Full range measurement method and system for linear variable differential transformer displacement sensors

Abstract

This invention discloses a method and system for full-range displacement measurement using a linear variable differential transformer (LVDT) displacement sensor. The method demodulates the differential output signal of the LVDT secondary coil, simultaneously extracting amplitude and phase information. Based on pre-calibrated zero-point residual vector parameters, two displacement estimates are calculated in parallel using amplitude-domain and phase-domain solution models, respectively. A fusion weight is generated based on the current amplitude information, and the two displacement estimates are adaptively weighted and fused to obtain the final displacement measurement value. This invention utilizes the complementary characteristics of high phase sensitivity near the zero point and high amplitude linearity far from the zero point, achieving continuous and smooth switching between the two observation quantities through fusion weights. This eliminates the zero-point measurement dead zone without altering the sensor's structure, realizing continuous high-precision displacement measurement across the entire range.

Description

Technical Field

[0001] This invention belongs to the field of precision displacement measurement technology, specifically relating to a method and system for full-range measurement of displacement using a linear variable differential transformer sensor. Background Technology

[0002] A linear variable differential transformer (LVDT) is a typical AC inductive displacement sensor that converts mechanical displacement into an electrical signal using the principle of electromagnetic induction. With its significant advantages such as simple structure, no mechanical contacts, high theoretical resolution, good repeatability, and strong environmental adaptability, it has been widely used in aerospace, industrial automation, and precision metrology. Traditional LVDT signal conditioning mainly relies on analog circuits or digital signal processing techniques. This involves phase-sensitive detection or synchronous demodulation of the differential signal from the secondary coil to extract a DC signal proportional to the displacement amplitude, thereby achieving displacement measurement.

[0003] However, in practical applications of LVDTs, due to non-ideal factors such as manufacturing errors in the coil windings, deviations in core processing and assembly, magnetic circuit asymmetry, and parasitic parameters of the conductors, the sensor inevitably has an unavoidable residual induced voltage near the electrical zero point (i.e., the theoretical output zero point). This residual component manifests as a fixed orthogonal bias in the signal vector space, causing the LVDT's output amplitude characteristic curve to change from an ideal V-shape to a smooth-bottomed hyperbola near the zero point. Under this physical constraint, when the core is located in a small region near the zero point, the sensitivity of the output signal amplitude to displacement changes (i.e., the slope of the amplitude-displacement curve) drops sharply and approaches zero. This physical degradation of sensitivity makes it difficult for small displacement changes in this region to be reflected in amplitude changes, thus forming a substantial measurement dead zone or low-sensitivity region in the center of the effective range, severely limiting the sensor's measurement accuracy in small ranges near the zero point.

[0004] To address the impact of residual voltage at zero points, existing techniques typically employ mechanical zeroing, the addition of external analog compensation networks, or nonlinear correction of the amplitude curve using polynomial fitting and neural networks in the digital domain. However, a common limitation of these methods is that they essentially still use amplitude as the sole or primary observable for displacement calculation. When the physical sensitivity of the amplitude observable at zero points has become ineffective, relying solely on mathematical fitting or compensation algorithms, while improving linearity, inevitably amplifies system noise, making it difficult to fundamentally improve the signal-to-noise ratio and resolution within the dead zone. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the aforementioned background technology and provide a full-range measurement method and system for a linear variable differential transformer displacement sensor. This overcomes the physical limitations of single-amplitude observations and, without altering the sensor's structure, fully utilizes the high phase sensitivity near zero to achieve high-precision measurement without dead zones across the entire range.

[0006] The technical solution adopted in this invention is: a method for full-range measurement of displacement using a linear variable differential transformer sensor, comprising the following steps: Obtain the differential output signal of the secondary coil of the linear variable differential transformer; The differential output signal is demodulated to extract amplitude information reflecting the signal strength and phase information reflecting the time delay of the signal relative to the excitation signal. A predetermined zero-point residual vector is provided, including the zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient, to characterize the inherent residual induction characteristics and displacement conversion relationship of the linear variable differential transformer at the electrical zero point; Based on the amplitude information, the zero-point residual amplitude, and the sensitivity coefficient, a first displacement estimate is calculated using a first solution model. The first solution model eliminates the influence of the zero-point residual amplitude on the measurement signal, extracts the effective signal component that is linearly related to the displacement, and inversely calculates the displacement. Based on the phase information and the zero-point residual vector, the second displacement estimate is calculated using the second solution model; the second solution model uses the phase reference established by the zero-point residual vector in the vector plane composed of the in-phase component and the orthogonal component to analyze the mapping relationship between the angle of phase deviation from the reference and the small displacement. Generate fusion weights, which are determined based on the amplitude information, and are used to characterize the credibility of the first displacement estimate in the current state; Based on the fusion weight, the first displacement estimate and the second displacement estimate are weighted and calculated to obtain the final displacement measurement value; In the weighted calculation, the fusion weights allocate a greater weight to the second displacement estimate than to the first displacement estimate in the electrical zero-point region, and a greater weight to the first displacement estimate than to the second displacement estimate in the linear region. Furthermore, the fusion weights change continuously between regions, resulting in a smooth transition in the final displacement measurement. The fusion weights ensure that the measurement result is primarily determined by the second displacement estimate in the electrical zero-point region, and primarily determined by the first displacement estimate in regions far from the electrical zero-point region, achieving a continuous and smooth transition between the two.

[0007] In the above technical solution, the demodulation step of the differential output signal includes: Generate a sine reference sequence and a cosine reference sequence with the same frequency and phase-locked as the excitation signal of the linear variable differential transformer; The digital sampling sequence of the differential output signal is multiplied with the sine reference sequence and the cosine reference sequence respectively, and then low-pass filtered to separate the in-phase component and the quadrature component. Based on the in-phase and quadrature components, the amplitude information is calculated by square root operation and the phase information is calculated by arctangent operation.

[0008] In the above technical solution, the calculation logic of the first solution model is as follows: Calculate the difference between the square of the amplitude information and the square of the zero-point residual amplitude, and take the square root of the difference to obtain the effective signal amplitude; The direction of the core displacement relative to zero is determined by using the sign of the in-phase component, and the direction determination result is obtained; Divide the effective signal amplitude by the sensitivity coefficient and combine it with the direction discrimination result to output the first displacement estimate.

[0009] In the above technical solution, the calculation logic of the second solution model is as follows: The phase reference is determined by the zero-point residual phase offset based on the position of the zero-point residual vector in the vector plane formed by the in-phase component and the quadrature component. Calculate the phase difference between the phase information and the phase reference; Based on the phase difference, combined with the zero-point residual amplitude and the sensitivity coefficient, the second displacement estimate is calculated using the inverse trigonometric function relationship.

[0010] In the above technical solution, the step of generating fusion weights is implemented through deterministic function mapping: Set a switching threshold related to the zero-point residual amplitude; A mapping relationship from the amplitude information to the weighting coefficients is constructed using an S-shaped smoothing function. The output of the S-shaped smoothing function changes continuously and monotonically with the amplitude, and the weighting coefficients are normalized values. When the amplitude information is lower than the switching threshold, the weight of the weight coefficient assigned to the second displacement estimate is greater than the weight of the first displacement estimate; when the amplitude information is higher than the switching threshold, the weight of the weight coefficient assigned to the first displacement estimate is greater than the weight of the second displacement estimate. The switching threshold and the steepness of the S-shaped function are set to ensure that the weight switching occurs entirely within the zero-neighborhood where the second solution model has high sensitivity, in order to avoid errors caused by using the second solution model in regions far from the electrical zero point.

