A method for characterizing a strain sensor

Abstract

The present invention discloses a method for characterizing a strain sensor in the technical field of strain sensors. This method for characterizing a strain sensor extracts and separates based on the background noise and input noise of the object to be measured. The main analysis methods include two analysis methods in the time domain and the frequency domain. The time domain analysis method includes an amplitude analysis method and a harmonic analysis method. The frequency domain analysis method includes a spectrum analysis method and a wavelet analysis method. The present invention determines the state of the device after power-off and restart through one or more specific parameters, and then discriminates true and false zeros, eliminating the "pseudo zeros" caused by prestress existing before power-off and restart, which can effectively prevent the reference following algorithm from incorrectly tracking the "pseudo zeros" and affecting the operation of the device before the next power-off and restart, effectively solving the defects existing in the base following algorithm and improving the reliability and consistency of pressure-sensitive interaction.

Description

Technical Field

[0001] The present invention relates to the technical field of strain sensors, and specifically to a method for characterizing strain sensors. Background Art

[0002] The main application scenarios of stress sensors include: high-precision metrology, such as weighing scales, etc.; relatively low-precision sensing, such as pressure-sensitive interaction in consumer electronics, etc.

[0003] Regardless of the application scenario, stress measurement requires that the creep of the object to be measured (including its material and structure) be as small as possible.

[0004] Creep is a phenomenon in which the strain of a solid material increases with time under the condition of constant stress. Due to creep, the stress state of the object to be measured at a certain moment is related not only to the deformation at that instant but also to the deformation process before that instant.

[0005] Currently, there is no unified creep theory applicable to all materials. For metal materials, there are mainly the aging theory, the strengthening theory, and the creep aftereffect theory as follows:

[0006]

[0007] Among them, σ is the stress and ε is the strain.

[0008] The first term is the basic part, and the second term is the aftereffect influence part. K is called the influence function, which is the deformation caused by unit stress at a later time per unit time at a certain moment.

[0009] In actual application scenarios, the main influence of creep is the difference between the actual stress acting on the object to be measured and the measurement result.

[0010] In view of this, high-precision metrology often modifies the material and optimizes the structure of the object to be measured (usually an elastomer) to avoid creep generation within the measurement range as much as possible; while relatively low-precision sensing mostly "eliminates" the influence of creep through algorithms - that is, tracking the object to be measured from beginning to end, and marking the stress application, stress removal, and creep influence (mainly manifested as zero drift), thereby reducing the influence of creep on stress sensing (also known as the reference following algorithm).

[0011] Since there is no stable and constant stress action in actual application scenarios, the above formula is adjusted to:

[0012]

[0013] Among them, E is a constant, which is the elastic modulus or Young's modulus, as Figure 2 shown.

[0014] The reference following algorithm regards the continuously differentiable change in the time domain after algorithm filtering as zero drift, and regards the discontinuous mutation in the time domain as stress action, as Figure 3 shown.

[0015] The reference following algorithm can largely solve the influence of creep on stress sensing and compensate for zero drift. However, in some 1 / 3 edge cases, it is very likely to cause the zero point to "run away" due to overcorrection, as Figure 4 shown,[[]] Figure 4 in,

[0016] Before the system power-off and restart, it happens to be in the prestress state (for example: the brush head of an electric toothbrush is pressed against the tooth surface and then started, the stylus is powered off and restarted under the stress state, etc.). After power-on again, the following algorithm will track the current state as the zero point, resulting in overcorrection of the zero point. Assuming that the input stress before power-off and restart is σ0, the output result after the reference following algorithm is:

[0017]

[0018] The influence of this zero-point overcorrection will continue until the next power-off and restart, seriously affecting the reliability and consistency of pressure-sensing interaction. Summary of the Invention

[0019] The purpose of the present invention is to provide a method for characterizing a strain sensor to solve the problem that before the system power-off and restart, it happens to be in the prestress state (for example: the brush head of an electric toothbrush is pressed against the tooth surface and then started, the stylus is powered off and restarted under the stress state, etc.). After power-on again, the following algorithm will track the current state as the zero point, resulting in overcorrection of the zero point. The influence of this zero-point overcorrection will continue until the next power-off and restart, seriously affecting the reliability and consistency of pressure-sensing interaction.

