Pure digital fluxgate current sensor based on magnetization residence time and neural network

Abstract

The invention discloses a pure digital fluxgate current sensor based on magnetization residence time and a neural network. The sensor is composed of a composite amplifier enhanced Howland current source, a fluxgate probe, a signal conditioning circuit, a time sequence trigger counter and a neural network analyzer. The fluxgate probe is composed of an excitation coil, an induction coil, a non-magnetic framework and a magnetic core, the non-magnetic framework is designed to be of an annular structure, the excitation coil and the induction coil are alternately wound on the annular framework, and the magnetic core is arranged in a groove of the annular framework. The composite amplifier enhanced Howland current source can generate excitation current required by the fluxgate current sensor system. In the whole signal processing process, the output of the digital signal can be obtained without any digital-to-analog conversion circuit, so that the complexity of circuit design is reduced, the signal-to-noise ratio of the output signal is improved, and the utilization efficiency of the output signal is greatly improved.

Description

Technical Field

[0001] The invention belongs to the field of fluxgate technology and current sensors, and in particular relates to a pure digital fluxgate current sensor based on magnetization retention time and a neural network. Background Art

[0002] Fluxgate current sensors play a vital role in many electronic devices. With the widespread popularity of electric vehicles in the automotive field, they are very dependent on accurate current in battery management systems, motor control, charging systems, vehicle electrical systems, etc. At the same time, the environment will also affect the accuracy of current measurement, resulting in measurement accuracy that does not meet the requirements. Therefore, achieving accurate measurement of DC and AC currents while having strong anti-interference capabilities has become a research focus.

[0003] To achieve accurate current measurement of fluxgate current sensors, the probe structure is generally improved and the circuit is optimized, such as "A signal processing system for wide-range closed-loop fluxgate current detection" disclosed in CN117907674A, "A probe and single-probe dual-range fluxgate current sensor" disclosed in CN117054733A, and "A closed-loop fluxgate current sensor coil structure" disclosed in CN220626494U.

[0004] CN115684708A "A digital fluxgate current sensor" discloses a digital fluxgate current sensor, including an MCU, an energy supply module, an oscillation module and a current acquisition module, but the MCU includes a digital-to-analog conversion circuit module, which to a certain extent causes a nonlinear distortion relationship between the output signal and the input signal, reduces the signal-to-noise ratio of the output signal, and increases the complexity of the circuit design.

[0005] At present, the accuracy of fluxgate current sensors is greatly affected by external interference. The surrounding magnetic field, temperature changes, etc. will affect the current measurement of fluxgate current sensors, resulting in errors in the measurement results. There are various problems with the existing current testing methods for fluxgate current sensors. In contrast, the pure digital fluxgate current sensor based on magnetization retention time and neural network has a strong ability to suppress magnetic interference. It has the characteristics of simple structure and high accuracy in measuring current. It is expected to have more extensive applications in power system detection, industrial manufacturing, smart grid and other fields. Summary of the invention

[0006] The purpose of the present invention is to address a series of problems such as the common use of digital-to-analog conversion circuits in existing fluxgate current sensors, which leads to a nonlinear distortion relationship between the output signal and the input signal and increases the complexity of circuit design. Therefore, a pure digital fluxgate current sensor based on magnetization residence time and neural network is proposed.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A pure digital fluxgate current sensor based on magnetization residence time and neural network includes a composite amplifier enhanced Howland current source, a fluxgate probe, a signal conditioning circuit, a timing trigger counter and a neural network analyzer.

[0009] With the cooperation of the low noise amplifier 1, the low noise amplifier 2 and the high current output driver of the composite amplifier enhanced Howland current source, the Howland current source can provide high current output and low output error.