[0011] In the above technical solution, the step of generating fusion weights is implemented through a neural network model: A multilayer perceptron network is constructed, whose input feature vector includes the amplitude information, the phase information, the ratio of the in-phase component to the quadrature component, and the difference between the amplitude and phase in adjacent sampling periods; During the calibration phase, the error between the final displacement measurement value and the known reference displacement during the calibration process is used as the loss function to train the multilayer perceptron network to learn the optimal weight allocation under different amplitude and phase states. Using the network parameters that have been fixed after training, normalized weight coefficients are output based on the feature vectors input in real time; the weight coefficients are used to adaptively identify whether the current measurement point is in the amplitude sensitivity attenuation region based on the multidimensional signal characteristics.

[0012] In the above technical solution, the step of providing a predetermined zero-point residual vector includes: The core of the linear variable differential transformer is controlled to move within the full range, and the differential output signal is demodulated to obtain the in-phase component and the quadrature component. The signal amplitude is calculated based on the in-phase and quadrature components. Several sampling points are selected in the neighborhood of the minimum amplitude value to perform local polynomial fitting. The position corresponding to the extreme point of the fitted curve is calculated as the electrical zero. Record the in-phase component value and quadrature component value corresponding to the electrical zero point, and calculate the zero point residual amplitude and the zero point residual phase offset accordingly. In the linear region far from the electrical zero point, the sensitivity coefficient is calibrated using the correspondence between amplitude and known displacement; The zero-point residual amplitude, the zero-point residual phase offset, and the sensitivity coefficient are stored as zero-point residual vector parameters.

[0013] This invention provides a linear variable differential transformer displacement sensor measurement system, comprising: The excitation signal generation module is configured to generate an excitation signal to drive the primary coil of a linear variable differential transformer; The signal acquisition module is configured to synchronously acquire the differential output signal of the secondary coil of the linear variable differential transformer and convert it into a digital signal; The quadrature demodulation module is configured to perform vector decomposition on the differential output signal using an orthogonal reference signal that originates from the same source as the excitation signal, and to obtain in-phase components, quadrature components, amplitude information and phase information; The storage unit is configured to store pre-calibrated zero-point residual vector parameters, which include zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient. The processor is configured to execute the full-range measurement method steps of the linear variable differential transformer displacement sensor as described in the above technical solution; the processor internally runs amplitude calculation logic and phase calculation logic in parallel, as well as a fusion control algorithm for arbitrating the weights of the two, thereby eliminating the measurement dead zone without changing the mechanical structure of the sensor.

[0014] In the above technical solution, the full-range measurement system of the linear variable differential transformer displacement sensor adopts a synchronous architecture with the same source clock: The excitation signal generation module uses a digital frequency synthesizer to generate an excitation signal and a quadrature demodulation reference signal, both of which share the same phase accumulator to ensure strict phase coherence. The sampling clock of the signal acquisition module and the update clock of the digital frequency synthesizer are triggered by the same reference timer to ensure timing synchronization between the excitation output, signal sampling and demodulation reference; The orthogonal demodulation module adopts a double-buffered pipeline mechanism, which divides the data buffer into two blocks. While the processor is directly accessing and transmitting one block of data, it is simultaneously performing reference sequence calculation and sampling data processing on the other block, thus achieving parallel transmission and computation.

[0015] The present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the steps of the full-range measurement method of the linear variable differential transformer displacement sensor as described in the above technical solution.

[0016] The beneficial effects of this invention are as follows: This invention proposes a full-range LVDT measurement method and system based on amplitude-phase fusion. Its core advantage lies in completely solving the measurement dead zone problem of traditional LVDTs near the electrical zero point from the perspective of measurement mechanism. Existing technologies rely solely on amplitude observation near the zero point, which is limited by the presence of residual voltage, leading to physical degradation of sensitivity. In contrast, this invention utilizes the high sensitivity of phase near the zero point, using phase information as a complementary observation means for amplitude information. By constructing an adaptive fusion mechanism, the system can automatically switch to phase-dominant mode in the zero-point region and to amplitude-dominant mode in regions far from the zero point. This design eliminates the need for complex mechanical zeroing of the sensor or the addition of extra analog compensation circuitry, enabling continuous, high-resolution, and high signal-to-noise ratio measurements from minute displacements to a large range without altering the sensor's structure, significantly enhancing the application potential of LVDTs in the field of precision positioning.

[0017] Furthermore, in terms of low-level signal processing, this invention employs digital synchronous demodulation technology based on orthogonal reference sequences. Compared to traditional analog phase-sensitive detection, this method can accurately separate the in-phase component related to displacement and the quadrature component related to residual voltage. This fully digital demodulation method not only avoids errors introduced by analog circuit temperature drift and DC bias, but also effectively suppresses broadband noise through low-pass filtering, providing a high-quality data foundation for subsequent accurate calculation of pure amplitude and phase information, ensuring that stable phase characteristics can still be extracted under zero-point residual voltage interference.

[0018] Furthermore, this invention designs two complementary physical displacement calculation models. The first calculation model (amplitude model) recovers the linear relationship between the effective signal and displacement by mathematically subtracting the residual amplitude at the zero point, effectively solving the nonlinear bending problem caused by residual voltage in the large range of traditional methods, and ensuring the linearity of macroscopic measurements. The second calculation model (phase model) cleverly utilizes the phase reference constructed by the residual vector, transforming the small displacement changes near the zero point into drastic angle changes for calculation, thereby achieving extremely high measurement resolution in the dead zone where amplitude sensitivity fails. The combination of the two models ensures that the system has no performance bottlenecks throughout the entire measurement range.

[0019] Furthermore, to achieve seamless integration of the two measurement modes, this invention provides two weight generation strategies: a deterministic function (S-shaped function) and a data-driven approach (multilayer perceptron). This mechanism eliminates the numerical jumps or discontinuities that may occur at the switching points in traditional piecewise linearization methods. In particular, the fusion scheme based on multilayer perceptrons (MLPs) can comprehensively utilize multidimensional features such as amplitude, phase, and rate of change to intelligently identify the current operating state of the sensor. This not only achieves a smooth transition but also adaptively compensates for nonlinear characteristics caused by environmental changes or individual sensor differences, further improving the system's robustness and environmental adaptability.

[0020] Furthermore, the parameter calibration method proposed in this invention greatly simplifies the installation and commissioning process of LVDT. Traditional methods often require extremely high-precision mechanical alignment to reduce residual zero-point voltage, while this method allows for a certain degree of zero-point residual voltage in the sensor. The residual zero-point vector parameters (amplitude and phase offset) can be accurately extracted through automatic scanning and fitting algorithms. This electronic zeroing method, which replaces mechanical zeroing, reduces the requirements for mechanical installation accuracy, reduces manual commissioning costs, and also improves the system's tolerance to mechanical zero-point drift after long-term operation.

