A low-voltage wire harness terminal insertion quality detection method

Abstract

The application specifically relates to a low-voltage wire harness terminal plug-in quality detection method, and belongs to the technical field of electric connection quality testing.The method comprises the following steps: injecting a broadband composite excitation signal into the wire harness and collecting a response, separating extremely low-frequency and medium-high-frequency components through Fourier transform; performing unbiased frequency domain filtering on the extremely low-frequency components to remove power frequency interference, extracting an impedance real part to obtain a temperature drift coefficient; stripping parasitic parameters through multi-frequency point impedance circle fitting on the medium-high-frequency components to obtain an equivalent series resistance, constructing a dispersion compensation factor based on the relationship between the skin effect and the square root of the frequency, combining the temperature drift coefficient to correct a nominal resistance, and obtaining a terminal contact resistance; and establishing a Gaussian confidence score model according to historical statistics and dynamically updating a standard deviation. The application realizes non-invasive internal temperature sensing through the electric parameters of the wire harness itself, so that the temperature compensation reference can truly reflect the actual temperature field in the closed sheath, and the compensation error caused by the arrangement deviation of the external sensor is eliminated.

Description

Technical Field

[0001] This invention relates to a method for testing the insertion quality of low-voltage wire harness terminals, belonging to the technical field of electrical connection quality testing. More specifically, this invention relates to a method for testing the insertion quality of low-voltage wire harness terminals. Background Technology

[0002] In the quality inspection of low-voltage wiring harness assemblies before shipment, the terminals are completely sealed by plastic sheaths and tape, making optical visual inspection impossible. Therefore, electrical measurement methods are commonly used to indirectly assess the connection status. The conventional approach is to measure the DC resistance of the circuit or the AC impedance at a specific frequency and compare it with a preset acceptable threshold. However, the ohmic resistance of the conductor in a long wiring harness undergoes a non-negligible drift with ambient temperature, while the additional resistance at the terminal contact interface caused by defects such as oxidation and poor connection is usually only in the milliohm range, easily masked by temperature drift. When high-frequency excitation is used to obtain more contact characteristic information, the skin effect of the conductor causes the AC resistance to increase non-linearly with frequency, and this effect is coupled with temperature drift in the total resistance, making it difficult to isolate the terminal contact resistance. Some solutions add external temperature sensors to correct for temperature effects, but these sensors cannot accurately reflect the actual temperature distribution of the conductor inside the sealed sheath, limiting compensation accuracy and increasing system cost.

[0003] Furthermore, factors such as power frequency electromagnetic interference, quantization noise, and unstable probe contact in the workshop can cause random fluctuations in measured values. Using a fixed hard threshold for quality judgment is prone to false alarms or missed detections because it cannot adapt to dynamic changes in random noise levels. Existing methods lack effective means to decouple temperature drift and skin effect in the frequency domain and accurately extract terminal contact resistance without relying on external temperature sensors, and have not established a dynamic statistical judgment model with anti-interference capabilities.

[0004] Therefore, how to utilize the electrical parameters of the wire harness itself to achieve intrinsic temperature sensing, decouple and compensate for temperature drift and skin effect, thereby accurately calculating terminal contact resistance and establishing a robust judgment mechanism to adapt to noise fluctuations is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for detecting the connection quality of low-voltage wire harness terminals, employing the following technical solution: A broadband composite voltage excitation signal is injected into the wire harness under test, and a current response signal is acquired; the broadband composite voltage excitation signal and the current response signal are subjected to frequency domain transformation to separate extremely low-frequency components and mid-to-high-frequency components; the extremely low-frequency components are subjected to unbiased frequency domain filtering to eliminate power frequency interference, and the real part of the extremely low-frequency impedance is extracted; the temperature drift coefficient is calculated based on the real part of the extremely low-frequency impedance and a pre-stored nominal resistance; the mid-to-high-frequency components are subjected to multi-frequency impedance circle fitting to remove parasitic capacitive reactance and inductive reactance. The system first obtains the initial equivalent series resistance; then constructs a dispersion compensation factor based on the nonlinear mapping relationship between the skin effect and the square root of frequency. The nominal resistance is then jointly corrected for temperature drift and skin effect using the temperature drift coefficient and dispersion compensation factor. Finally, the corrected nominal resistance is subtracted from the initial equivalent series resistance to obtain the terminal contact resistance. A Gaussian distribution confidence scoring model is established based on historical statistical priors, and the model standard deviation is updated dynamically online. The confidence score is calculated based on the deviation between the terminal contact resistance and the target resistance. When the confidence score is lower than a preset threshold, the connection quality is deemed poor.

