CN109086523B - Automatic generation method of power supply design experiment questions based on cognitive computation model - Google Patents

Abstract

The invention discloses a method for automatically generating power supply design experiment questions based on a cognitive computation model, which comprises the following steps: constructing an integral framework of the cognitive computation model; analyzing each experiment in the PMLK-BUCK by taking the power management experiment suite as a research platform, and excavating a mutual constraint relation among the experiments; based on the constraint relation, specific data are obtained by utilizing WEBENCH simulation and are imported into an SPSS tool to construct a cognitive computation model; developing an intelligent system for automatically generating experimental questions; students select the electric energy conversion performance indexes needing to be researched in the intelligent system, and the system can automatically generate design experiment questions around the indexes. The students independently design an experiment platform according to the questions and finish the experiment on the platform, so that the students can completely master theoretical knowledge points and apply the theoretical knowledge points to actual operation in the process from the design experiment to the completion experiment, and then culture the students according to the standard of engineering certification in the true sense.

Description

Automatic generation method of power supply design experiment questions based on cognitive computing model

technical field

The invention relates to a method for automatically generating experimental questions of power supply design based on a cognitive computing model.

Background technique

The "Washington Accord" (Washington Accord) is an international agreement. Its purpose is to formulate a standard for universities all over the world to train contemporary college students according to this standard. Before this agreement, countries cultivated talents according to their own methods, resulting in Engineers from various countries cannot trust each other, communicate, and cooperate. In order to solve this dilemma, educational organizations led by the United States called on the world to train engineering talents according to a unified standard, and the Washington Accord was born. Since the agreement uses a relatively complete engineering education professional certification system, it is generally recognized internationally, which shows that it is authoritative, professional, and has a high degree of internationalization. In 2016, my country officially became a signatory member of the "Washington Agreement", which means that my country's education level has reached the international advanced level, and international standards can be used to train talents. Therefore, various universities in China have applied to join the agreement, hoping to train engineers in accordance with international standards and improve the quality of engineering and technical personnel training. At the same time, the "Washington Agreement" is also the basis and key to promote the international mutual recognition of Chinese engineer qualifications, which is of great significance for China's engineering technology field to cope with international competition and go global.

The literature "Zhang Jingxian. The role of experimental teaching in the teaching of electrical engineering [J]. Experimental Technology and Management, 2011, 28 (08): 297-298+304." mentioned that experiments are an effective way to help electrical engineering students master professional knowledge . Generally speaking, the abstract knowledge in the course is difficult to understand, but the experiment can make the abstract content concrete, and students can deepen their understanding of abstract knowledge points with practice, so as to quickly master relevant knowledge points and cultivate practical ability. However, in traditional experimental teaching, students are required to go to the laboratory to do the same experiment, and there are ready-made experimental platforms and experimental instructions. Students only need to follow the steps in the experiment guide to complete the experiment easily. But this kind of confirmatory experimental mode is not very effective and cannot fundamentally achieve the real purpose of the experiment. Therefore, it is necessary to develop an automatic experiment generation system, which can automatically generate design questions for students. According to the questions, students can independently design the experimental platform and complete the experiment on this platform. In this way, students have fully mastered the theoretical knowledge points from the design of the experiment to the completion of the experiment.

At the end of 1950, when Joms studied language architecture, he not only believed that syntax is autonomous, but also that language itself is an autonomous modular system. However, due to the limited cognitive ability of human beings, cognitive linguists believe that language cannot be independent of other disciplines, thus producing cognitive model theory. The cognitive process of human beings comes from the perception of the outside world. As the number of times of contact with the same thing increases, the brain begins to store the characteristics of the thing, and summarizes it, and finally morphs, conceptualizes and categorizes the thing.