[0020] To achieve the above purpose, the present invention provides the following technical solution: A method for characterizing a strain sensor, which extracts and separates based on the background noise and input noise of the object to be measured. The main analysis methods include two analysis methods in the time domain and the frequency domain. The time domain analysis method includes an amplitude analysis method and a harmonic analysis method, and the frequency domain analysis method includes a spectrum analysis method and a wavelet analysis method.

[0021] Preferably, the amplitude analysis method includes the following steps:

[0022] S1: When powering on for the first time, automatically obtain the peak-to-peak value of the background noise amplitude and store it in the register;

[0023] S2: Obtain the peak-to-peak value of the total noise amplitude in real time;

[0024] S3: In case of power failure and restart, the peak-to-peak value of the total noise amplitude immediately after the restart is compared with the peak-to-peak value of the background noise amplitude;

[0025] S4: If the difference is higher than the peak-to-peak value of the noise floor amplitude by 8-12%, it is considered that the device was in the input state before the restart, and the initial value in the memory is used as the zero point basis for tracking;

[0026] S5: Otherwise, execute the benchmark automatic following algorithm.

[0027] Preferably, the harmonic analysis method comprises the following steps:

[0028] The periodic law of structural vibration is determined by Fourier series expansion, and the expression is: Among them, n=1 represents the fundamental wave, n>1 represents the harmonic wave, A n is the peak-to-peak amplitude of the nth wave;

[0029] In the local input state, only the fundamental wave exists, that is:

[0030] The first harmonic under load is expressed as:

[0031] By performing Fourier expansion on the real-time data after power failure and restart, we can obtain its fundamental wave and first harmonic, and then identify whether the device is in a background state or a load state, so as to decide whether to execute the following algorithm for the zero point at the moment of power-on.

[0032] Preferably, the spectrum analysis method comprises the following steps:

[0033] In the local input state, the noise frequency is Among them, t max , t min is the relative time when each deformation wave reaches the peak and trough;

[0034] In the input state, the noise frequency is:

[0035] Perform frequency domain transformation on the real-time data after power failure and restart. If the ratio of the difference between the real-time noise frequency and the background noise frequency to the background noise is higher than 8-12%, that is:

[0036]

[0037] If the sensor is in a free state, the reference automatic following algorithm is executed. Otherwise, the register zero point data before power failure is filled in before the reference automatic following is performed.

[0038] Preferably, the wavelet analysis method includes the following steps:

[0039] The set obtained by scaling and translating the mother wavelet and the scaling function has both high frequency and low frequency, and also covers the time domain, as follows:

[0040] f(t) = ∑ m ∑ j a j,k ψ j,k (t), where ψ j,k (t) is the wavelet series;

[0041] By using the wavelet series to locate the signal in both the time domain and the frequency domain, the outputs after denoising under the background state and the load state are respectively as follows:

[0042] f(t, ε = 0) = ∑ k ∑ j a j,k ψ j,k (t, ε = 0);

[0043] f(t, ε) = ∑ k ∑ j a j,k ψ j,k (t, ε);

[0044] That is, the wavelet series of the two are ψ j,k (t, ε = 0) and ψ j,k (t, ε).

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows: This method for characterizing a strain sensor uses one or more specific parameters to determine the state of the device after power-off and restart, and then distinguishes true and false zeros, eliminating the "pseudo-zeros" caused by prestress existing before power-off and restart. It can effectively prevent the reference following algorithm from incorrectly tracking the "pseudo-zeros" and affecting the operation of the device before the next power-off and restart, effectively solving the defects existing in the base following algorithm and improving the reliability and consistency of pressure-sensitive interaction. Brief Description of the Drawings

[0046] Figure 1 Schematic diagram showing the state of the sensor after power-off and restart characterized by the method of the present invention;

[0047] Figure 2 Schematic diagram of the stress-strain curve under the creep state of the present invention;

[0048] Figure 3 Schematic diagram showing the tracking of the stress-strain curve by the reference following algorithm of the present invention;

[0049] Figure 4This is a schematic diagram showing the error following caused by the prestress acting state of the device after the power failure and restart of the present invention. Detailed implementation manners