[0006] Optionally, the frequency domain transformation employs the discrete Fourier transform.

[0007] Optionally, the frequency corresponding to the extremely low frequency component is selected based on the condition that the skin depth is greater than the conductor radius, and the frequency corresponding to the mid-to-high frequency component is selected based on the condition that the skin depth is less than the conductor radius.

[0008] Optionally, the unbiased frequency domain filtering is a windowed notch filter for power frequency, and the fixed attenuation factor is abolished to avoid systematic amplitude deviation.

[0009] Optionally, the temperature drift coefficient is obtained by dividing the difference between the real part of the very low frequency impedance and the nominal resistance by the nominal resistance.

[0010] Optionally, the construction of the dispersion compensation factor based on the nonlinear mapping relationship between the skin effect and the square root of the frequency includes: obtaining the center frequency ratio between the mid-high frequency and the extremely low frequency, and multiplying the power of the center frequency ratio by the basic nonlinear gain weight that has been pre-calibrated or statistically learned to obtain the dispersion compensation factor.

[0011] Optionally, the temperature drift coefficient and dispersion compensation factor jointly correct the nominal resistance for temperature drift and skin effect, and the corrected nominal resistance is subtracted from the initial equivalent series resistance to obtain the terminal contact resistance, including: multiplying the pre-stored nominal resistance by the dispersion compensation factor, and then multiplying it by a value plus the temperature drift coefficient to obtain the corrected nominal resistance; subtracting the corrected nominal resistance from the initial equivalent series resistance to obtain the terminal contact resistance.

[0012] Optionally, the online dynamic update of the model standard deviation includes: using a sliding window method or a Bayesian update strategy to update the standard deviation in real time using current and historical detection data; wherein the standard deviation is initialized using a small sample initialization method.

[0013] Optionally, the target resistance is a pre-calibrated standard contact resistance value.

[0014] The embodiments of the present invention have at least the following beneficial effects:

[0015] 1. Since the skin effect and distributed capacitance can be ignored under extremely low frequency excitation, the real part of the impedance is determined only by the ohmic resistance of the conductor. After removing power frequency interference by using unbiased frequency domain filtering, the real part of the extremely low frequency impedance is extracted, and the dimensionless temperature drift coefficient is calculated based on the nominal resistance. Thus, non-invasive internal temperature sensing is achieved using the electrical parameters of the wire harness itself, so that the temperature compensation reference can truly reflect the actual temperature field inside the enclosed sheath and eliminate the compensation error caused by the deviation of the external sensor arrangement.

[0016] 2. Due to the nonlinear mapping relationship between the skin effect and the square root of the frequency (exponential) A dispersion compensation factor is constructed and jointly corrected by the temperature drift coefficient at extremely low frequencies. Combined with multi-frequency impedance circle fitting, parasitic capacitive and inductive reactance are removed. This ensures that the remaining resistance after deducting the effects of temperature drift and skin effect from the initial equivalent series resistance only represents the terminal contact interface state. Thus, the contact resistance changes caused by micron-level oxidation or poor connection will not be masked by interference from the main circuit, and reliable contact characteristic indicators are obtained.

[0017] 3. Because a historical prior Gaussian distribution reliability scoring model is used instead of a fixed threshold, and the standard deviation is updated online dynamically. The terminal contact resistance deviation is mapped to a normalized confidence score through the ratio of the square term to the variance. The judgment boundary can be adaptively adjusted with the noise level, avoiding random spike triggering misjudgments caused by power frequency interference, quantization noise and fixture fluctuations, thus improving the robustness of quality judgment in a statistical sense. Attached Figure Description

[0018] Figure 1 The schematic diagram illustrates the steps of a method for detecting the connection quality of low-voltage wire harness terminals according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To avoid program calculation errors, numerical overflows, and illegal NaN values ​​caused by the denominator approaching 0 under extreme conditions such as hardware sampling failure, insufficient calibration samples, and signal loss, this invention introduces a global minimum protection constant ε= .

[0021] Quantity description: The value of ε is much smaller than the order of magnitude of resistance, current, frequency, and countermeasures normally measured by this system. Under normal testing conditions, the calculation error for the final contact resistance and confidence score is less than [value missing]. It does not affect the accuracy of detection and judgment at all.

[0022] Dimension matching rules: Current denominator is matched with dimension A, frequency denominator is matched with Hz, resistance denominator is matched with Ω, and dimensionless denominator for counting types is directly matched with dimensionless ε; Modify the general rule: Add +ε to the end of the overall expression of the denominator that has the risk of returning to zero, without changing the mathematical logic of normal operation.