[0010] In the composite amplifier enhanced Howland current source, the excitation voltage source is connected to the ground terminal and resistor three respectively, resistor three is connected to the non-inverting input terminal of the low noise amplifier one and resistor four respectively, resistor one is connected to the inverting input terminal of the low noise amplifier one, resistor two, capacitor one and the ground terminal respectively, the inverting input terminal of the low noise amplifier one is connected to resistor one, resistor two and capacitor one respectively, the output terminal of the low noise amplifier one is connected to capacitor one and the non-inverting input terminal of the large current output driver respectively, the output terminal of the large current output driver is connected to resistor five and resistor six respectively, resistor five is connected to resistor two, resistor six, the inverting input terminal of the large current output driver and the output terminal of the large current output driver respectively, the non-inverting input terminal of the low noise amplifier two is connected to resistor six and the excitation coil respectively, and the output terminal of the low noise amplifier two is connected to the inverting input terminal of the low noise amplifier two and resistor four respectively.

[0011] The fluxgate probe includes an excitation coil, an induction coil, an annular frame, and a magnetic core. The excitation coil and the induction coil are alternately wound on the annular frame, and the magnetic core is placed inside the groove of the annular frame. The excitation coil and the induction coil have the same winding direction.

[0012] Furthermore, the ring skeleton was evenly divided into 10 segments with a radius of 40 mm;

[0013] Furthermore, the number of turns of the excitation coil and the induction coil are both 360 turns;

[0014] Furthermore, the magnetic core can be made of a high permeability soft magnetic material such as amorphous alloy, permalloy, yttrium iron garnet and nickel iron alloy, with a thickness of 20 μm;

[0015] Furthermore, the skeleton groove where the magnetic core is located has a size of 1 mm.

[0016] The composite amplifier enhanced Howland current source is connected to the excitation coil of the fluxgate probe, and the induction coil of the fluxgate probe is connected to the signal conditioning circuit.

[0017] The signal conditioning circuit is composed of a voltage follower, an inverting amplifier, a precision full-wave rectifier circuit and an inverter with a Schmitt trigger input. In the voltage follower, resistor seven is connected to the non-inverting input terminal of the low-noise amplifier three and the induction coil respectively, the output terminal of the low-noise amplifier three is connected to the inverting input terminal of the low-noise amplifier three and resistor eight in the inverting amplifier respectively, resistor eight is connected to the inverting input terminal of the low-noise amplifier four and resistor ten respectively, resistor ten is connected to the inverting input terminal of the low-noise amplifier four, resistor eight and resistor eleven respectively, resistor nine is connected to the non-inverting input terminal of the low-noise amplifier four and the ground terminal respectively, the output terminal of the low-noise amplifier four is connected to resistor ten and resistor twelve and resistor sixteen in the full-wave rectifier circuit respectively, and in the full-wave rectifier circuit, resistor twelve is connected to the low-noise amplifier The inverting input terminal of the low noise amplifier five, the cathode of the diode one and the resistor fourteen, the resistor thirteen are respectively connected to the non-inverting input terminal of the low noise amplifier five and the ground terminal, the output terminal of the low noise amplifier five is respectively connected to the anode of the diode one and the cathode of the diode two, the cathode of the diode one is respectively connected to the inverting input terminal of the low noise amplifier five, the resistor twelve and the resistor fourteen, the anode of the diode two is respectively connected to the cathode of the low noise amplifier six, the resistor fourteen, the resistor fifteen and the resistor sixteen, the resistor seventeen is respectively connected to the inverting input terminal of the low noise amplifier six and the ground terminal, and the output terminal of the low noise amplifier six is ​​respectively connected to the resistor fifteen and the inverter with a Schmitt trigger input.

[0018] The signal conditioning circuit can be used to extract the dwell time of each magnetic state from the analog output signal of the induction coil, and a digital signal output can be realized without any digital-to-analog conversion circuit.

[0019] In order to identify the rectangular wave S out The starting position of the trigger signal S is first obtained. trig When S out Experience decline and S trig = 0, t n0 It can be uniquely determined within the excitation cycle. Then all the peak times can be obtained, and the residence time of each magnetic state can be obtained according to equation (3). The above process can be completed by a timing trigger counter, so the timing trigger counter can realize the timing function by controlling the digital signal input state.

[0020] The neural network parser includes a static neural network parser and a dynamic LSTM neural network parser, which can be used to extract the target current from the residence time.