[0021] Furthermore, in terms of hardware and system implementation, this invention employs a synchronous architecture with the same clock source. The generation of the excitation signal and the acquisition of the feedback signal are triggered by the same clock source and timer, ensuring strict phase locking between excitation and sampling and eliminating the impact of clock jitter on high-sensitivity phase measurement. Combined with a double-buffered pipelined DMA transfer mechanism, the system can achieve continuous data throughput and real-time computation even with limited microcontroller resources, ensuring high dynamic response capability for displacement measurement and making it suitable for real-time closed-loop control applications in industrial settings. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 A schematic diagram comparing the amplitude and phase characteristics near the zero point of an LVDT. Figure 3 The measured amplitude and phase variations with displacement are shown in the example. Figure 4 For the example, the signal amplitude at the zero point within the dead zone was measured. Figure 5 The measured signal phase at the zero point within the dead zone is shown in the example. Figure 6 This is a test case demonstrating the measurement error within the dead zone. Figure 7 This is a system architecture diagram of the present invention. Detailed Implementation

[0023] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments to facilitate a clear understanding of the present invention, but these descriptions do not constitute a limitation on the present invention.

[0024] like Figure 1 As shown, this invention provides a full-range measurement method for a linear variable differential transformer displacement sensor, comprising the following steps: Obtain the differential output signal of the secondary coil of the linear variable differential transformer; The differential output signal is demodulated to extract amplitude information reflecting the signal strength and phase information reflecting the time delay of the signal relative to the excitation signal. A predetermined zero-point residual vector is provided, including the zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient, to characterize the inherent residual induction characteristics and displacement conversion relationship of the linear variable differential transformer at the electrical zero point; Based on the amplitude information, the zero-point residual amplitude, and the sensitivity coefficient, a first displacement estimate is calculated using a first solution model. The first solution model eliminates the influence of the zero-point residual amplitude on the measurement signal, extracts the effective signal component that is linearly related to the displacement, and inversely calculates the displacement. Based on the phase information and the zero-point residual vector, the second displacement estimate is calculated using the second solution model; the second solution model uses the phase reference established by the zero-point residual vector in the vector plane composed of the in-phase component and the orthogonal component to analyze the mapping relationship between the angle of phase deviation from the reference and the small displacement. A fusion weight is generated, which is determined based on the amplitude information and is used to characterize the credibility of the first displacement estimate in the current state; Based on the fusion weight, the first displacement estimate and the second displacement estimate are weighted and calculated to obtain the final displacement measurement value; The fusion weights ensure that the measurement results are primarily determined by the second displacement estimate in the electrical zero-point region, and primarily determined by the first displacement estimate in the region far from the electrical zero-point region, achieving a continuous and smooth transition between the two.

[0025] A linear variable differential transformer (LVDT) is a typical AC inductive displacement sensor. Its core structure consists of a primary excitation coil and two symmetrically arranged secondary coils. By changing the mutual inductance between the primary and secondary coils through the movement of the iron core, the LVDT converts mechanical displacement into an electrical signal.

[0026] This invention is based on the principle of amplitude complementarity through orthogonal decomposition, first establishing a vector model of the residual voltage at the zero point of the LVDT. Due to factors such as manufacturing asymmetry, parasitic capacitance, and iron loss, actual LVDTs exhibit an uneliminable residual voltage at the zero displacement point. This residual voltage mainly manifests as an interference component orthogonal to the effective signal. Let the ideal effective output signal of the LVDT be... It is related to the displacement of the iron core. Proportional, that is ,in Let be the sensitivity coefficient. Let the residual voltage at the zero point be... It typically consists of higher harmonics of quadrature and in-phase components, and can be considered a constant over a short time. Actual output voltage vector. amplitude It can be represented as:

[0027] For displacement Taking the derivative, we obtain the amplitude sensitivity. :

[0028] when hour, This mathematically proves that near the zero point, the amplitude is extremely insensitive to changes in displacement, forming a measurement dead zone.

[0029] However, consider the phase of the output signal relative to the excitation signal. The following relationship is satisfied near zero:

[0030] phase Taking the derivative, we obtain the amplitude sensitivity. :

[0031] when When moving from the negative direction across zero to the positive direction, the phase It will happen A violent upheaval. In Within a tiny region, the rate of phase change is extremely high, meaning it has extremely high sensitivity.

[0032] like Figure 2 As shown, the changes in amplitude and phase characteristics indicate that the LVDT secondary output signal possesses two distinct physical characteristics: first, the amplitude characteristic, where the secondary output amplitude A is approximately equal to the square root of the sum of the square of the effective signal and the square of the residual voltage. Near zero, the derivative of the amplitude with respect to displacement, i.e., the amplitude sensitivity, approaches zero. At this point, even small noise can lead to significant displacement calculation errors, thus creating a measurement dead zone. Second, the phase characteristic, where the phase difference between the output signal and the excitation signal undergoes a sharp reversal near zero. Within this small interval, the derivative of the phase with respect to displacement is extremely large, resulting in a very high signal-to-noise ratio. This invention utilizes these differences in physical characteristics to replace the phase sensitivity for measurement in the interval where amplitude sensitivity fails.

[0033] The technical solution of the present invention will be further described below in conjunction with the above technical principles and embodiments.

[0034] Example 1 This embodiment provides a full-range measurement method for a linear variable differential transformer displacement sensor. This method utilizes the complementary characteristics of high phase sensitivity and low amplitude sensitivity within a small zero-point region, and achieves high-precision, dead-zone-free measurement across the entire range without altering the sensor's structure through an adaptive weighted fusion model. Specifically, it includes the following steps: S1, Signal Acquisition and Demodulation.

[0035] First, the differential output signal of the secondary coil of the linear variable differential transformer (LVDT) is obtained. This differential output signal is formed by differential processing of the induced voltage generated by the two secondary coils of the LVDT under the action of core displacement, and it contains amplitude and phase information related to the core displacement.

[0036] The differential output signal is demodulated to extract amplitude information reflecting the signal strength and phase information reflecting the time delay of the signal relative to the excitation signal. Specifically, the demodulation steps include: (1) Generate a sine reference sequence and a cosine reference sequence with the same frequency and phase-locked as the excitation signal of the linear variable differential transformer. Direct digital synthesis (DDS) technology is used inside the microcontroller, with a 48-bit frequency control word (FTW) and a phase accumulator to achieve discrete phase recursion:

[0037] in This represents the value of the 48-bit phase register. Assume the DAC update frequency is... Then the DDS output frequency The relationship with FTW is as follows:

[0038] By configuring FTW, the excitation frequency can be adjusted. Precise programmable settings and scanning.

[0039] To generate the excitation sinusoidal sequence and the orthogonal reference sequence, the system uses table lookup or equivalent numerical methods to obtain them:

[0040] in This is the initial phase. Used for driving signal generation , Simultaneously, it serves as an orthogonal demodulation reference, achieving strict coherence between the excitation signal and the demodulated signal.

[0041] (2) The digital sampling sequence of the differential output signal is multiplied by the sine reference sequence and the cosine reference sequence respectively, and then low-pass filtered to separate the in-phase component and the quadrature component. For the sampled signal x[n], the system performs synchronous quadrature demodulation: ,

[0042] in This indicates a digital low-pass filter, used to remove high-frequency components generated by multiplication and extract baseband components.

[0043] (3) Based on the in-phase and quadrature components, the amplitude information is calculated by square root operation and the phase information is calculated by arctangent operation: ; ; The square root and arctangent operations can be implemented using lookup tables, CORDIC algorithms, or approximation algorithms to meet the real-time requirements of the microcontroller. Through the above orthogonal demodulation process, amplitude and phase data reflecting the core displacement state are obtained, providing a data foundation for subsequent displacement calculations. The advantages of this technique are: generating excitation and quadrature demodulation reference signals using a co-source DDS ensures the stability of the phase reference; digital low-pass filtering effectively removes high-frequency interference and improves signal quality; and simultaneous acquisition of amplitude and phase provides complementary observations for subsequent fusion calculations.