[0023] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings. (Refer to...) Figure 1 A method for detecting the connection quality of low-voltage wire harness terminals includes steps S1-S7, as detailed below: S1: Inject a broadband composite voltage excitation signal into the wire harness under test and acquire the current response signal.

[0024] During the inspection of the connection quality of low-voltage wiring harness terminals, the contact interface of the terminal under inspection is completely covered by plastic sheath and tape, making it impossible to observe directly by optical means. It is necessary to use electrical excitation and response measurement to obtain information on the contact electrical characteristics of the terminal.

[0025] However, the DC resistance of the long wire harness body will drift with changes in ambient temperature. This temperature drift can easily mask the slight impedance changes at the terminal contact interface caused by minor oxidation or poor connection. At the same time, if a single high-frequency excitation is used, the skin effect of the wire will cause a nonlinear increase in resistance, which will be coupled with the temperature drift.

[0026] Therefore, this step injects a broadband composite voltage excitation signal into the circuit of the harness under test. This signal contains both extremely low-frequency and mid-to-high-frequency components, and simultaneously acquires the current response signal generated in the circuit, thereby obtaining an electrical parameter sequence that can simultaneously reflect the temperature effect and the contact interface impedance characteristics during a single excitation response process.

[0027] One feasible signal generation method is as follows: a broadband composite voltage excitation signal in the form of a superposition of multi-frequency sine waves is output using a digital signal synthesizer, and its time-domain expression is:

[0028] In the formula, This is the instantaneous value of the excitation voltage, measured in volts (V). In this embodiment, the number of superimposed frequency components is... ; For the first The amplitude of each frequency component, measured in volts (V), is used to suppress high-frequency energy while maintaining a sufficient signal-to-noise ratio for low-frequency components. Follow Increase and according to Decreasing Get experience points The unit is volts (V). This value ensures that the voltage amplitudes generated by each frequency component across the sampling resistor are on the same order of magnitude. It can also be adjusted by the implementer according to the length and cross-sectional area of ​​the wire harness under test. For the first The frequency of each frequency component is measured in Hertz (Hz), and the frequency range is from... Hz coverage kHz, and a logarithmically uniform distribution is used to select the frequency to ensure sufficient spectral resolution in both the very low and mid-high frequency regions; For the first The initial phase of each frequency component is taken as an empirical value of 0 radians (rad).

[0029] The overall peak voltage of the excitation signal does not exceed 0.1V, which is a non-destructive low-level excitation, avoiding the generation of current at the contact interface that could cause local overheating or dielectric breakdown.

[0030] In the testing device, the broadband composite voltage excitation signal generated by the digital signal synthesizer is amplified by the voltage drive amplifier and then applied to the beginning and end of the circuit under test of the wire harness through the shielded clamp probe.

[0031] A precision sampling resistor is connected in series in the circuit. Its nominal resistance is 0.1. With an accuracy of ±0.1%, it is used to linearly convert loop current into a voltage signal.

[0032] The voltage difference across the sampling resistor is acquired using a differential amplifier. Then the loop current It can be obtained from the following formula:

[0033] In the formula, The unit is volt (V). The unit is ohm ( ), The unit is ampere (A).

[0034] Data acquisition card with sampling frequency MHz for excitation voltage and current Perform synchronous, equally spaced sampling; the time length of a single acquisition window. s, to obtain the discrete-time voltage sequence and current sequence ,in , One sampling point.

[0035] Sampling frequency The Nyquist sampling condition of the signal must be met, i.e. This is to avoid spectral aliasing of high-frequency components during the digitization process.

[0036] During data acquisition, to reduce interference from power frequency electromagnetic fields, spatial radiation noise, and contact fluctuations of the fixture probes, the entire wire harness under test and the test fixture are placed inside a metal shielding box. The supporting hardware quantitative specifications are as follows: Shielding box: Cold-rolled steel plate wall thickness ≥1.2mm, protective grounding impedance ≤0.1Ω; Signal transmission cable: 50Ω double-shielded coaxial cable, with the total cable length controlled within 1m; Grounding specifications: The shielding layer should only be connected to the protective ground of the test equipment at one end; multiple grounding points at both ends of the harness to form a ground loop are prohibited. Front-end differential amplifier: Input equivalent noise ≤10nV / The operating bandwidth fully covers 0.5Hz~200kHz; Sampling resistor selection: Low-temperature drift metal film resistor with a temperature drift coefficient ≤5ppm / ℃ to eliminate additional measurement errors introduced by temperature during the sampling process.

