A multilayer ceramic capacitor defect prediction device

Abstract

A multilayer ceramic capacitor defect prediction device comprises a heating temperature control unit, a detection unit, a model acquisition unit, a data fitting module, a durability prediction module, and a processing unit. The heating temperature control unit is formed with a heating zone to perform temperature variation cycles on multiple batches of multilayer ceramic capacitors to be inspected. The detection unit is used to obtain the corresponding electrical values of the capacitors. The model acquisition unit collects these electrical values, and the data fitting module generates a model fitting curve based on the electrical values using a peak identification algorithm. It then calculates a threshold through regression standard deviation. The durability prediction module uses the threshold and employs a bidirectional hetero-associative memory (BHAM) model to classify the defect levels corresponding to the electrical values of the multilayer ceramic capacitors. Finally, the processing unit predicts the service life of the capacitors based on their defect level classification.

Description

[Technical Field]

[0001] This invention relates to a defect prediction device for multilayer ceramic capacitors, and more particularly to a multilayer ceramic capacitor defect prediction device that directly obtains various parameters of multilayer ceramic capacitors through a high and low temperature test furnace and predicts defects through an artificial intelligence model. [Previous Technology]

[0002] With the advancement of technology, electronic products are becoming increasingly diverse and widespread. Since each electronic product requires a varying number of passive components, capacitors have the largest market share and can be categorized into ceramic capacitors, aluminum capacitors, and tantalum capacitors, among others. Currently, the most popular MLCC (Multi-layer Ceramic Capacitor) is also known as a surface-mount capacitor. Multi-layer ceramic capacitors on the market are available with various discharge times and current ratings.

[0003] Before leaving the factory, dozens of multilayer ceramic capacitors must be sampled from each batch and subjected to multiple reliability tests (life tests) such as thermal shock tester / high temperature and humidity tester / temperature cycling tester. The long-term sampling test requires manual sampling, measurement records, data classification and organization, and final judgment by engineers. The operation is quite complicated and labor-intensive. [Summary of the Invention]

[0004] The present invention provides a defect prediction device for multilayer ceramic capacitors, which can predict defects by obtaining various test values ​​of multilayer ceramic capacitors through a multi-channel rapid detection method in a high and low temperature test furnace.

[0005] A defect prediction device for multilayer ceramic capacitors according to the present invention includes: at least one heating temperature control unit, wherein at least one heating zone is formed on the heating temperature control unit for performing at least one temperature change cycle on multiple batches of multilayer ceramic capacitors to be inspected; a detection unit for obtaining an electrical value corresponding to the multiple batches of multilayer ceramic capacitors to be inspected; at least one model acquisition unit connected to the detection unit for collecting the electrical values, wherein the model acquisition unit further includes an electrical value acquisition unit and a prediction model; a numerical fitting module connected to the model acquisition unit for generating a model fitting curve based on the electrical values ​​through a peak discrimination algorithm, and then generating a threshold through a regression standard deviation calculation; a durability prediction module connected to the numerical fitting module for classifying a defect level of the multilayer ceramic capacitor corresponding to the electrical value based on the threshold through a bidirectional hetero-associative memory (BHAM) model; and a processing unit for predicting a service life of the multilayer ceramic capacitor based on the defect level classification.

[0006] In one embodiment of the present invention, the above-mentioned multilayer ceramic capacitor defect prediction device further includes a plurality of carrier chassis disposed in the heating zones for loading the plurality of batches of multilayer ceramic capacitors to be inspected, and the carrier chassis are, but are not limited to, a printed circuit board (PCB).

[0007] In one embodiment of the present invention, the above-mentioned multilayer ceramic capacitor defect prediction device further includes at least one memory unit, which is electrically coupled to the model acquisition units and used to store the electrical values, wherein the memory unit further includes a bidirectional heterogeneous associative memory model database.

[0008] In one embodiment of the present invention, the temperature range in which the above-mentioned heating temperature control units perform the temperature change cycles is between -60 degrees and 180 degrees.

[0009] In one embodiment of the present invention, the detection unit is used to obtain the electrical values, including a loss tangent parameter, an insulation resistance value parameter, a high voltage withstand parameter, an aging rate parameter, and a measurement frequency parameter.

[0010] In one embodiment of the present invention, the detection unit described above is used to obtain these electrical values, which further include a resonant frequency parameter, an equivalent series resistance parameter, an equivalent series inductance parameter, a DC bias voltage parameter, and a thermal shock parameter.