[0044] S2, calibration of the zero-point residual vector.

[0045] A predetermined zero-point residual vector is provided, including the zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient, to characterize the inherent residual induction characteristics and displacement conversion relationship of the linear variable differential transformer at the electrical zero point.

[0046] The calibration steps for the zero-point residual vector include: (1) Control the core of the linear variable differential transformer to move within the full range and demodulate the differential output signal to obtain the in-phase component and the quadrature component. The control mechanism drives the core to slowly scan within the full range of the LVDT, while simultaneously acquiring the in-phase component I, the quadrature component Q, and the corresponding displacement data.

[0047] (2) Calculate the signal amplitude based on the in-phase and quadrature components: .

[0048] Several sampling points are selected in the neighborhood of the minimum amplitude value to perform local polynomial fitting, and the positions corresponding to the extreme points of the fitted curve are calculated as electrical zeros. Due to sampling discreteness and noise effects, directly selecting the minimum voltage value as the zero point will introduce system bias. In this embodiment, a quadratic polynomial is used to locally fit the relationship between amplitude and displacement, i.e. A(x) = a·x 2 + b·x+c, Calculate the displacement corresponding to the extreme point as the physical zero point. =-b / (2a). This method can find sub-pixel-level physical zero points and improve zero-point positioning accuracy.

[0049] (3) Record the in-phase component value I corresponding to the electrical zero point. r With orthogonal component value Q r Based on this, the zero-point residual amplitude is calculated: ; and the zero-point residual phase bias: ; All subsequent amplitude and phase calculations are based on Let the origin be the coordinate system, and let the coordinate system be the origin. As the residual vector parameter at the zero point.

[0050] (4) In the linear region far from the electrical zero point, the sensitivity coefficient is calibrated using the correspondence between amplitude and known displacement. Calibration data is selected for the linear region, based on: Solve for sensitivity coefficients using linear regression This calibration method avoids the influence of the nonlinear region near the zero point by calibrating in the linear region; by eliminating the influence of the residual amplitude at the zero point, it extracts the effective signal component proportional to the displacement, thus ensuring the accuracy of the sensitivity coefficient.

[0051] (5) Store the zero-point residual amplitude, the zero-point residual phase offset and the sensitivity coefficient as zero-point residual vector parameters for subsequent real-time measurement.

[0052] S3, amplitude domain displacement estimation.

[0053] Based on the amplitude information, the zero-point residual amplitude, and the sensitivity coefficient, a first displacement estimate is calculated using a first solution model. The first solution model eliminates the influence of the zero-point residual amplitude on the measurement signal, extracts the effective signal component that is linearly related to the displacement, and then calculates the displacement inversely.

[0054] The calculation logic of the first solution model is as follows: (1) The effective signal amplitude is obtained by taking the square root of the difference between the square of the amplitude information and the square of the zero-point residual amplitude: .

[0055] (2) The direction of the core displacement relative to zero is determined by the sign of the in-phase component, and the direction determination result is obtained: sign = sgn(I).

[0056] (3) Divide the effective signal amplitude by the sensitivity coefficient, and combine it with the direction discrimination result to output the first displacement estimate: .

[0057] The first solution model exhibits high linearity over a large range and high accuracy and stability in the linear region far from zero. By subtracting the square term of the residual amplitude at zero, the influence of residual voltage on the measurement is eliminated, enabling a linear output proportional to displacement to be obtained in the region far from zero. The displacement direction of the core is correctly identified by distinguishing the signs of the in-phase components.

[0058] S4, Phase Domain Shift Estimation.

[0059] Based on the phase information and the zero-point residual vector, a second displacement estimate is calculated using a second solution model. The second solution model utilizes a phase reference established by the zero-point residual vector in a vector plane composed of in-phase and quadrature components to analyze the mapping relationship between the angle of phase deviation from this reference and the minute displacement.

[0060] The calculation logic of the second solution model is as follows: (1) Based on the position of the zero-point residual vector in the vector plane formed by the in-phase component and the quadrature component, the zero-point residual phase offset φ is determined. r Determine the phase reference.

[0061] (2) Calculate the phase difference between the phase information and the phase reference: Δφ = φ - φ r .

[0062] (3) Based on the phase difference value, combined with the zero-point residual amplitude and the sensitivity coefficient, the second displacement estimate is calculated using the inverse trigonometric function relationship: .

[0063] This phase domain solution model has a resolution far exceeding that of the amplitude channel in the small displacement range near zero. When the core is near zero, the phase changes drastically, and the phase sensitivity is extremely high. At this time, the displacement can be solved by phase inversion with a much higher accuracy than the amplitude solution. By introducing the zero-point residual phase offset as a reference, an accurate mapping relationship between phase and displacement is established.

[0064] S5, Weighted Fusion Generation A fusion weight is generated, determined based on the magnitude information, to characterize the reliability of the first displacement estimate in the current state. This invention provides two methods for generating the fusion weight: one based on a deterministic function mapping and the other based on a neural network model.

[0065] In one implementation, the step of generating a fusion weight is achieved through a deterministic function mapping: (1) Set a switching threshold V related to the zero-point residual amplitude. th Based on the residual voltage V obtained from calibration r Set the fusion switching threshold, typically V th =β×V r , where β is a coefficient greater than 1, to ensure that the switching region is within the high sensitivity range where the phase model is effective.

[0066] (2) A mapping relationship from the amplitude information to the weighting coefficients is constructed using an S-shaped smoothing function. The output of the S-shaped smoothing function changes continuously and monotonically with the amplitude, and the weighting coefficients are normalized values. The fusion weights are calculated using the Sigmoid function, and this factor is based on the current measured amplitude. Dynamic calculation: ;

[0067] in, To switch the threshold voltage, The slope factor is used to adjust the steepness of the transition band. This invention employs a fast-in, fast-out strategy to ensure that the switching process occurs entirely within the effective boundary of the phase model fitting, and the switching is completed immediately upon recovery of the amplitude signal-to-noise ratio, thus avoiding extrapolation errors of the phase model at the far end.

[0068] (3) When the amplitude information is lower than the switching threshold, the weighting coefficient tends to make the final result mainly depend on the second displacement estimate; when the amplitude information is higher than the switching threshold, the weighting coefficient tends to make the final result mainly depend on the first displacement estimate. Specifically: When A< <V th When the weights approach 0, the system fully accepts the phase calculation results; When A>>V th When the weights rapidly approach 1, the system fully accepts the amplitude calculation results; In A≈V th In the transition region, the weights change continuously to achieve a smooth switching.

[0069] (4) The switching threshold and the steepness of the S-shaped function are set to ensure that the weight switching occurs entirely within the zero-neighborhood of the second solution model with high sensitivity, so as to avoid errors caused by using the second solution model in regions far from the zero point. This invention adopts a fast-in, fast-out strategy to ensure that the switching process occurs entirely within the effective boundary of the phase model fitting, and the switching is completed immediately at the moment the amplitude signal-to-noise ratio is recovered, thus avoiding the extrapolation error of the phase model at a far end.

[0070] The technical advantages of this deterministic fusion method based on the Sigmoid function are: the fusion weights can be calculated without training data; the Sigmoid function ensures the continuity and smoothness of weight changes, avoiding abrupt changes during switching; by reasonably setting the switching threshold and slope factor, it can be ensured that the switching occurs within the effective range of the phase model, while quickly transitioning to the amplitude model.