[0037] This step completes the injection of the broadband composite voltage excitation signal and the acquisition of the current response signal, resulting in a synchronously aligned discrete voltage sequence. and current discrete sequence .

[0038] S2: Perform frequency domain transformation on the broadband composite voltage excitation signal and current response signal to separate the extremely low frequency component and the mid-to-high frequency component.

[0039] The time-domain voltage sequence obtained in step S1 and time-domain current sequence In the middle, the extremely low frequency signal components and the mid-to-high frequency signal components overlap on the time axis and cannot be directly distinguished by amplitude or waveform characteristics.

[0040] Because long conductors exhibit significantly different physical effects at different frequency bands—that is, at extremely low frequencies, the current is uniformly distributed across the conductor cross-section, and the impedance is mainly determined by the bulk ohmic resistance and drifts with temperature changes; in the mid-to-high frequency bands, the skin effect gradually increases, the current concentrates on the conductor surface, and the impedance is affected by both the skin effect and parasitic reactance—it is necessary to transform the time-domain signal to the frequency domain and separate the two types of components based on their frequency.

[0041] right and Perform separately Point-based discrete Fourier transform yields a complex spectrum sequence. and :

[0042]

[0043] In the formula, For time-domain sampling point index, , dimensionless integer; For frequency domain indexing, , dimensionless integer; The first frequency domain voltage sequence A complex value, in volts (V). The first frequency domain current sequence A complex value, in amperes (A). It is the imaginary unit.

[0044] Frequency domain index corresponding physical frequency It is given by the following formula:

[0045] In the formula, Hz is the sampling frequency. Each point represents the number of sampling points per window, and the frequency resolution. Hz.

[0046] Because the discrete Fourier transform of a real signal has conjugate symmetry, it typically only retains... to The spectral components correspond to the frequency range DC to This allows for a complete description of the signal characteristics.

[0047] In this embodiment, the Discrete Fourier Transform is directly implemented using the Fast Fourier Transform algorithm. The Fast Fourier Transform technique is a well-known technique and will not be described in detail here.

[0048] To define the boundary between extremely low-frequency components and mid-to-high-frequency components in the frequency domain, the skin depth concept from electromagnetic field theory is utilized. With frequency Relationship:

[0049] In the formula, Let be the resistivity of the conductor, for copper conductors. ; Let be the permeability of the conductor, and for non-ferromagnetic materials, take . ; The signal frequency is expressed in Hertz (Hz). For frequency The skin depth is measured in meters (m).

[0050] Skin depth characterizes the degree to which alternating current concentrates on the surface of a conductor cross section: when the skin depth is much greater than the conductor radius, the current is approximately uniformly distributed and the skin effect is negligible; when the skin depth is less than the conductor radius, the skin effect is significant.

[0051] Therefore, the skin depth is equal to the conductor radius. The corresponding frequency is defined as the separation frequency. ,satisfy Solving for:

[0052] In the formula, The nominal radius of the conductor in the wire harness to be inspected is expressed in meters (m). The unit is Hertz (Hz).

[0053] For a copper wire with a cross-sectional area of ​​0.75 mm² in a typical automotive low-voltage wiring harness, the conductor radius is approximately... Substituting m into the above equation, we get Hz.

[0054] During implementation, The specifications are determined by the conductor specifications and material parameters of the actual wire harness to be inspected.

[0055] In the frequency domain, for each index Calculate its corresponding frequency .

[0056] Will satisfy All indexes The set formed is denoted as The spectral components within this set are the extremely low frequency components; they will satisfy... All indexes The set formed is denoted as The spectral components within this set are the mid-to-high frequency components.

[0057] Through the above separation, the extremely low frequency voltage frequency domain component is obtained from the frequency domain transformation result. and extremely low frequency current frequency domain components and the frequency domain components of mid-to-high frequency voltage. and mid-to-high frequency current frequency domain components .

[0058] In the frequency domain, the voltage and current relationship of the frequency band corresponding to the extremely low frequency component is mainly determined by the ohmic resistance of the conductor body and is modulated by temperature changes; the voltage and current relationship of the frequency band corresponding to the mid-to-high frequency component is simultaneously affected by the skin effect, parasitic reactance and terminal contact interface impedance.

[0059] S3: Perform unbiased frequency domain filtering on the extremely low frequency components to eliminate power frequency interference, extract the real part of the extremely low frequency impedance, and calculate the temperature drift coefficient based on the real part of the impedance and the pre-stored nominal resistance.