[0021] The static neural network parser has an input layer, a hidden layer and an output layer. The hidden layer uses the sigmoid activation function to convert the input signal into a nonlinear output, which greatly enhances the expression and learning ability of the neural network. The output layer can directly output the linear relationship learned by the neural network, improving the computing efficiency.

[0022] The dynamic LSTM neural network parser consists of an input layer and a hidden layer, which can be used to process long-term dependencies in time series data. At the same time, it has stronger adaptability when facing different data characteristics and working conditions, and can achieve better performance and generalization capabilities without redesigning the entire model of the fluxgate current sensor, thereby improving detection accuracy and efficiency.

[0023] The beneficial effects of the present invention are:

[0024] Compared with the existing fluxgate products, the fluxgate current sensor based on the residence time effect and neural network designed by the present invention can realize digital signal output without any digital-to-analog conversion circuit, thereby reducing the complexity of circuit design and improving the signal-to-noise ratio of the output signal, greatly improving the utilization efficiency of the output signal. At the same time, compared with the traditional method of measuring current, the neural network method of measuring current can indirectly estimate the current by processing the data collected by the sensor. This method helps to reduce the interference between the measuring equipment and the object being measured. At the same time, the neural network can adapt to various complex current waveforms and change patterns during the learning process. In short, the method of measuring current by the neural network provides a new solution for current measurement through the advantages of non-contact and strong adaptability. Therefore, the fluxgate current sensor based on the residence time effect and neural network can be widely used in the fields of power systems, industrial production and electronic equipment maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the overall structure of the fluxgate current sensor of the present invention;

[0026] Figure 2 It is the circuit diagram of the composite amplifier enhanced Howland current source;

[0027] Figure 3 This is the front view of the fluxgate probe;

[0028] Figure 4 It is a top view of the fluxgate probe;

[0029] Figure 5 It is a schematic diagram of the structure of the annular magnetic core of the fluxgate sensor;

[0030] Figure 6 This is the signal conditioning circuit diagram;

[0031] Figure 7 It is a timing diagram of rectangular output and excitation signal;

[0032] Figure 8 This is the structure diagram of the static neural network parser;

[0033] Fig. 9This is the structure diagram of the dynamic LSTM neural network parser;

[0034] Fig.10 This is the "M" waveform signal diagram;

[0035] Fig.11 A schematic diagram of an implementation scheme using a static neural network parser;

[0036] Fig.12 Schematic diagram of the implementation scheme using a dynamic LSTM neural network parser.

[0037] In the figure, 1. Howland current source module, 111. Excitation voltage source, 121. Low noise amplifier, 122. Low noise amplifier, 131. Resistor 1, 132. Resistor 2, 133. Resistor 3, 134. Resistor 4, 135. Resistor 5, 136. Resistor 6, 141. Capacitor 1, 151. High current output driver, 161. Ground terminal, 2. Fluxgate probe, 21. Excitation coil, 22. Induction coil, 23. Magnetic core, 24. Ring skeleton, 3. Signal conditioning circuit module, 31. Voltage follower, 311. Resistor 7, 312. Low noise amplifier 3, 32. Inverting amplifier, 321. Resistor 8, 322. Resistor 9, 323. Resistor 10, 324. Resistor 11, 325. Low noise amplifier 4, 33. Precision full-wave rectifier circuit, 331. Resistor 12, 332. Resistor 13, 333. Resistor 14, 334. Resistor 15, 335. Resistor 16, 336. Resistor 17, 337. Diode 1, 338. Diode 2, 339. Low noise amplifier 5, 3310. Low noise amplifier 6, 34. Inverter with Schmitt trigger input, 4. Timing trigger counter, 5. Neural network parser, 51. Static neural network parser, 511. Input layer, 512. Hidden layer, 513. Output layer, 52. Dynamic LSTM neural network parser, 521. Input layer, 522. Hidden layer, 5221. Forget gate, 5222. Input gate, 5223. Output gate. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments:

[0039] The fluxgate current sensor based on residence time effect and neural network consists of a composite amplifier enhanced Howland current source 1, a fluxgate probe 2, a signal conditioning circuit 3, a timing trigger counter 4 and a neural network analyzer 5.