The computer cognitive model is a model established by simulating the cognitive process of the human brain to things. Different from the cognitive model in the human brain, the computer cognitive model is a mathematical model, which is a quadruple (L ₁ , L ₂ , L, H), where both L ₁ and L ₂ are complete lattices, the elements of L ₁ are called extension elements, the elements of L ₂ are called intension elements, L is the mapping from L ₁ to L ₂ , and H is the mapping from L ₂ to Mapping _of L1. L: L ₁ → L ₂ satisfies: L(0L ₁ )=1L ₂ ; L(1L ₁ )=0L ₁ ; L(a1∪a2)=L(a1)∩L(a2), a1, a2∈L ₁ ; Mapping L is called the extension and intension operator of the cognitive process, and for a∈L ₁ , L(a) is called the intension of a. H: L ₂ →L ₁ satisfies: H(0L ₂ )＝1L ₁ ; H(1L ₂ )＝0L ₁ ; H(b1∨b2)＝H(b1)∧H(b2), b1, b2∈L ₂ ; The mapping H is called the intension-extension operator of the cognitive process, and for b∈L ₂ , H(b) is called the extension of b. The two operators of L and H satisfy: H×L(a)≥a; L×H(b)≥b; therefore, in this cognitive model, the two operators of L and H describe the objects and property change process.

Contents of the invention

In order to solve the above-mentioned technical problems, the present invention provides a method for automatically generating experimental questions of power supply design based on a cognitive computing model with a wide application range.

The technical solution of the present invention to solve the above-mentioned problems is: a method for automatically generating experimental questions of power supply design based on a cognitive computing model, including the following steps:

Step 1: Construct the overall framework of the cognitive computing model;

Step 2: Using the power management experiment kit as the research platform, analyze each experiment in PMLK-BUCK, and dig out the mutual constraints between the experiments;

Step 3: Obtain specific data through WEBENCH simulation based on the mutual constraints between experiments;

Step 4: Import the specific data obtained from the simulation into the SPSS tool to build a cognitive computing model;

Step 5: Build an intelligent system for generating experimental questions based on the cognitive computing model;

Step 6: Generating experimental topics The intelligent system automatically generates experimental topics according to the performance indicators in the field of electric energy conversion selected by the students.

In the above-mentioned method for automatically generating design experiment topics based on cognitive computing models, in the first step, the overall framework of the cognitive computing model includes three independent levels, the first layer is problems and scenarios, including The second layer is the information source, which specifies the source of information; the third layer is the salient features of the project, including elements and constraints. Elements refer to various electric energy parameters. The role of constraints is to automatically generate questions constraints in the process to ensure that the resulting experiments are reliable.

In the method for automatically generating experimental topics for power supply design based on the cognitive computing model, in the third step, the power data obtained from the simulation includes various index values under different power parameters, and the power parameters include input voltage, load Current, switching frequency, input capacitance, equivalent resistance of input capacitor, output capacitor, equivalent resistance of output capacitor, indicators include efficiency, output voltage ripple, inductor current ripple, total loss, crossover frequency, phase margin, low frequency gain.

In the method for automatically generating experimental topics for power supply design based on the cognitive computing model, in the third step, the power data obtained by simulation is imported into the tool SPSS to obtain the Pearson correlation coefficient between each power parameter and each index, and the Pearson correlation The larger the value of the coefficient, the stronger the correlation between the electric energy parameter and the index.

In the above-mentioned automatic generation method of power supply design experiment topics based on the cognitive computing model, the specific steps of the sixth step are: the students select a certain index in the field of electric energy conversion research in the input interface of the experimental topic generation intelligent system in combination with the actual teaching situation; According to the cognitive computing model, the intelligent system of experimental topic generation queries the degree of correlation between the selected indicators and electric energy parameters, and screens out the electric energy parameters whose correlation degree is greater than the threshold value, and randomly selects one of the screened electric energy parameters for generating design experiments topic.

The beneficial effect of the present invention is that: the present invention first establishes a cognitive computing model, and then develops an intelligent system for generating experimental topics according to the cognitive computing model. Parameters, and then randomly select one of the screened electric energy parameters to automatically generate the experimental topic. According to the topic, the students independently design the experimental platform and complete the experiment on this platform. In this way, the students must completely Master relevant theoretical knowledge points and apply them in actual design, and then train students in a true sense in accordance with the standards of engineering certification.

Description of drawings

Fig. 1 is a flowchart in the present invention.

Fig. 2 is a framework diagram of the cognitive computing model in the present invention.

Fig. 3 is a working flow chart of the intelligent system for generating experimental questions in the present invention.