[0071] As another implementation method, the step of generating fusion weights can also be implemented using a neural network model: (1) In order to make full use of the high sensitivity of phase near zero and the high linearity of amplitude far from zero, and to achieve a continuous and smooth switching between the two, a multilayer perceptron (MLP) network is constructed to adaptively allocate the weights of amplitude calculation channel and phase calculation channel according to the current amplitude and phase state.

[0072] The MLP employs a feedforward neural network structure, including an input layer, at least one hidden layer, and an output layer. The input layer receives the amplitude and phase feature vectors obtained from orthogonal demodulation. The hidden layer establishes a nonlinear mapping between the amplitude and phase features and the zero-point residual state. The output layer is a single-node structure used to output the fusion weights. ,in The output layer preferably uses a Sigmoid function or an equivalent compression function to ensure that the weight values ​​are within the effective range for weighted fusion. The input feature vector of the MLP consists of physical quantities that characterize the relative relationship between the zero-point residual vector and the effective displacement vector, including the current magnitude. Phase The ratio of in-phase to quadrature components, I / Q, and the first-order difference between amplitude and phase. and .

[0073] Among them, amplitude Characterizing the overall magnitude and phase of the current signal relative to the zero-point residual. Characterizing the direction of the current vector in the I-Q plane, I / Q provides sign information about zero crossings and residual directions, while and These characteristics reflect the instantaneous sensitivity of amplitude and phase to displacement changes, respectively. Together, they characterize whether the system is currently in the amplitude-dominated or phase-dominated physical state.

[0074] The output of the MLP is the fusion weight. Its physical meaning is the reliability of the amplitude calculation channel under the current measurement conditions.

[0075] (2) During the calibration stage, the error between the final displacement measurement value and the known reference displacement during the calibration process is used as the loss function to train the multilayer perceptron network to learn the optimal weight allocation under different amplitude and phase states.

[0076] During the calibration process, the actual displacement of the iron core is known. Synchronous acquisition of amplitude under the condition Phase And the corresponding I and Q data, forming a sample set. The amplitude calculation results are performed for each sample separately. Phase solution results and output in a fusion manner With actual displacement The mean squared error between the two phases is used as the loss function, and the network weights are optimized through backpropagation, enabling the MLP to automatically learn how to allocate weights under different amplitude and phase states. To minimize the overall displacement error.

[0077] (3) Using the network parameters that have been fixed after training, normalized weight coefficients are output based on the real-time input feature vectors. After training, the network weights are fixed and stored in the microcontroller or external memory. During system operation, only forward inference is performed to calculate the weights. No online training is required, thus avoiding the introduction of additional real-time computational burden. The weighting coefficients are used to adaptively identify whether the current measurement point is in the amplitude sensitivity attenuation region based on multidimensional signal characteristics.

[0078] The technical advantages of this MLP-based fusion weight generation method are as follows: it learns the optimal weight allocation through a data-driven approach, which can adapt to the differences in characteristics of different sensors; multi-dimensional feature input enables the MLP to more accurately identify the current measurement state and achieve smoother switching in the transition region; after the network parameters are solidified, only forward inference is performed, the computational complexity is controllable, and the real-time requirements are met.

[0079] S6, weighted fusion output. Based on the fusion weights, the first displacement estimate and the second displacement estimate are weighted and calculated to obtain the final displacement measurement value: ; The fusion weights ensure that the measurement results are primarily determined by the second displacement estimate in the electrical zero-point region, and primarily determined by the first displacement estimate in the region far from the electrical zero-point region, achieving a continuous and smooth transition between the two.

[0080] When LVDT is at the center of the dead zone, i.e., A≈V r hour, →0, the system fully accepts the phase calculation results and uses the high sensitivity of the phase to fill the dead zone; when the LVDT leaves the dead zone, i.e., A>>V r hour, →1. The system fully accepts the amplitude calculation results and utilizes the large range linearity of the amplitude. At the transition edge, a weighted average is used to achieve smooth splicing, thereby eliminating the measurement dead zone caused by the zero-point residual voltage in mathematical principle.

[0081] This weighted fusion method solves the nonlinear error and noise amplification problems caused by the residual voltage at the zero point of the LVDT without changing the sensor hardware structure or requiring precise mechanical zeroing. It achieves full-range, high signal-to-noise ratio submicron-level continuous high-precision measurement and significantly reduces the system's dependence on sensor manufacturing processes.

[0082] like Figure 3 As shown, the measured amplitude (blue curve) and phase (yellow curve) vary with displacement. It can be observed that in the region far from zero, the amplitude and displacement exhibit a good linear relationship; however, near zero, the amplitude change tends to level off, forming a measurement dead zone. Simultaneously, the phase undergoes a sharp flip near zero, exhibiting extremely high sensitivity. This verifies the technical principle of this invention, which utilizes the complementary amplitude characteristics for fusion measurement.

[0083] like Figure 4 and Figure 5 As shown, the signal amplitude and phase at the zero point within the dead zone are illustrated. It can be seen that within a small displacement range near the zero point, the amplitude remains almost constant, exhibiting extremely low sensitivity; while the phase changes significantly with displacement, demonstrating high resolution. This further proves the necessity and effectiveness of phase calculation in the zero-point region.

[0084] like Figure 6 As shown, the measurement error within the dead zone after applying the method of this invention is illustrated. Compared to the traditional method that only uses amplitude calculation, this invention significantly reduces the measurement error in the zero-point region through amplitude-phase fusion, achieving continuous high-precision measurement across the entire measurement range.

[0085] Example 2 This embodiment provides a linear variable differential transformer displacement sensor measurement system for implementing the method described in Embodiment 1 above. Figure 7 As shown, the system includes an LVDT sensor, a coil drive unit, an analog signal conditioning unit, a digital-to-analog converter, an analog-to-digital converter, a direct memory access module, and a microcontroller. Each unit is connected in sequence according to the signal flow direction and works together.

[0086] The LVDT sensor is a linear variable differential transformer, whose core structure includes a primary excitation coil and two symmetrically arranged secondary coils. By moving the iron core, the mutual inductance between the primary and secondary coils is changed, and the LVDT converts mechanical displacement into an electrical signal. The primary coil receives the AC excitation signal from the coil drive unit, and the outputs of the two secondary coils form a pair of analog differential signals related to the iron core displacement. These differential signals are then input to the analog signal conditioning unit for further processing.

[0087] The coil drive unit is configured to receive the analog excitation signal from the digital-to-analog converter unit, amplify it, and then drive the primary coil of the LVDT. For example... Figure 7 As shown, the coil driving unit includes a differential driving circuit and a low-pass filter circuit.

[0088] The differential drive circuit converts single-ended signals into differential signals to improve common-mode interference immunity and enhance drive capability. The low-pass filter circuit filters out high-frequency quantization noise and harmonic components generated during digital-to-analog conversion, making the output excitation signal smoother and reducing high-frequency interference to the LVDT sensor.

[0089] The coil drive unit improves the signal's anti-interference capability through differential drive; the low-pass filter effectively suppresses the step effect of the DAC output, improves the waveform quality of the excitation signal, and thus improves the measurement accuracy of the LVDT.

[0090] The analog signal conditioning unit is configured to condition the differential signal output from the LVDT secondary coil to make it suitable for subsequent analog-to-digital conversion. For example... Figure 7 As shown, the analog signal conditioning unit includes a differential amplifier circuit and an anti-aliasing filter circuit.