[0040] The composite amplifier enhanced Howland current source 1 is connected to the excitation coil 21 of the fluxgate probe 2 , the induction coil 22 of the fluxgate probe 2 is connected to the signal conditioning circuit 3 , and the signal conditioning circuit 3 is connected to the timing trigger counter 4 .

[0041] With the cooperation of the noise amplifier 121 , the low noise amplifier 122 and the high current output driver 151 of the composite amplifier enhanced Howland current source, the Howland current source can provide high current output and low output error.

[0042] The excitation voltage source 111 is respectively connected to the ground terminal 161 and the resistor 3 133, the resistor 3 133 is respectively connected to the non-inverting input terminal of the low noise amplifier 121 and the resistor 4 134, the resistor 131 is respectively connected to the inverting input terminal of the low noise amplifier 121, the resistor 2 132, the capacitor 141 and the ground terminal 161, the inverting input terminal of the low noise amplifier 121 is respectively connected to the resistor 131, the resistor 2 132, the capacitor 141, and the output terminal of the low noise amplifier 121 is respectively connected to the capacitor 141 and the large current output driver The non-inverting input terminal of the driver 151 and the output terminal of the high current output driver 151 are respectively connected to resistor five 135 and resistor six 136, resistor five 135 are respectively connected to resistor two 132, resistor six 136, the inverting input terminal of the high current output driver 151 and the output terminal of the high current output driver 151, the non-inverting input terminal of the low noise amplifier two 122 is respectively connected to resistor six 136 and the excitation coil 21, and the output terminal of the low noise amplifier two 122 is respectively connected to the inverting input terminal of the low noise amplifier two 122 and resistor four 134.

[0043] The fluxgate probe 2 consists of an excitation coil 21, an induction coil 22, a magnetic core 23, and an annular skeleton 24. The excitation coil 21 and the induction coil 22 are alternately wound on the annular skeleton 24, and the magnetic core 23 is placed inside the groove of the annular skeleton 24. The excitation coil 21 and the induction coil 22 have the same winding direction.

[0044] The signal conditioning circuit 3 is composed of a voltage follower 31, an inverting amplifier 32, a precision full-wave rectifier circuit 33 and an inverter 34 with a Schmitt trigger input. In the voltage follower, resistor seven 311 is respectively connected to the non-inverting input terminal of low noise amplifier three 312 and induction coil 22, the output terminal of low noise amplifier three 312 is respectively connected to the inverting input terminal of low noise amplifier three 312 and resistor eight 321 in the inverting amplifier, resistor eight 321 is respectively connected to the inverting input terminal of low noise amplifier four 325 and resistor ten 323, resistor ten 323 is respectively connected to the inverting input terminal of low noise amplifier four 325, resistor eight 321 and resistor eleven 324, resistor nine 322 is respectively connected to the non-inverting input terminal of low noise amplifier four 325 and ground terminal 161, the output terminal of low noise amplifier four 325 is respectively connected to resistor eleven 324 and resistor twelve 331 and resistor sixteen 335 in the full-wave rectifier circuit, in the full-wave rectifier circuit, resistor twelve 331 is respectively connected to low noise amplifier five 339 The inverting input terminal, the cathode of diode 1 337 and resistor 14 333 are connected to the inverting input terminal of low noise amplifier 5 339 and the ground terminal 161 respectively, the output terminal of low noise amplifier 5 339 is connected to the anode of diode 1 337 and the cathode of diode 2 338 respectively, the cathode of diode 1 337 is connected to the inverting input terminal of low noise amplifier 5 339, resistor 12 331 and resistor 14 333 respectively, the anode of diode 2 338 is connected to the cathode of low noise amplifier 6 3310, resistor 14 333, resistor 15 334 and resistor 16 335 respectively, the resistor 17 336 is connected to the inverting input terminal of low noise amplifier 6 3310 and the ground terminal 161 respectively, and the output terminal of low noise amplifier 6 3310 is connected to resistor 15 334 and inverter 34 with Schmitt trigger input respectively.