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0051] The present invention provides a method for characterizing a strain sensor, which determines the state of the device after power failure and restart through one or more specific parameters, and then distinguishes true and false zero points, eliminating the "pseudo zero points" caused by the existence of prestress before power failure and restart. It can effectively prevent the reference following algorithm from incorrectly tracking the "pseudo zero points" and affecting the operation of the device before the next power failure and restart, effectively solving the defects existing in the base following algorithm, improving the reliability and consistency of pressure sensing interaction. This strain sensor characterization method extracts and separates the background noise (all noises irrelevant to the input, mainly composed of the intrinsic noise and background noise of the system) and the input noise of the measured object. The main analysis methods include two analysis methods in the time domain and the frequency domain. The time domain analysis method includes an amplitude analysis method and a harmonic analysis method, and the frequency domain analysis method includes a spectrum analysis method and a wavelet analysis method;

[0052] The amplitude analysis method includes the following steps:

[0053] Assume that the peak-to-peak value of the background noise amplitude of the system is:

[0054] The peak-to-peak value of the input noise amplitude is:

[0055] In the direction of stress input, there is:

[0056] Where represents the peak-to-peak value of the real-time total noise amplitude;

[0057] The characterization process is as follows:

[0058] S1: When powering on for the first time, automatically obtain and store it in the register;

[0059] S2: Obtain the peak-to-peak value of the total noise amplitude in real time

[0060] S3: In case of power failure and restart, compare the peak-to-peak value of the total noise amplitude at the moment after restart with the peak-to-peak value of the background noise amplitude, that is:

[0061] S4: If the above difference is higher than the ratio of 8 - 12%, preferably 10%, of the peak - to - peak value of the background noise amplitude, it is considered that the device was in the input state before restart, and the initial value in the memory is used as the zero - point basis for following;

[0062] S5: Otherwise, execute the reference automatic following algorithm, as Figure 1 shown;

[0063] The harmonic analysis method includes the following steps:

[0064] The structural vibration conforms to a periodic law. Measure the periodic law of the structural vibration and expand it with Fourier series. The expression is: where n = 1 represents the fundamental wave, and n > 1 represents the harmonic. A n is the peak - to - peak value of the amplitude of the nth wave;

[0065] In the local input state, only the fundamental wave exists, that is:

[0066] The first - order harmonic in the load state is expressed as:

[0067] Perform Fourier expansion on the real - time data after power - off restart. The fundamental wave and the first - order harmonic can be obtained, and then the state of the device, whether it is the background state or the load state, can be identified to determine whether to execute the following algorithm for the zero - point at the moment of power - on again;

[0068] The spectrum analysis method includes the following steps:

[0069] In the local input state, the noise frequency is where t max and t min are the relative times when each deformed wave reaches the peak and trough;

[0070] In the input state, the noise frequency is:

[0071] Perform frequency - domain transformation on the real - time data after power - off restart. If the ratio of the difference between the real - time noise frequency and the background noise frequency to the background noise is higher than 8 - 12%, preferably 10%, that is:

[0072]

[0073] Then it is considered that the sensor is in the free state, and the reference automatic following algorithm is executed. Otherwise, after filling the zero - point data of the register before power - off, the reference automatic following is performed again;

[0074] The wavelet analysis method includes the following steps:

[0075] The set obtained by scaling and translating the mother wavelet and the scaling function has both high frequencies and low frequencies, and also covers the time domain, as follows:

[0076] f(t) = ∑ m ∑ j a j,k ψ j,k (t), where ψ j,k (t) is the wavelet series;

[0077] By localizing the signal in both the time domain and the frequency domain through the wavelet series, the outputs after denoising in the background state and the load state are respectively as follows:

[0078] f(t, ε = 0) = ∑ k ∑ j a j,k ψ j,k (t, ε = 0);

[0079] f(t, ε) = ∑ k ∑ j a j,k ψ j,k (t, ε);

[0080] That is, the wavelet series of the two are respectively ψ j,k (t, ε = 0) and ψ j,k (t, ε);

[0081] The noise amplitude method can be completed within an extremely short time period (within milliseconds), but it is vulnerable to environmental noise;

[0082] The noise frequency method can be completed within a relatively short time period (in the millisecond range), but it is also vulnerable to environmental noise;

[0083] The harmonic method can be completed within a slightly longer time period (from milliseconds to seconds) and is hardly affected by environmental noise;

[0084] The wavelet analysis method requires a relatively long time period (in the second range) to complete, but it is not affected by environmental noise.