[0091] The differential amplifier circuit amplifies the weak induced signal from the LVDT to a voltage range suitable for analog-to-digital conversion, while suppressing common-mode interference and DC drift. The anti-aliasing filter circuit is a low-pass filter with a cutoff frequency set below half the ADC sampling frequency to filter out high-frequency components in the signal and prevent spectral aliasing during analog-to-digital conversion.

[0092] The differential amplification of the analog signal conditioning unit effectively improves the signal level and suppresses common-mode interference; the anti-aliasing filter eliminates the influence of high-frequency noise on sampling, ensuring the fidelity of the digitized signal and laying a good foundation for subsequent digital signal processing.

[0093] The digital-to-analog converter (DAC) is configured to convert the digital excitation signal generated by the microcontroller's internal digital signal processing unit into an analog signal. The DAC is connected to the microcontroller via a direct memory access (DMA) channel, enabling the digital excitation sequence to be continuously transmitted to the DAC at a fixed update rate, thereby ensuring the real-time and continuous output of the excitation signal.

[0094] The output value of the digital-to-analog converter unit is given by the following formula: Given, among which The midpoint code is 2048. For the excitation amplitude control coefficient, To excite the sinusoidal sequence.

[0095] The digital-to-analog converter reduces the CPU load through DMA transfer, ensuring continuous output of the excitation signal; the programmable amplitude control coefficient G enables the system to adapt to the excitation amplitude requirements of different LVDT specifications.

[0096] The analog-to-digital converter (ADC) is configured to synchronously sample the LVDT secondary signal processed by the analog signal conditioning unit and convert it into a digital signal. The ADC is connected to the microcontroller via a direct memory access (DMA) channel, enabling the sampled data to be continuously transmitted to the microcontroller's internal memory at a fixed sampling rate, thereby ensuring the real-time performance and consistency of data acquisition.

[0097] The sampling clock of the analog-to-digital converter (ADC) and the update clock of the digital-to-analog converter (DAC) are triggered by the same reference timer inside the microcontroller, ensuring timing synchronization between the excitation output and signal sampling. This synchronization mechanism guarantees that each sampling point is strictly aligned with the corresponding excitation phase, avoiding phase jitter caused by asynchronous sampling.

[0098] Synchronous sampling by the analog-to-digital converter ensures a stable phase relationship between the excitation and response signals, providing an accurate timing reference for quadrature demodulation; the DMA transfer mode enables continuous and seamless acquisition of sampled data, avoiding data loss.

[0099] The Direct Memory Access (DMA) module is configured to establish efficient data transfer channels between the digital-to-analog converter (DAC) and the microcontroller, as well as between the DAC and the microcontroller, enabling data transfer without CPU intervention. For example... Figure 7 As shown, the system is equipped with two DMA channels: one for transferring digital excitation sequences from the microcontroller to the digital-to-analog converter (DAC), and the other for transferring sampled data from the DAC to the microcontroller's internal memory.

[0100] To ensure parallel data transmission and processing, the system employs a ping-pong update method combining DMA half-transfer and full-transfer interrupts. The specific implementation is as follows: For DMA transfer on the analog-to-digital converter side: the data buffer is divided into two blocks, the first half and the second half. While the DMA outputs the data in the first half of the buffer, the microcontroller calculates and fills the data to be output in the second half in real time. When the DMA outputs the data in the second half, the data in the first half is updated, thus realizing a pipelined output structure of half-block output and half-block update, ensuring the continuity of the output waveform and no phase jump.

[0101] For DMA transfer on the analog-to-digital conversion unit side: the half-transfer interrupt mechanism is also adopted, so that when half of the data is written by DMA, the microcontroller can process the other half of the data that has been acquired, thereby avoiding data overwriting and ensuring real-time demodulation under continuous sampling conditions.

[0102] The ping-pong buffering mechanism of the direct memory access module enables data transmission and computation to be performed in parallel, significantly improving the real-time performance of the system; it avoids frequent CPU intervention in data movement, freeing up processor resources for complex signal processing and displacement calculation; and it ensures that data is not lost and waveforms are uninterrupted under continuous sampling conditions.

[0103] The microcontroller is the core processing unit of the system, and it integrates a digital signal processing unit, such as... Figure 7 As shown, the digital signal processing unit includes a direct digital synthesis module, a quadrature demodulation module, and an amplitude-phase fusion module.

[0104] The Direct Digital Synthesis (DDS) module is configured to generate the AC excitation signal for the LVDT and the reference signal required for quadrature demodulation. The microcontroller incorporates a 48-bit frequency control word (FTW) and a phase accumulator to implement discrete phase recursion during DMA transfer half-complete interrupts.

[0105] in This represents the value of the 48-bit phase register. Assume the DAC update frequency is... Then the DDS output frequency The relationship with FTW is as follows:

[0106] By configuring FTW, the excitation frequency can be adjusted. Precise programmable settings and scanning.

[0107] To generate the excitation sinusoidal sequence and the orthogonal reference sequence, the system uses table lookup or equivalent numerical methods to obtain them: ; in This is the initial phase. Used for driving signal generation , Simultaneously, it serves as an orthogonal demodulation reference, achieving strict coherence between the excitation signal and the demodulated signal.

[0108] The excitation signal is output to the DAC via a timer-triggered circuit, and data is continuously transferred from the circular buffer to the DAC via direct DMA. To ensure the continuity of the output waveform and the absence of phase jumps, the system employs a ping-pong update method combining DMA half-transfer and full-transfer interrupts. While the DMA outputs the first half of the data in the buffer, the microcontroller calculates and fills the second half of the data to be output in real time. The first half of the data is updated again when the DMA outputs the second half, thus achieving a pipelined output structure of half-block output and half-block update. The DAC output value is generated by... Given. Meanwhile, synchronously generated based on the same phase accumulator. and After quantization, it is stored in the reference buffer as a quadrature demodulation reference for subsequent multiplication and demodulation of the sampled signal.

[0109] The Direct Digital Synthesis (DDS) module provides a precisely adjustable excitation signal based on DDS digital frequency synthesis technology; a 48-bit phase accumulator ensures extremely high frequency resolution; the excitation signal generated from the same source is strictly phase coherent with the demodulation reference signal, eliminating the impact of phase drift on demodulation accuracy.

[0110] The quadrature demodulation module is configured to perform vector decomposition on the LVDT secondary signal using an orthogonal reference signal originating from the same source as the excitation signal, obtaining in-phase components, quadrature components, amplitude information, and phase information. For the sampled signal x[n], the system performs synchronous quadrature demodulation: , ; in This represents a digital low-pass filter, used to remove high-frequency components generated by multiplication and extract the baseband component. The resulting in-phase component I and quadrature component Q are used to calculate the amplitude and phase of the output signal.

[0111] The square root and arctangent operations can be implemented using lookup tables, CORDIC, or approximation algorithms to meet the real-time requirements of microcontrollers.

[0112] The synchronous quadrature demodulation module can simultaneously acquire the amplitude and phase information of the signal, providing complete observations for subsequent fusion calculations; the digital low-pass filter effectively removes the high-frequency components generated by multiplication, improving demodulation accuracy; the square root and arctangent operations implemented by CORDIC or lookup table algorithms meet the real-time requirements of the microcontroller.