[0045] Example 1

[0046] In the composite amplifier enhanced Howland current source 1, the excitation voltage source 111 is respectively connected to the ground terminal 161 and the resistor 3 133, the resistor 3 133 is respectively connected to the non-inverting input terminal of the low noise amplifier 121 and the resistor 4 134, the resistor 131 is respectively connected to the inverting input terminal of the low noise amplifier 121, the resistor 2 132, the capacitor 1 141 and the ground terminal 161, the inverting input terminal of the low noise amplifier 121 is respectively connected to the resistor 1 131, the resistor 2 132, the capacitor 1 141, and the output terminal of the low noise amplifier 121 is respectively connected to the capacitor 1 141 and the non-inverting input terminal of the high-current output driver 151, the output terminal of the high-current output driver 151 is respectively connected to resistor five 135 and resistor six 136, resistor five 135 is respectively connected to resistor two 132, resistor six 136, the inverting input terminal of the high-current output driver 151 and the output terminal of the high-current output driver 151, the non-inverting input terminal of the low-noise amplifier 122 is respectively connected to resistor six 136 and the excitation coil 21, and the output terminal of the low-noise amplifier 122 is respectively connected to the inverting input terminal of the low-noise amplifier 122 and resistor four 134.

[0047] The composite amplifier enhanced Howland current source 1 is connected to the excitation coil 21 of the fluxgate probe 2 , the induction coil 22 of the fluxgate probe 2 is connected to the signal conditioning circuit 3 , and the signal conditioning circuit 3 is connected to the timing trigger counter 4 .

[0048] The signal conditioning circuit module 3 is composed of a voltage follower 31, an inverting amplifier 32, a precision full-wave rectifier circuit 33 and an inverter 34 with a Schmitt trigger input. In the voltage follower, resistor seven 311 is respectively connected to the non-inverting input terminal of low noise amplifier three 312 and induction coil 22, the output terminal of low noise amplifier three 312 is respectively connected to the inverting input terminal of low noise amplifier three 312 and resistor eight 321 in the inverting amplifier, resistor eight 321 is respectively connected to the inverting input terminal of low noise amplifier four 325 and resistor ten 323, resistor ten 323 is respectively connected to the inverting input terminal of low noise amplifier four 325, resistor eight 321 and resistor eleven 324, resistor nine 322 is respectively connected to the non-inverting input terminal of low noise amplifier four 325 and ground terminal 161, the output terminal of low noise amplifier four 325 is respectively connected to resistor eleven 324 and resistor twelve 331 and resistor sixteen 335 in the full-wave rectifier circuit, in the full-wave rectifier circuit, resistor twelve 331 is respectively connected to low noise amplifier five 339 The inverting input terminal, the cathode of diode 1 337 and resistor 14 333 are connected to the inverting input terminal of low noise amplifier 5 339 and the ground terminal 161 respectively, the output terminal of low noise amplifier 5 339 is connected to the anode of diode 1 337 and the cathode of diode 2 338 respectively, the cathode of diode 1 337 is connected to the inverting input terminal of low noise amplifier 5 339, resistor 12 331 and resistor 14 333 respectively, the anode of diode 2 338 is connected to the cathode of low noise amplifier 6 3310, resistor 14 333, resistor 15 334 and resistor 16 335 respectively, the resistor 17 336 is connected to the inverting input terminal of low noise amplifier 6 3310 and the ground terminal 161 respectively, and the output terminal of low noise amplifier 6 3310 is connected to resistor 15 334 and inverter 34 with Schmitt trigger input respectively.

[0049] like Fig.10 As shown, the output of the signal consists of positive peaks and negative peaks, and there are positive "M" waves and negative "M" waves. There are two peaks p0 and p1 in the positive "M" wave, and two peaks n0 and n1 in the negative "M" wave.