[0011] In one embodiment of the present invention, the calculation formula of the above-mentioned peak identification algorithm is: ;

[0012] where i=1,…,n; k=1,…,m; is the value of the independent variable corresponding to the i-th observation; represents the i-th observation; and is the random error.

[0013] In one embodiment of the present invention, the linearity of the above-mentioned model fitting curve is calculated by a mathematical strength model to obtain a determination coefficient. The calculation formula of the mathematical strength model is: ; ; ;

[0014] Where SSE represents error variance; SST represents total variance; is the variance in total variance that can be explained by a regressive pattern, which comes from the total variance of the dependent variable Y. For example, Y1, Y2, ..., Y10 represent observations with 10 Y values, and the total variance of the variable Y is 10.

[0015] In one embodiment of the present invention, the above-mentioned durability prediction module sets the threshold value through the threshold value of the regression standard deviation and converts it into a binary value, the formula of which is: ;

[0016] where n represents the sample data; and and represent the sample mean.

[0017] In one embodiment of the present invention, the above-mentioned durability prediction module inputs the binary value into the bidirectional heterogeneous associative memory model for calculation, and classifies the output result into different defect level classifications.

[0018] The advantage of this invention is that, by using this multilayer ceramic capacitor defect prediction device and an automated high and low temperature testing furnace, various data of multiple batches of multilayer ceramic capacitors can be measured in real time, a defect prediction model can be established, the quality of the capacitors can be judged in advance, and combined with the real-time measured data and the defect prediction model, a new type of comprehensive reliability test can be provided, which can judge the life test of all products, thereby improving the yield of multilayer ceramic capacitors and reducing waiting time waste and ineffective energy consumption.

Implementation Method

[0019] In order to make the above-mentioned features and advantages of the present invention more apparent and understandable, embodiments are described below in detail with reference to the accompanying drawings.

[0020] Figure 1 is a block diagram of a multilayer ceramic capacitor defect prediction device according to the present invention. In Figure 1, a multilayer ceramic capacitor defect prediction device includes: at least one heating temperature control unit 11, which has at least one heating zone formed thereon for performing at least one temperature change cycle on multiple batches of multilayer ceramic capacitors to be inspected; a detection unit 12 for acquiring an electrical value corresponding to the multiple batches of multilayer ceramic capacitors to be inspected; at least one model acquisition unit 13, connected to the detection unit 12, for collecting the electrical values, the model acquisition unit 13 further including an electrical value acquisition unit and a prediction model; a numerical fitting module 14, connected to the model acquisition unit 13, which, based on the electrical values, generates a model fitting curve using a peak discrimination algorithm, and then generates a threshold by calculating a regression standard deviation; and a durability prediction module 15, connected to the numerical fitting module 14, which, based on the threshold, uses a bidirectional hetero-associative memory... A memory (BHAM) model is used to classify the defect level of the multilayer ceramic capacitor corresponding to the electrical value; and a processing unit 16 predicts the service life of the multilayer ceramic capacitor based on the defect level classification.

[0021] In this embodiment, the multilayer ceramic capacitor defect prediction device further includes a plurality of carrier chassis disposed in the heating zones for loading the plurality of batches of multilayer ceramic capacitors to be inspected. The carrier chassis are, but are not limited to, a printed circuit board (PCB).

[0022] In this embodiment, the heating temperature control unit 11 can also be used to adjust the humidity state of the heating zones.

[0023] In this embodiment, the heating zones may be divided into, but are not limited to, high temperature zones, low temperature zones, and humidity circulation zones, for at least one temperature change cycle.

[0024] In this embodiment, the multilayer ceramic capacitor defect prediction device further includes at least one memory unit, which is electrically coupled to the model acquisition unit 13 and used to store the electrical values, wherein the memory unit further includes a bidirectional heterogeneous associative memory model database.

[0025] The bidirectional heterogeneous associative memory model database is either a physical database or a cloud database.

[0026] In this embodiment, the temperature range in which the heating temperature control units 11 perform these temperature change cycles is between -60 degrees and 180 degrees.

[0027] The heating temperature control unit 11 is installed in a programmable high-temperature furnace.

[0028] In this embodiment, the detection unit 12 is used to obtain the electrical values, including a loss tangent parameter, an insulation resistance value parameter, a high voltage withstand parameter, an aging rate parameter, and a measurement frequency parameter.

[0029] In this embodiment, the detection unit 12 is used to obtain these electrical values, including a resonant frequency parameter, an equivalent series resistance parameter, an equivalent series inductance parameter, a DC bias voltage parameter, and a thermal shock parameter.

