal into standby mode (or waits for the screen to time out and turn off, which is also acceptable), the operating system enters standby mode after a brief adjustment. The machine learning algorithm can predict that the terminal under test is currently in standby mode and may be awakened at any time by a single click, turning on the screen and performing facial recognition.
[0113] When the identification program is executed, the following steps are implemented:
[0114] S1. Using floating-point data with preset precision to represent the current value of the programmable power supply obtained in real time;
[0115] S2, forming a time series data stream from the floating point data according to a preset time series rule;
[0116] S3, performing Kalman filtering on the time series data stream;
[0117] S4. Use the perceptron classification method to identify the current and predict the three states of the terminal under test;
[0118] S5. Calculate the average current and standard current in each state and compare them to determine the leakage conditions in the three states.
[0119] To sum up, with the help of the above-mentioned technical scheme of the present invention, by adopting Kalman filtering with different step lengths and multi-classification perceptron learning algorithm and prediction, it is possible to predict the three basic states of the shutdown current and state, the startup current and state, and the standby current and state of the terminal under test, realize automatic analysis, reduce the occurrence of manual judgment, and thus effectively solve the errors in manual test current data. At the same time, it also greatly reduces the labor workload and operation complexity of production line workers, thereby effectively improving production line efficiency, improving accuracy, and improving product quality. In addition, compared with traditional machine recognition algorithms, the present invention adopts Kalman filtering, which effectively solves the problem of filtering peak current during current testing, solves the interference of peak current on the detection of the three states of the terminal, and adopts a multi-classification perceptron algorithm to identify current, so that the three states of shutdown, startup and standby of the terminal under test can be predicted.
[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A sensing method for analyzing the working status of a smart terminal through current, characterized in that: The perception method includes the following steps: S1, using floating point data with preset precision to represent the current value of the program-controlled power supply obtained in real time; S2, forming a time series data stream from the floating point data according to a preset time series rule; S3, performing Kalman filtering on the time series data stream; The Kalman filter processing of the time series data stream in S3 includes the following steps: S31, adopt the collection Represents the time series data stream, let represents the vector at time t, and uses F t Represents the prediction matrix, and the current estimate at time t is calculated as: S32, add the total current consumption of each component in the terminal under test, add them up and draw power from the power line, based on the fact that the power consumption of each component changes dynamically at all times, and get the increment at time t as a vector Then the estimated current at time t is: Among them, B t Represents the dynamic current change increment matrix; S33, the measured peak current and the measured interference of the programmable power supply are collectively referred to as interference, and the covariance matrix Q t Represents the interference current noise, then the total covariance matrix P t It is expressed as: S34, combined set Using matrix H t As each t moment takes out the set The conversion matrix is used to calculate the estimated distribution of the current consumption of the terminal under test at time t: Among them, H t is the identity matrix, H t T H t The transposed matrix of , Σ represents the covariance of the predicted Gaussian distribution; S35, combined with the interference covariance matrix fusion part, calculate and obtain: in, represents the predicted current value at time t, Indicates that at time t The iteration of represents the distribution mean, K represents the Kalman gain matrix, K′ represents the Kalman gain iteration, R t represents the noise covariance, P t ′ represents the iteration result; S4. Use the perceptron classification method to identify the current and predict three states of the terminal under test; In S4, the current is identified by using a perceptron classification method, and the three states of the terminal under test are predicted, including the following steps: S41, adopt the collection Represents the time series data stream, let Represents the vector at time t, let the vector is the weight at time t, then we get; Where b represents the bias; S42, add bias b to vector Get And in the vector Add 1 to get The above calculation formula Converting to vector form, we get: S43, with formula It is a decision boundary model that predicts a binary classification problem. The calculation formula for the current prediction value is as follows: S44, marking the current data in the power-on state, the power-off state and the standby state respectively; S45, using the above three sets of current data sequences to train the power-on state A and other states B, And in state B, train the system to distinguish between shutdown state and standby state. In S45, the three sets of current data are used to train the power-on state A and other states B to obtain And in state B, train the system to distinguish between shutdown state and standby state. The following steps are also included: S451, read training samples The superscript t represents the tth group of vectors, and the prediction is performed S452, if represents the predicted current value, represents the standard answer of the annotation, then update the parameter S453, repeat S452 until the gradient drops below 0.0001, and finally obtain the S5. Calculate the average current and standard current in each state and compare them to determine the leakage status in the three states; The comparison formula of the average current and the standard current in each state in S5 is as follows: in, It represents the average current in the case of no leakage, and |δ| represents the threshold radius of n leakage-free machines, or tolerance.
2. The sensing method for analyzing the working status of a smart terminal through current according to claim 1, characterized in that: The current value of the programmable power supply in S1 is obtained every 50 milliseconds, and the preset accuracy is 0.001mA; The step of using floating point data with preset precision to represent the current value of the programmable power supply acquired in real time in S1 also includes the following steps: Arrange the acquired current values into pairs in chronological order, and continuously add the newly acquired current values to the end of the queue.
3. The sensing method for analyzing the working status of a smart terminal through current according to claim 1, characterized in that: The S31x t-1 、x t and x t+1 All conform to the Gaussian distribution, that is, each component has a mean μ and variance δ 2 , indicating uncertainty; the interference current noise Q in S33 t It also conforms to the Gaussian distribution.
4. The sensing method for analyzing the working status of a smart terminal through current according to claim 1, characterized in that: The current data marked in S44 includes the current measurement value in the power-on state, the current measurement value in the power-off state and the current measurement value in the standby state.
5. A sensing system for analyzing the working state of a smart terminal through current, used to implement the steps of the sensing method for analyzing the working state of a smart terminal through current as described in any one of claims 1 to 4, characterized in that: The system includes an application software layer, a hardware layer and a semi-finished machine layer to be tested; the application software layer includes an identification program, the hardware layer includes a computer and a programmable power supply, and the identification program runs on the computer, and the semi-finished machine layer to be tested includes a number of terminals to be tested; The identification program is composed of Kalman filtering with different step lengths and multi-classification perceptron learning algorithms and predictions, which are used to predict the three basic states of the terminal under test: shutdown current and state, startup current and state, and standby current and state; When the identification program is executed, the following steps are implemented: S1, using floating point data with preset precision to represent the current value of the program-controlled power supply obtained in real time; S2, forming a time series data stream from the floating point data according to a preset time series rule; S3, performing Kalman filtering on the time series data stream; S4. Use the perceptron classification method to identify the current and predict three states of the terminal under test; S5. Calculate the average current and standard current in each state and compare them to determine the leakage conditions in the three states.
6. The sensing system for analyzing the working status of a smart terminal through current according to claim 5, characterized in that: The recognition program also includes, before prediction, training using several groups of manually annotated current measurement values in three states to obtain a prediction model.