[0113] The amplitude-phase fusion module is configured to execute the displacement calculation and fusion algorithm described in Embodiment 1 above. Internally, the module runs amplitude calculation logic and phase calculation logic in parallel, as well as a fusion control algorithm for arbitrating their weights, thereby eliminating the measurement dead zone without altering the sensor's mechanical structure.

[0114] The amplitude-phase fusion module constructs two displacement estimation paths in parallel based on the physical model: The amplitude domain solution path calculates the first displacement estimate using the following formula: ; This path exhibits high accuracy and stability in linear regions far from zero.

[0115] The phase domain solution path calculates the second displacement estimate using the following formula: ; This path has a much higher resolution than the amplitude path in the small displacement range near zero.

[0116] To fully utilize the high sensitivity of phase near zero and the high linearity of amplitude far from zero, and to achieve a continuous and smooth switching between the two, the amplitude-phase fusion module introduces an adaptive gating mechanism. This gating mechanism can adopt a deterministic mapping method based on the Sigmoid function or a data-driven method based on a multilayer perceptron (MLP).

[0117] When an MLP is used as the gating module, the MLP employs a feedforward neural network structure, including an input layer, at least one hidden layer, and an output layer. The input layer receives the amplitude-phase feature vector obtained from orthogonal demodulation, including the current amplitude A, phase φ, the ratio I / Q of the in-phase and quadrature components, and the first-order differences ΔA and Δφ between the amplitude and phase. The hidden layer establishes a nonlinear mapping relationship between the amplitude-phase features and the zero-point residual state. The output layer is a single-node structure used to output the fusion weights. ,in ∈[0,1].

[0118] The physical meaning of the fusion weight is the reliability of the amplitude calculation path under the current measurement conditions. The system uses this weight to weight and fuse the displacement calculation results in the amplitude and phase domains. The final displacement is given by the following formula: .

[0119] When the core is near zero, the amplitude sensitivity decreases, and the phase changes drastically, the gating mechanism automatically outputs a smaller w, causing the system to primarily rely on the phase calculation path; when the core moves away from zero and enters the linear region, the gating mechanism automatically outputs a larger w. This makes the system mainly rely on the amplitude calculation path; in the transition region between the two, the weights change continuously, thus achieving smooth splicing without jumps or jitter.

[0120] The amplitude-phase fusion module transforms the amplitude dead zone of the LVDT zero-point region into a fusion zone that can be compensated by phase information and data-driven models through parallel dual-path solution and adaptive gating fusion. This enables continuous, highly sensitive, and robust displacement measurement across the entire range without altering the sensor's structure.

[0121] The displacement output unit is configured to output the final displacement measurement value calculated by the amplitude-phase fusion module to an external device or system. This displacement result can be output to a host computer or control system via a digital interface (such as SPI, I2C, UART, etc.), an analog interface (such as DAC output voltage signal), or a communication bus (such as CAN, RS485, etc.) to achieve high-precision displacement measurement and closed-loop control.

[0122] As a preferred embodiment, the system employs a synchronous architecture with the same clock source to ensure timing synchronization and phase coherence among the modules: (1) The direct digital synthesis module inside the microcontroller generates the excitation signal and the quadrature demodulation reference signal, both of which share the same phase accumulator to ensure strict phase coherence. The 48-bit phase accumulator ensures that the excitation signal and the demodulation reference signal originate from the same phase source, eliminating demodulation errors caused by phase drift.

[0123] (2) The sampling clock of the analog-to-digital converter and the update clock of the digital-to-analog converter are triggered by the same reference timer to ensure timing synchronization between the excitation output, signal sampling and demodulation reference. This synchronization mechanism ensures that the demodulation reference value corresponding to each sampling point is accurate and avoids phase jitter caused by asynchronous sampling.

[0124] (3) The direct memory access module adopts a double buffer pipeline mechanism, which divides the data buffer into two blocks. While the DMA transfers one block of data, the processor performs reference sequence calculation and sampling data processing on the other block, thus realizing parallel transmission and calculation.

[0125] Through the above system structure, this invention realizes a complete measurement link from excitation generation, signal acquisition, amplitude and phase extraction to fusion calculation without changing the structure of the LVDT sensor itself. It provides a system foundation for filling the measurement dead zone by utilizing the high sensitivity of the phase near zero point and achieving high-precision measurement across the entire range by utilizing the high linearity of the amplitude over a large range.

[0126] The overall technical advantages of this system include: a co-clock architecture that ensures strict synchronization of excitation, sampling, and demodulation, eliminating phase drift and timing jitter, and improving the stability and accuracy of amplitude and phase measurements; a double-buffered pipeline mechanism that enables parallel data transmission and computation, guaranteeing the real-time performance of the system under continuous sampling conditions and avoiding data loss and processing delays; digital signal processing that allows for flexible parameter adjustment to adapt to different LVDT sensor specifications, improving the system's versatility; an adaptive weighting of amplitude and phase calculation results based on a gated fusion mechanism, reducing the system's reliance on engineering experience; and the ability of the entire system to run on resource-constrained microcontroller platforms, with low cost and power consumption, making it suitable for embedded applications.

[0127] Example 3 This embodiment provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the steps of the full-range measurement method of the linear variable differential transformer displacement sensor described in Embodiment 1 above.

[0128] The computer-readable storage medium may include, but is not limited to: flash memory, electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), hard disk, optical disk, magnetic tape, or any other non-transitory medium capable of storing computer programs.

[0129] The computer program includes the implementation code for the following functional modules: (1) Signal acquisition and demodulation module: used to acquire the differential output signal of the LVDT secondary coil, generate a quadrature reference sequence, and perform synchronous quadrature demodulation to extract in-phase components, quadrature components, amplitude information and phase information; (2) Zero-point residual vector parameter reading module: used to read the pre-calibrated zero-point residual amplitude, zero-point residual phase offset and sensitivity coefficient from the storage unit; (3) First solution module: used to calculate the first displacement estimate based on amplitude information and zero-point residual vector parameters by eliminating the influence of zero-point residual amplitude; (4) Second solution module: used to calculate the second displacement estimate based on phase information and zero-point residual vector parameters through phase inverse solution relationship; (5) Fusion weight generation module: used to generate fusion weights based on amplitude information, which can be implemented using Sigmoid function mapping or multilayer perceptron network; (6) Weighted fusion output module: used to perform weighted calculation on the first displacement estimate and the second displacement estimate according to the fusion weight, and output the final displacement measurement value.

[0130] When the processor loads and executes the above-mentioned computer program, it can realize the full-range measurement method described in this invention, eliminate the measurement dead zone without changing the mechanical structure of the sensor, and realize continuous high-precision displacement measurement across the entire range.

[0131] In summary, this invention, by constructing a displacement calculation mechanism based on a physical amplitude-phase model and adaptive fusion, enables the system to automatically switch to a phase-dominant measurement mode near the zero point and to an amplitude-dominant measurement mode far from the zero point. This allows for the simultaneous acquisition of high zero-point sensitivity of the phase and high linear dynamic range of the amplitude within the same measurement link. Specific technical effects include: (1) Elimination of measurement dead zone: This invention utilizes the high phase sensitivity of the zero-point region and uses phase calculation to calculate displacement in the range where amplitude sensitivity degrades, effectively eliminating the measurement dead zone that exists near the zero point in the traditional LVDT measurement method.

[0132] (2) High-precision measurement across the entire range: By adaptively fusing amplitude and phase, this invention utilizes the high linearity of amplitude in the linear region far from zero and the high sensitivity of phase near zero to achieve continuous high-precision measurement across the entire range.