[0085] In summary, after the device is powered off and restarted, the time-domain and frequency-domain characteristics of the noise signal are extracted (or combined), and the current state (background, load) of the sensor is identified through the characterization method, and then the zero-point tracking of the reference following algorithm is assisted, which can very accurately solve the problem of excessive zero-point correction caused by edge cases. For the application scenario of realizing innovative interaction through stress sensors, the solution involved in this patent can solve the defects existing in the reference (zero-point) following algorithm and improve the reliability and consistency of pressure-sensitive interaction.

[0086] Although the present invention has been described above with reference to the embodiments, various modifications thereof can be made and components thereof can be replaced with equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the features of the embodiments disclosed in the present invention can be combined with each other in any way, and the exhaustive description of these combinations is not given in this specification only for the sake of saving space and resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for characterizing a strain sensor, characterized in that: This method is based on extracting and separating the background noise and input noise of the object under test. The analysis methods include time-domain analysis methods and frequency-domain analysis methods; The time-domain analysis methods include amplitude analysis methods and harmonic analysis methods; The frequency-domain analysis methods include spectrum analysis methods and wavelet analysis methods; The amplitude analysis method is used to distinguish the equipment state through the peak-to-peak difference of the noise amplitude; The harmonic analysis method is used to distinguish the equipment state through the difference between the fundamental wave and harmonics expanded by the Fourier series; The spectrum analysis method is used to distinguish the equipment state through the noise frequency difference; The wavelet analysis method is used to distinguish the equipment state through the wavelet series difference.

2. The method for characterizing a strain sensor according to claim 1, wherein: The amplitude analysis method includes the following steps: S1: When powered on for the first time, automatically obtain the peak-to-peak value Vpp0 of the background noise and store it in the register; S2: Real-time obtain the peak-to-peak value Vpp of the total noise; S3: In case of power-off and restart, calculate the amplitude difference ΔV = ∣Vpp - Vpp0∣ at the moment after restart; S4: If ΔV > Vpp0×10%, it is considered that the equipment was in the input state before restart, and the stored initial value is used as the zero-point basis; S5: Otherwise, execute the reference automatic following algorithm.

3. A method for characterizing a strain sensor according to claim 2, characterized in that: The harmonic analysis method includes the following steps: Measure the periodic law of the structural vibration and expand it with the Fourier series. The expression is: Among them, when n = 1, it represents the fundamental wave, and when n > 1, it represents the harmonic wave. A n is the peak-to-peak value of the amplitude of the nth wave; In the background input state, only the fundamental wave exists, that is: The fundamental harmonic under load conditions is expressed as: Perform Fourier expansion on the real-time data after power-off and restart, and the fundamental wave and the first harmonic can be obtained, so as to distinguish whether the state of the equipment is the background state or the load state, and decide whether to execute the following algorithm for the zero point at the moment of power-on again.

4. A method for characterizing a strain sensor according to claim 3, wherein: The spectrum analysis method includes the following steps: In the background input state, the noise frequency is wherein, t max , t min are the relative times when each deformation wave reaches the wave crest and the wave trough; In the input state, the noise frequency is: Perform frequency-domain transformation on the real-time data after power-off and restart. If the ratio of the difference between the real-time noise frequency and the background noise frequency to the background noise is higher than 8 - 12%, that is: It is considered that the sensor is in the free state and execute the reference automatic following algorithm. Otherwise, after filling the zero-point data of the register before power-off, perform the reference automatic following again.

5. A method for characterizing a strain sensor according to claim 4, characterized in that: The wavelet analysis method includes the following steps: The set of the mother wavelet and the scaling function after scaling and translation has both high frequency and low frequency, and also covers the time domain, as follows: f(t) = ∑ m ∑ j a j，k ψ j，k (t), where ψ j，k (t) is a wavelet series; Locate the signal in both the time domain and the frequency domain through the wavelet series. The outputs after denoising in the background state and the load state are as follows: f(t, ε = 0) = ∑ k ∑ j a j，k ψ j，k (t, ε = 0); f(t, ε) = ∑ k ∑ j a j，k ψ j，k (t, ε); That is, the wavelet series of the two are ψ j，k (t, ε = 0) and ψ j，k (t, ε), respectively.

Images