Fig. 4 is an operation diagram of the input interface of the intelligent system for generating experimental questions in the present invention.

Fig. 5 is a schematic diagram of output results of the intelligent system for generating experimental questions in the present invention.

Fig. 6 is a circuit diagram of an embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

As shown in Figure 1, a method for automatically generating experimental questions for power supply design based on cognitive computing models includes the following steps:

Step 1: Construct the overall framework of the cognitive computing model.

The overall framework of the cognitive computing model includes three independent levels. The first level is problems and scenarios, including various indicators in the field of electric energy conversion; the second level is information sources, which specify information sources, and information sources can be Specific to a specific problem, it can also be a general problem, which is applicable to many types of problems; the third layer is the salient features of the project, including elements and constraints. Elements refer to various electric energy parameters. Limits are imposed during the generation process to ensure that the generated experiments are reliable.

Step 2: Using the power management experiment kit as the research platform, analyze each experiment in PMLK-BUCK, and dig out the mutual constraints between the experiments.

This paper uses TI-PMLK (power management experiment kit) as the research platform. PMLK is a development platform for power electronics experiment teaching designed by TI company invited by Stanford University, mainly to serve their university plan. The PMLK kit includes three experimental boards, namely the BUCK experimental board, the BOOST experimental board and the LDO experimental board. Each experimental board includes 6 verification experiments, through which students can understand some basic concepts of power supply such as ripple, efficiency and transient response. This kit can be used not only for students majoring in electrical related majors to understand the basic knowledge of power supplies, but also for teaching professional courses such as switching power supply design. Each experimental board is equipped with an experimental instruction book, which introduces the experimental principle, steps and result analysis in detail. The experimental generation system developed in this project can automatically generate innovative experimental topics by using the cognitive computing model on the basis of the original experiment. Educators require students to complete innovative experiments, which can enable students to have a deep understanding of power from theory to practice. Then, in a true sense, students are trained in accordance with the standards of engineering certification.

In the third layer of the cognitive computing model, it is necessary to give experiment generation setting rules in order to ensure that the generated experiments are valid and reliable. According to the characteristic points of complex engineering problems, that is, there are correlations between sub-problems and conflicts between multiple factors, this rule is the so-called constraint relationship. PMLK itself is a typical complex engineering problem. After fully analyzing each experiment in PMLK-BUCK, the mutual constraint relationship between each experiment was excavated, and a constraint relationship table was summarized to provide cognition for the cognitive computing model. rule. The constraint relationship is shown in Table 1, where 11 is the constraint relationship between Experiment 1 and Experiment 1, 12 is the constraint relationship between Experiment 1 and Experiment 2, and so on, and 66 is the constraint relationship between Experiment 6 and Experiment 6.

Table 1

Among them, 21: the input capacitor Cin and the output capacitor Cout have power loss; so Cin and Cout have an impact on the efficiency, and the ripple also has an impact on the efficiency.

12: Load current, the change of switching frequency will affect the input current ripple. The input current ripple is proportional to the load current and inversely proportional to the switching frequency.

31: The switching frequency has loss on the power components, so setting the switching frequency will also affect the efficiency.

13: The input voltage has an influence on the load transient response.

41: Inductor magnetic core material and winding have power loss _PL , W=ESRLI ² α _PP , C=K ₁ f ^x (K ₂ Δi _PP ) ^y , which affects efficiency.

14: Input voltage and load current have an influence on the inductor current ripple, and under constant load current, different magnetic core materials have different effects on the inductor current ripple.

15: The input voltage and load current will affect the current limit function.

51: The current limit function determines the maximum current that the buck regulator can supply and thus the minimum value of the efficiency.

61: Experiment 1 has no constraint relationship with Experiment 6.

16: The input voltage greatly affects the switching frequency, while the tolerance and uncertainty of the output capacitor ESR may cause the switching frequency to have a different value than expected.

32: Setting the switching frequency has an effect on the inductor current ripple.

23: Both switching frequency and output voltage have an impact on load transient response.

42: Under constant load current, different magnetic core materials have different effects on output voltage ripple. The same kind of magnetic core material has different saturation levels on the output voltage ripple of the regulator.