[0030] Specifically, the detection unit 12 can be used to measure multiple electrical characteristics of multilayer ceramic capacitors, including withstand voltage, capacitance tolerance, insulation resistance (IR), and dissipation factor (DF), etc., and the present invention is not limited thereto. The detection unit 12 can be a single device with the above-mentioned multiple electrical characteristic measurement functions, or it can be composed of multiple instruments, such as high resistance meter, insulation resistance tester, inductance, capacitance and resistance meter (LCR) tester, impedance analyzer, etc., which are instruments for measuring related electrical characteristics.

[0031] In this embodiment, the calculation formula of the peak identification algorithm is: ;

[0032] where i=1,…,n; k=1,…,m; is the value of the independent variable corresponding to the i-th observation; represents the i-th observation; and is the random error.

[0033] In this embodiment, the linearity of the model fitting curve is calculated using a mathematical strength model to obtain a coefficient of determination. The calculation formula for the mathematical strength model is: ; ; ;

[0034] Where SSE represents error variance; SST represents total variance; is the variance in total variance that can be explained by a regressive pattern, which comes from the total variance of the dependent variable Y. For example, Y1, Y2, ..., Y10 represent observations with 10 Y values, and the total variance of the variable Y is 10.

[0035] The determination coefficient is used in statistics to measure the proportion of the variation of strain numbers that can be explained by the independent variable, thereby judging the explanatory power of the regression model. It represents the percentage of total variation that can be explained by the regression model and is an indicator of the goodness of fit of the regression model, so as to reduce the false rate (FR).

[0036] Among them, the larger the value, the more the measured multilayer ceramic capacitor meets the factory standard.

[0037] In this embodiment, the durability prediction module 15 sets the threshold through the threshold of the regression standard deviation and converts it into a binary value, the formula of which is: ;

[0038] where n represents the sample data; and represent the sample mean.

[0039] In this embodiment, the durability prediction module 15 inputs the binary value into the bidirectional heterogeneous associative memory model for calculation, and classifies the output results into different defect level categories.

[0040] Figure 2 is a schematic diagram of the model fitting curve of a multilayer ceramic capacitor defect prediction device according to the present invention. The numerical fitting module 14 is used to connect the model acquisition unit 13. Based on the electrical values, the peak identification algorithm is used to generate the model fitting curve. Then, the threshold is generated by calculating the regression standard deviation.

[0041] First, the measurement curve 21 is plotted using these electrical values, and the model fitting curve 22 is generated using the peak identification algorithm, as shown in Figure 2.

[0042] In this embodiment, the calculation formula of the peak identification algorithm is: ;

[0043] where i=1,…,n; k=1,…,m; is the value of the independent variable corresponding to the i-th observation; represents the i-th observation; and is the random error.

[0044] In this embodiment, the linearity of the model fitting curve 22 is calculated through a mathematical strength model to obtain a coefficient of determination. The calculation formula of the mathematical strength model is: ; ; ;

[0045] Where SSE represents error variance; SST represents total variance; is the variance in total variance that can be explained by a regressive pattern, which comes from the total variance of the dependent variable Y. For example, Y1, Y2, ..., Y10 represent observations with 10 Y values, and the total variance of the variable Y is 10.

[0046] The determination coefficient is used in statistics to measure the proportion of the variation of strain numbers that can be explained by the independent variable, thereby judging the explanatory power of the regression model. It represents the percentage of the total variation that can be explained by the regression model and is an indicator for judging the goodness of fit of the regression model, so as to reduce the false rate (FR).

[0047] Among them, the larger the value, the more the measured multilayer ceramic capacitor meets the factory standard.

[0048] Figure 3 is a schematic diagram of the defect level classification of a multilayer ceramic capacitor defect prediction device according to the present invention. The durability prediction module 15 is connected to the numerical fitting module 14. Based on the threshold 31, the defect level classification of the multilayer ceramic capacitor corresponding to the electrical value is performed through the bidirectional hetero-associative memory (BHAM) model.

[0049] In this embodiment, the durability prediction module 15 sets the threshold through the threshold of the regression standard deviation and converts it into a binary value, the formula of which is: ;

[0050] Where n represents the sample data; and represents the sample mean.

[0051] Preferably, the threshold is divided into four intervals, as shown in Figure 3.

[0052] In this embodiment, the durability prediction module 15 inputs the binary value into the bidirectional heterogeneous associative memory model for calculation, and classifies the output results into different defect level categories.