[0133] (3) No need to change the sensor structure: The present invention solves the zero dead zone problem from the signal processing level, without the need for mechanical zeroing or structural compensation of the LVDT sensor body, thus reducing the requirements for sensor manufacturing process.

[0134] (4) Improve system versatility: Through data-driven fusion, this invention reduces the dependence on coil matching accuracy and processing symmetry, and improves the versatility and long-term stability of LVDTs of different specifications.

[0135] (5) Good real-time performance: The same clock architecture and double buffer pipeline mechanism adopted in this invention enable the system to run in real time on a resource-constrained microcontroller platform, meeting the real-time requirements of engineering applications.

[0136] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for full-range measurement using a linear variable differential transformer displacement sensor, characterized in that, Includes the following steps: Obtain the differential output signal of the secondary coil of the linear variable differential transformer; The differential output signal is demodulated to extract amplitude information reflecting the signal strength and phase information reflecting the time delay of the signal relative to the excitation signal. A predetermined zero-point residual vector is provided, including the zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient, to characterize the inherent residual induction characteristics and displacement conversion relationship of the linear variable differential transformer at the electrical zero point; Based on the amplitude information, the zero-point residual amplitude, and the sensitivity coefficient, the first displacement estimate is calculated using the first solution model. The first solution model eliminates the influence of zero-point residual amplitude on the measured signal, extracts the effective signal component that is linearly related to the displacement, and inversely calculates the displacement. Based on the phase information and the zero-point residual vector, the second displacement estimate is calculated using the second solution model; The second solution model uses the phase reference established by the zero-point residual vector in the vector plane composed of the in-phase component and the orthogonal component to analyze the mapping relationship between the angle of phase deviation from the reference and the small displacement. Generate fusion weights, which are determined based on the amplitude information, and are used to characterize the credibility of the first displacement estimate in the current state; Based on the fusion weight, the first displacement estimate and the second displacement estimate are weighted and calculated to obtain the final displacement measurement value; In the weighted calculation, the weighting weight is assigned to the second displacement estimate in the electrical zero-point region as a greater weight than the first displacement estimate, and to the first displacement estimate in the linear region as a greater weight than the second displacement estimate. The weighting weight changes continuously between regions to ensure a smooth transition in the final displacement measurement value.

2. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The steps for demodulating the differential output signal include: Generate a sine reference sequence and a cosine reference sequence with the same frequency and phase-locked as the excitation signal of the linear variable differential transformer; The digital sampling sequence of the differential output signal is multiplied with the sine reference sequence and the cosine reference sequence respectively, and then low-pass filtered to separate the in-phase component and the quadrature component. Based on the in-phase and quadrature components, the amplitude information is calculated by square root operation and the phase information is calculated by arctangent operation.

3. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The calculation logic of the first solution model is as follows: Calculate the difference between the square of the amplitude information and the square of the zero-point residual amplitude, and take the square root of the difference to obtain the effective signal amplitude; The direction of the core displacement relative to zero is determined by using the sign of the in-phase component, and the direction determination result is obtained; Divide the effective signal amplitude by the sensitivity coefficient and combine it with the direction discrimination result to output the first displacement estimate.

4. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The calculation logic of the second solution model is as follows: The phase reference is determined by the zero-point residual phase offset based on the position of the zero-point residual vector in the vector plane formed by the in-phase component and the quadrature component. Calculate the phase difference between the phase information and the phase reference; Based on the phase difference, combined with the zero-point residual amplitude and the sensitivity coefficient, the second displacement estimate is calculated using the inverse trigonometric function relationship.

5. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The step of generating the fusion weights is implemented through deterministic function mapping: Set a switching threshold related to the zero-point residual amplitude; A mapping relationship from the amplitude information to the weighting coefficients is constructed using an S-shaped smoothing function. The output of the S-shaped smoothing function changes continuously and monotonically with the amplitude, and the weighting coefficients are normalized values. When the amplitude information is lower than the switching threshold, the weight of the weight coefficient assigned to the second displacement estimate is greater than the weight of the first displacement estimate; When the amplitude information is higher than the switching threshold, the weight of the weight coefficient assigned to the first displacement estimate is greater than the weight of the second displacement estimate; The switching threshold and the steepness of the sigmoid function are set so that the weight switching occurs entirely within the zero neighborhood of the second solution model.

6. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The step of generating fusion weights is implemented through a neural network model: A multilayer perceptron network is constructed, whose input feature vector includes the amplitude information, the phase information, the ratio of the in-phase component to the quadrature component, and the difference between the amplitude and phase in adjacent sampling periods; During the calibration phase, the error between the final displacement measurement value and the known reference displacement during the calibration process is used as the loss function to train the multilayer perceptron network to learn the optimal weight allocation under different amplitude and phase states. Using the network parameters that have been fixed after training, normalized weight coefficients are output based on the feature vectors input in real time; the weight coefficients are used to adaptively identify whether the current measurement point is in the amplitude sensitivity attenuation region based on the multidimensional signal characteristics.

7. The full-range measurement method of the linear variable differential transformer displacement sensor according to claim 1, characterized in that, The step of providing a predetermined zero-point residual vector includes: The core of the linear variable differential transformer is controlled to move within the full range, and the differential output signal is demodulated to obtain the in-phase component and the quadrature component. The signal amplitude is calculated based on the in-phase and quadrature components. Several sampling points are selected in the neighborhood of the minimum amplitude value to perform local polynomial fitting. The position corresponding to the extreme point of the fitted curve is calculated as the electrical zero. Record the in-phase component value and quadrature component value corresponding to the electrical zero point, and calculate the zero point residual amplitude and the zero point residual phase offset accordingly. In the linear region, the sensitivity coefficient is calibrated using the correspondence between the amplitude and the known displacement; The zero-point residual amplitude, the zero-point residual phase offset, and the sensitivity coefficient are stored as zero-point residual vector parameters.

8. A linear variable differential transformer displacement sensor measurement system, characterized in that, include: The excitation signal generation module is configured to generate an excitation signal to drive the primary coil of a linear variable differential transformer; The signal acquisition module is configured to synchronously acquire the differential output signal of the secondary coil of the linear variable differential transformer and convert it into a digital signal; The quadrature demodulation module is configured to perform vector decomposition on the differential output signal using an orthogonal reference signal that originates from the same source as the excitation signal, and to obtain in-phase components, quadrature components, amplitude information and phase information; The storage unit is configured to store pre-calibrated zero-point residual vector parameters, which include zero-point residual amplitude, zero-point residual phase offset, and sensitivity coefficient. The processor is configured to perform the steps of the full-range measurement method for a linear variable differential transformer displacement sensor as described in any one of claims 1 to 7.

9. The full-range measurement system for a linear variable differential transformer displacement sensor according to claim 8, characterized in that, The linear variable differential transformer displacement sensor full-range measurement system adopts a synchronous architecture with the same source clock. The excitation signal generation module uses a digital frequency synthesizer to generate an excitation signal and a quadrature demodulation reference signal, both of which share the same phase accumulator. The sampling clock of the signal acquisition module and the update clock of the digital frequency synthesizer are triggered by the same reference timer; The orthogonal demodulation module adopts a double-buffered pipeline mechanism, which divides the data buffer into two blocks. While one block of data is being transmitted via direct memory access, the processor performs reference sequence calculation and sampled data processing on the other block.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the full-range measurement method of the linear variable differential transformer displacement sensor as described in any one of claims 1 to 7.

Images