24: Input and output voltages, switching frequency, and load current can cause the ripple of the current in the inductor L to differ from the expected triangular waveform, thus also affecting the ripple of the output capacitor C _out .

52: As the load current increases, the inductor tends to be deeply saturated, and the ripple of the inductor current increases. If the output capacitor value does not change, it will cause the output voltage ripple to increase.

25: Input voltage V _in , switching frequency fs and inductance L will affect the average load current, peak-to-peak value of the inductor current, and thus affect the current limit behavior. Larger output capacitors produce lower output voltage ripple, which may increase the current limit level. The sensitivity of its current limit to the size of the output capacitor depends on the voltage loop gain crossover frequency.

62: Experiment 6 has no constraints on Experiment 2.

26: The equivalent resistance ESR of the output capacitor has an influence on the transient response of the hysteresis voltage regulator.

43: The saturation of the inductance affects the crossover frequency and also affects the transient response of the load.

34: Experiment 3 has no constraints on Experiment 4.

53: The core saturation of different types of inductors will affect the load transient response.

35: Switching frequency and output capacitance affect buck regulator current limit based on inductor type. A higher switching frequency should increase the level of current limit because the magnitude of the inductor current ripple will be smaller and then the control voltage level will be smaller for a given load current. Larger output capacitors result in lower output voltage ripple, which may increase the current limit level.

36: The characteristics of the output capacitor have an impact on the output voltage DC accuracy, output voltage ripple, switching frequency, input voltage, and load current of the hysteresis buck regulator.

63: The output capacitor has an effect on the load transient response.

54: The degree of saturation of the inductor magnetic core has an impact on the equivalent value of the inductor and the ripple of the inductor power supply.

45: Depending on the type of core material, as the current increases, the inductance may saturate at high currents, and the inductance value of the inductance of different core types decreases to different degrees during saturation, so the type of inductance will affect the current limit behavior .

64: Input voltage and output voltage, switching frequency and load current can make the ripple of the current in the inductor L different from the expected triangular waveform, which will affect the output voltage ripple.

46: Experiment 4 has no constraint relationship to Experiment 6.

56: Experiment 5 has no constraint relationship to Experiment 6.

65: Experiment 6 has no constraint relationship to Experiment 5.

Step 3: Obtain electrical energy data through simulation, and build a cognitive computing model using the tool SPSS based on the electrical energy data and the constraint relationship obtained in Step 1.

The power data obtained by simulation includes various index values under different power parameters, and the power parameters include input voltage, load current, switching frequency, input current, equivalent resistance of input capacitor, output capacitor, output capacitor, etc. Effective resistance, the indicators include efficiency, output voltage ripple, inductor current ripple, total loss, crossover frequency, phase margin, and low frequency gain.

The constraint relationship between each experiment in PMLK is only a macroscopic representation method. It is necessary to convert these constraint relationships into the form of data and serve the information source of the second layer of the cognitive computing model framework. The information source in the second layer The data provided for the automatic generation of the topic provides content, and the data in the current information source is obtained by WEBENCH simulation. The data in the information source is shown in Table 2.

Table 2

Send the power data obtained from the simulation and the constraint relationship obtained in step 1 into the tool SPSS software to obtain the Pearson correlation coefficient between each power parameter and each index, as shown in Table 3.

table 3

The numbers in Table 3 are the Pearson correlation coefficients, which constitute a cognitive computing model. The larger the value of the Pearson correlation coefficient, the stronger the correlation between the electric energy parameters and the indicators. Using this cognitive computing model can be used for the next Chapter's experimental generation intelligent system automatically generates experimental questions to provide a theoretical basis.

Step 4: Build an intelligent system for generating experimental questions based on the cognitive computing model.

The design purpose of the intelligent system is to automatically generate experimental questions and automatic project generation. It is the process of using computer technology to generate items through models, and this process can be divided into three steps: 1. Using cognitive computing models to identify the content of item generation. 2. Automatically generate the structured content of the project. 3. Generate items using computer algorithms.

3.1 Use the cognitive computing model to identify the generated content of the topic

Cognitive computing models are used to generate test items. Cognitive computing models generated by automated projects highlight the knowledge, skills, and abilities required to solve problems in specific domains.