[0050] Peak time t p and t n They are:

[0051]

[0052] t p0 ,t p1 ,t n1 ,t n0 are the time positions of the p0 peak, p1 peak, n1 peak and n0 peak respectively,

[0053]

[0054] The residence time for each magnetic field state can be given as,

[0055]

[0056] The residence time difference (RTD) between the positive region and the negative region can be expressed as:

[0057]

[0058] Arcsinx≈x near zero, so when H c +|H amb |+|H dut |<<|H exc |, the above formula can be approximately rewritten as:

[0059]

[0060] The current detection sensitivity is:

[0061]

[0062] The linear relationship between the dwell time difference at zero point and the target current can be obtained by equation (6), thereby determining the magnitude of the target current.

[0063] The neural network analyzer 5 includes a static neural network analyzer 51 and a dynamic LSTM neural network analyzer 52, and the neural network analyzer 5 is used to extract the target current from the residence time.

[0064] The hidden layer 512 of the static neural network parser 51 uses a sigmoid activation function to convert the input signal into a nonlinear output, which greatly enhances the expression and learning ability of the neural network. The output layer 513 can directly output the linear relationship learned by the neural network, which improves the calculation efficiency. The neural network can indirectly estimate the current by processing the data collected by the sensor. At the same time, the neural network can adapt to various complex current waveforms and change patterns during the learning process. This significantly improves the linearity of the sensor and obtains the ability to suppress environmental magnetic field interference.

[0065] Example 2

[0066] like Fig. 9 As shown, the signal data containing the time series is obtained in the signal conditioning circuit 3, the data is preprocessed, and the preprocessed training set data is input into the input layer 521 of the dynamic LSTM neural network parser 52. In the forget gate 5221 part, the sigmoid activation function is used to convert the information X of the current time step into t and the short-term information h of the previous time step t-1 Integrate to determine the long-term memory C used to influence the previous time stept-1 The ratio f t In the input gate 5222 part, the tanh activation function is used to determine how much new information is absorbed in the current time step. At the same time, the sigmoid activation function is used and according to the information X t With h t-1 Determine the proportion of i used to generate new information for screening t . The C obtained by the forget gate and the input gate t-1 ×f t and Add them together to get the long-term information C of the next time step t In the output gate 5223, the new long-term information C t Filter out the short-term information h that is most suitable for the current time step t The calculation unit of long-term information C t Standardization is performed to prevent the historical information from becoming larger and larger during the transmission process, and a certain nonlinear property is given to the output gate to ensure the stability of training. To prevent problems such as gradient disappearance, f t The ratio is set between 0 and 1. Dropout is applied between hidden layers to shut down some neurons during data training to prevent data overfitting. The LSTM neural network 52 can effectively process long-term dependencies in time series data, and has stronger adaptability when facing different data features and working conditions, and can achieve better performance and generalization capabilities without redesigning the entire model of the fluxgate current sensor, thereby improving detection accuracy and efficiency.

[0067] The embodiments described above are only preferred specific implementation methods of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of the technical solutions that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belong to the protection scope of the present invention.