[0053] In summary, by using this multilayer ceramic capacitor defect prediction device, an automated high and low temperature testing furnace can be used to measure various data of multiple batches of multilayer ceramic capacitors in real time, establish a defect prediction model, judge the quality of the capacitors in advance, and combine the real-time measured data and the defect prediction model to provide a new type of comprehensive reliability test. It can judge the life test of all products to improve the yield of multilayer ceramic capacitors, reduce waiting time waste and ineffective energy consumption. In fact, the multilayer ceramic capacitor defect prediction device of the present invention can not only be used to predict multilayer ceramic capacitors (MLCCs), but also can be applied to general ceramic capacitors, aluminum capacitors and tantalum capacitors, etc.

[0054] Although the present invention has been disclosed above with reference to the foregoing embodiments, it is not intended to limit the present invention. Any modifications and equivalent substitutions made by those skilled in the art without departing from the spirit and scope of the present invention shall still be within the scope of patent protection of the present invention. [Simplified Explanation of the Diagram]

[0055] Figure 1 is a block diagram of a defect prediction device for multilayer ceramic capacitors according to the present invention.

[0056] Figure 2 is a schematic diagram of the model fitting curve of a defect prediction device for multilayer ceramic capacitors according to the present invention;

[0057] Figure 3 is a schematic diagram of the defect level classification of a multilayer ceramic capacitor defect prediction device according to the present invention.

Claims

1. A defect prediction device for multilayer ceramic capacitors, comprising: The system includes at least one heating temperature control unit with at least one heating zone for performing at least one temperature change cycle on multiple batches of multilayer ceramic capacitors to be inspected; a detection unit for acquiring an electrical value corresponding to the multiple batches of multilayer ceramic capacitors to be inspected; at least one model acquisition unit connected to the detection unit for collecting the electrical values, the model acquisition units further including an electrical value acquisition unit and a prediction model; a numerical fitting module connected to the model acquisition unit for generating a model fitting curve based on the electrical values ​​using a peak discrimination algorithm, and then generating a threshold by calculating a regression standard deviation; a durability prediction module connected to the numerical fitting module for classifying a defect level of the multilayer ceramic capacitor corresponding to the electrical value based on the threshold using a bidirectional hetero-associative memory (BHAM) model; and a processing unit for predicting a service life of the multilayer ceramic capacitor based on the defect level classification.

2. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the multilayer ceramic capacitor defect prediction device further includes a plurality of carrier chassis disposed in the heating zones for loading the plurality of batches of multilayer ceramic capacitors to be inspected, wherein the carrier chassis are, but are not limited to, a printed circuit board (PCB).

3. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the multilayer ceramic capacitor defect prediction device further includes at least one memory unit, which is electrically coupled to the model acquisition units and used to store the electrical values, wherein the memory unit further includes a bidirectional heterogeneous associative memory model database.

4. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the temperature range for the temperature change cycles performed by the heating temperature control units is between -60 degrees and 180 degrees.

5. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the detection unit is used to obtain the electrical values, including a loss tangent parameter, an insulation resistance value parameter, a high voltage withstand parameter, an aging rate parameter, and a measurement frequency parameter.

6. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the detection unit is used to obtain the electrical values, further including a resonant frequency parameter, an equivalent series resistance parameter, an equivalent series inductance parameter, a DC bias parameter, and a thermal shock parameter.

7. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the calculation formula of the peak identification algorithm is: ; where, i=1,…,n; k=1,…,m; represents the value of the independent variable corresponding to the i-th observation; represents the i-th observation; represents the random error.

8. The defect prediction device for multilayer ceramic capacitors as described in claim 6, wherein the model fitting curve is calculated using a mathematical strength model to obtain a determination coefficient, and the calculation formula for the mathematical strength model is: ; ; where, SSE represents error variance; SST represents total variance; it is the variance in total variance that can be explained by a regressive pattern, which comes from the total variance of the dependent variable Y. For example, Y1, Y2, ..., Y10 represent observations with 10 Y values, in which case the total variance of the variable Y is 10.

9. The multilayer ceramic capacitor defect prediction device as described in claim 1, wherein the durability prediction module converts the threshold value into a binary value by setting the threshold value through the regression standard deviation, and the formula is: ; where, n represents the sample data; and represent the sample mean.

10. The multilayer ceramic capacitor defect prediction device as described in claim 9, wherein the durability prediction module inputs the binary value into the bidirectional heterogeneous associative memory model for calculation, and classifies the output result into different defect level categories.

Images