To recognize content in cognitive computing models, intelligent systems need to describe the knowledge, content, and reasoning skills required to create problem-solving tasks. The knowledge and skills specified in the cognitive computing model are identified through an inductive process by asking the intelligent system to look up a parent item and then identify and describe information that can be used to create a new item. The parent item is a high-quality topic, Generate robust item analysis results from existing performance tests.

3.2 Automatically generate the structured content of the project

In this step, an item model is developed to specify where content from the cognitive computing model is placed in order to generate new topics. A project model is a template that specifies actions for properties in newly generated items.

3.3 Using computer algorithms to generate questions

Computer-based algorithms place specified cognitive computing model content into a developed project model, subject to specified constraints. Content assembly is performed with computer algorithms because it is a complex constraint logic programming task, especially when large problems involving many information sources of different characteristics are specified in cognitive computing models. A cognitive computing model can be utilized as a general experimental topic generation system to perform content assembly tasks. The Experiment Auto-Generation System is a java-based program that assembles innovative topics using item models of all element combinations specified in the cognitive computing model.

Step 5: Experiment topic generation The intelligent system automatically generates experiment topics according to the indicators in the field of electric energy conversion selected by the students. The specific steps are: combining the actual teaching situation, students select a certain index in the field of electric energy conversion research on the input interface of the experimental topic generation intelligent system; the experimental topic generation intelligent system queries the degree of correlation between the selected index and the electric energy parameters according to the cognitive computing model , and screen out the electrical energy parameters whose correlation degree is greater than the threshold, and randomly select one of the screened electrical energy parameters to generate a new experimental topic. The specific flow chart is shown in Figure 3. For example, if a student wants to study the efficiency of a switching power supply, he only needs to select the performance index "efficiency" in the input interface, as shown in Figure 4. Based on the query of the cognitive computing model, the system obtains that the load current, the equivalent resistance of the output capacitor, and the switching frequency have the greatest impact on the efficiency. At this time, the random function in the system randomly selects one of the factors to generate the questions, so as to ensure that the students get inconsistent questions, thereby avoiding the phenomenon of mutual plagiarism. The random generation of questions is shown in Figure 5.

In order to verify the correctness of the experimental system, the article takes the topic "Analysis and Optimum Design of the Influence of Load Current on Efficiency" generated by the system as an example, designs the circuit according to the topic, and proves the generated topic by simulating and testing the designed circuit Effective and reliable, with TPS54160 as the core controller, the schematic design of the peripheral circuit is shown in Figure 6. Among them, MOSFET has been integrated in TPS54160 inside.

Theoretical efficiency calculation: Each component in the designed circuit has its own power loss. The specific calculation formula is as follows:

1) MOSFET conduction loss:

2) MOSFET switching loss:

P _MOS,SW ＝V _in I _out f _s t _sw (2)

3) MOSFET gate drive loss:

P _MOS,g = Q _g V _dr f _s (3)

Current Sensing:

IC current sense loss:

Diode D1:

Conduction loss:

P _diode = V _f D'I _out (5)

Inductor L1:

1) Winding loss:

2) Core loss:

Capacitive loss:

1) Input capacitor C ₁ loss:

P _Cin ＝ESRC _in I ² _out D'D (8)

2) Loss of output capacitor C ₁₁ :

other:

IC bias:

P _IC =V _in I _μ (10)

definition:

t _ON is the MOSFET conduction time; f _s =1/T _s is the switching frequency; D is the MOSFET duty cycle:

t _ON /T _s =V _out /V _in (11)

D'=1-D (12)

R _ds is the on-resistance of MOSFET; Q _g is the gate charge of MOSFET; T _s is the switching period of MOSFET; T _ON is the conduction time of MOSFET; T _sw is the switching time of MOSFET; f _s is the switching frequency of MOSFET; V _out is Output voltage; I _out is the output current; α _pp is the inductor ripple coefficient; V _f is the forward voltage drop of the diode; L is the inductance value; Δi _pp is the inductor current ripple; ESR _L is the equivalent series resistance of the inductor; ESRC _in is the equivalent series resistance _of the input capacitor; ESRC _out is the equivalent series resistance of the output capacitor; I _μ is the quiescent current of the controller; V _dr is the gate driver voltage of the MOSFET _; Depending on core material and size, switching frequency and temperature.