Claims

1. A pure digital fluxgate current sensor based on magnetization retention time and neural network, characterized in that: It includes a composite amplifier enhanced Howland current source (1), a fluxgate probe (2), a signal conditioning circuit (3), a timing trigger counter (4) and a neural network analyzer (5); The composite amplifier enhanced Howland current source (1) is connected to the excitation coil (21) of the fluxgate probe (2), the induction coil (22) of the fluxgate probe (2) is connected to the signal conditioning circuit (3), and the signal conditioning circuit (3) is connected to the timing trigger counter (4); The excitation voltage source (111) is respectively connected to the ground terminal (161) and the resistor three (133); the resistor three (133) is respectively connected to the non-inverting input terminal of the low noise amplifier one (121) and the resistor four (134); the resistor (131) is respectively connected to the inverting input terminal of the low noise amplifier one (121), the resistor two (132), the capacitor one (141) and the ground terminal (161); the inverting input terminal of the low noise amplifier one (121) is respectively connected to the resistor one (131), the resistor two (132) and the capacitor one (141); the output terminal of the low noise amplifier one (121) is respectively connected to the capacitor one (141) and the large current input terminal. The output end of the high-current output driver (151) is connected to the in-phase input end of the high-current output driver (151), the output end of the high-current output driver (151) is respectively connected to the resistor five (135) and the resistor six (136), the resistor five (135) is respectively connected to the resistor two (132), the resistor six (136), the inverting input end of the high-current output driver (151) and the output end of the high-current output driver (151), the in-phase input end of the low-noise amplifier two (122) is respectively connected to the resistor six (136) and the excitation coil (21), and the output end of the low-noise amplifier two (122) is respectively connected to the inverting input end of the low-noise amplifier two (122) and the resistor four (134); The fluxgate probe (2) is composed of an excitation coil (21), an induction coil (22), a magnetic core (23), and an annular frame (24); the excitation coil (21) and the induction coil (22) are alternately wound on the annular frame (24); the magnetic core (23) is placed inside a groove of the annular frame (24); and the excitation coil (21) and the induction coil (22) have the same winding direction; The signal conditioning circuit (3) is composed of a voltage follower (31), an inverting amplifier (32), a precision full-wave rectifier circuit (33) and an inverter (34) with a Schmitt trigger input; in the voltage follower, resistor seven (311) is respectively connected to the in-phase input terminal of low noise amplifier three (312) and the induction coil (22); the output terminal of low noise amplifier three (312) is respectively connected to the inverting input terminal of low noise amplifier three (312) and resistor eight (321) in the inverting amplifier; resistor eight (32 1) respectively connect the inverting input terminal of the low noise amplifier 4 (325) and the resistor 10 (323), the resistor 10 (323) respectively connects the inverting input terminal of the low noise amplifier 4 (325), the resistor 8 (321) and the resistor 11 (324), the resistor 9 (322) respectively connects the non-inverting input terminal of the low noise amplifier 4 (325) and the ground terminal (161), the output terminal of the low noise amplifier 4 (325) respectively connects the resistor 11 (324) and the resistor 12 (331) and the resistor 16 (335) in the full-wave rectifier circuit In the full-wave rectifier circuit, the resistor 12 (331) is respectively connected to the inverting input terminal of the low noise amplifier 5 (339), the cathode of the diode 1 (337) and the resistor 14 (333), the resistor 13 (332) is respectively connected to the non-inverting input terminal of the low noise amplifier 5 (339) and the ground terminal (161), the output terminal of the low noise amplifier 5 (339) is respectively connected to the anode of the diode 1 (337) and the cathode of the diode 2 (338), the cathode of the diode 1 (337) is respectively connected to the cathode of the low noise amplifier 5 (339). The inverting input terminal, resistor twelve (331) and resistor fourteen (333), the positive electrode of diode two (338) are respectively connected to the negative electrode of low noise amplifier six (3310), resistor fourteen (333), resistor fifteen (334) and resistor sixteen (335), the resistor seventeen (336) are respectively connected to the inverting input terminal of low noise amplifier six (3310) and the ground terminal (161), and the output terminal of low noise amplifier six (3310) is respectively connected to resistor fifteen (334) and an inverter (34) with a Schmitt trigger input.

2. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 1, characterized in that: The annular skeleton (24) is evenly divided into 10 sections, and the radius is 40 mm.

3. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 1 or 2, characterized in that: The number of turns of the excitation coil (21) and the induction coil (22) are both 360 turns.

4. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 1 or 2, characterized in that: The magnetic core (23) is made of one of amorphous alloy, Permalloy, yttrium iron garnet and nickel iron alloy, and has a thickness of 20 μm.

5. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 3, characterized in that: The magnetic core (23) is made of one of amorphous alloy, Permalloy, yttrium iron garnet and nickel iron alloy, and has a thickness of 20 μm.

6. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 1, 2 or 5, characterized in that: The skeleton groove where the magnetic core (23) is located has a diameter of 1 mm.

7. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 3, characterized in that: The skeleton groove where the magnetic core (23) is located has a diameter of 1 mm.

8. The pure digital fluxgate current sensor based on magnetization residence time and neural network as claimed in claim 4, characterized in that: The skeleton groove where the magnetic core (23) is located has a diameter of 1 mm.

Images