Inductor current ripple:

Δi _pp =V _out D'/(fs·L) (13)

Inductor ripple factor:

α _pp =1+(Δi _pp /I _out ) ² /12 (14)

4.2 Analysis of data results

The efficiency data obtained through theoretical calculation, simulation and experiment are shown in Table 4;

Table 4

Theoretical loss P _loss = P _MOSP + P _{MOS, SW} + P _{MOS, g} + P _SNS + P _diode + P _{L, W} + P _{L, C} + P _Cin + P _Cout + P _IC , theoretical efficiency η _theo = (V _out I _out -P _loss )/V _in I _in , the experimental efficiency: η _exp =V _out I _out /V _in I _in , the simulation efficiency is directly given by the simulation tool.

Under the same output current, comparing the data of theoretical efficiency η _theo , simulation efficiency η _sim and experimental efficiency η _exp , it is found that the theoretical efficiency and simulation efficiency are almost the same. In principle, the efficiency of the two should be exactly the same, but the simulation cannot be excluded In contrast, the experimental efficiency is about two percentage points lower than the theoretical efficiency of the simulation, because the actual efficiency calculation is relatively rough, that is, η _exp =P _out /P _in . The PCB wiring impedance and the equivalent impedance of passive components are ignored. The analysis of the data in Table 4 shows that the circuit designed based on the automatically generated questions is effective. The generated questions are correct and valid.

Claims (5)

1. A method for automatically generating a power supply design experiment topic based on a cognitive computing model, comprising the following steps: Step 1: Construct the overall framework of the cognitive computing model; Step 2: Using the power management experiment kit as the research platform, analyze each experiment in PMLK-BUCK, and dig out the mutual constraints between the experiments; Step 3: Obtain specific data through WEBENCH simulation based on the mutual constraints between experiments; Step 4: Import the specific data obtained from the simulation into the SPSS tool to build a cognitive computing model; Step 5: Develop an intelligent system for generating experimental questions based on the cognitive computing model; Step 6: Generating experimental topics The intelligent system automatically generates experimental topics according to the performance indicators in the field of electric energy conversion selected by the students. 2. the method for automatically generating experimental questions based on cognitive computing model for power supply design according to claim 1, characterized in that, in said step 1, the overall framework of cognitive computing model includes three independent levels, the first The first layer is problems and scenarios, including various problems in the field of electric energy conversion; the second layer is information sources, which specify information sources; the third layer is salient features of the project, including elements and constraints, and elements refer to various The function of the constraint conditions is to restrict the automatic generation of the problem, so as to ensure the reliability of the generated experiment. 3. The method for automatically generating the subject of a power design experiment based on a cognitive computing model according to claim 1, characterized in that, in said step 3, the power data obtained by simulation includes various indicators under different power parameters value, the electric energy parameter includes input voltage, load current, switching frequency, input capacitance, equivalent resistance of input capacitance, output capacitance, equivalent resistance of output capacitance, index includes efficiency, output voltage ripple, inductor current ripple , total loss, crossover frequency, phase margin, low frequency gain. 4. the automatic generation method of the power supply design experiment title based on cognitive computing model according to claim 3, it is characterized in that, in described step 4, the electric energy data that simulation obtains is imported in the tool SPSS, obtains each electric energy parameter The Pearson correlation coefficient with each index, the larger the value of the Pearson correlation coefficient, the stronger the correlation between the electric energy parameter and the index. 5. The method for automatically generating experimental topics for power supply design based on cognitive computing models according to claim 4, characterized in that, the specific steps of said step six are: students combine the actual teaching situation to generate the intelligent system in the experimental topics. The input interface selects a certain index in the field of electric energy conversion; the experimental topic generation intelligent system queries the degree of correlation between the selected index and the electric energy parameter according to the cognitive computing model, and screens out the electric energy parameters whose correlation degree is greater than the threshold value. One of the electric energy parameters is randomly selected to generate design experiment questions.

Images