[0060] Because of the presence of power frequency electromagnetic fields in the workshop environment, the fundamental frequency (50Hz or 60Hz) is located within the extremely low frequency component band. Therefore, the extremely low frequency voltage frequency domain component separated in step S2... and current frequency domain components In addition, power frequency interference is superimposed.

[0061] If these frequency domain components are used directly to calculate impedance, power frequency interference will cause the estimated real part of the impedance to deviate, thus failing to accurately reflect the resistance change of the wire harness conductor caused by temperature changes.

[0062] Therefore, this step first performs unbiased frequency domain filtering on the extremely low frequency components to eliminate power frequency interference components.

[0063] Specifically, determine the power frequency fundamental frequency. Its value is either 50Hz or 60Hz, depending on the power supply standard at the test site.

[0064] In the frequency domain, physical frequency , where frequency resolution Hz.

[0065] Find and closest frequency index To satisfy Minimum.

[0066] Power frequency interference energy is not absolutely concentrated at a single frequency point, but is distributed across... Within the adjacent narrow band, therefore, the following approach is adopted. Centered on, half width is Windowing and notch filtering for each index.

[0067] set up The empirical value of 3 is taken as a dimensionless integer, which corresponds to a physical frequency band half-width of about 6Hz. It can also be adjusted by the implementer according to the actual power frequency fluctuation amplitude.

[0068] For extremely low frequency sets Each index in Calculate its notch weights .

[0069] when At that time, the weights decay smoothly according to the Hanning window function, and the calculation formula is as follows:

[0070] when At that time, weight .

[0071] This processing only gradually attenuates the frequency components near the power frequency, instead of using... The fixed attenuation method, which directly sets the value to zero, eliminates the fixed attenuation factor and avoids systematic amplitude estimation errors caused by spectral leakage and frequency domain abrupt changes.

[0072] Using the above weights, unbiased frequency domain filtering is performed on the extremely low frequency voltage and current frequency domain components to obtain the filtered components:

[0073]

[0074] In the formula, and These are the frequency domain components of the extremely low frequency voltage and current after filtering, respectively, in volts (V) and amperes (A).

[0075] At extremely low frequencies, since the skin depth is much larger than the conductor radius, the current is uniformly distributed across the conductor cross-section, and the skin effect can be ignored. At the same time, the distributed capacitance of the insulation layer exhibits extremely high capacitive reactance at this frequency, and the distributed inductive reactance of the conductor is also extremely low. Therefore, the reactance component of the entire circuit is extremely small relative to the resistive component.

[0076] In other words, the imaginary part of the extremely low frequency impedance is approximately zero, while the real part is basically determined by the DC ohmic resistance of the wire harness conductor, and this DC ohmic resistance is a single-valued function of temperature.

[0077] Based on this physical property, the real part of the ultra-low frequency impedance is extracted from the filtered ultra-low frequency domain components.

[0078] In specific calculations, to reduce the impact of random quantization noise or residual interference at individual frequency points on impedance estimation, a set of... The impedance at all frequency points is taken as an arithmetic average.

[0079] First, for each Calculate complex impedance :

[0080] Then, take the real part and average it to obtain the real part of the extremely low frequency impedance. :

[0081] In the formula, For set The total number of frequency points in the range, a dimensionless integer; This indicates taking the real part of a complex number; This is the real part of the extremely low frequency impedance, in ohms. ).

[0082] Pre-heated at standard room temperature (e.g., at 25℃) Under these conditions, the DC resistance of wire harnesses of the same type and circuit structure is measured using the precision four-wire method to obtain the nominal resistance. And its value is stored in the detection system.

[0083] Should It reflects the reference resistance of the conductor when there is no temperature drift and no skin effect.

[0084] because and The differences mainly stem from deviations in ambient temperature. The resulting change in conductor resistivity can be used to calculate the temperature drift coefficient. :

[0085] In the formula, The unit is ohm ( ); It is a dimensionless coefficient.

[0086] When the temperature of the wiring harness is higher than the standard room temperature, the conductor resistance increases. , It is a positive value; when the temperature is lower than the standard room temperature, It is a negative value.

[0087] Thus, the temperature drift coefficient of the current wiring harness circuit under test at ambient temperature is obtained. .

[0088] S4: For mid-to-high frequency components, parasitic capacitive and inductive reactances are removed by multi-frequency impedance circle fitting to obtain the initial equivalent series resistance.

[0089] Obtain the initial equivalent series resistance.

[0090] The mid-to-high frequency voltage frequency domain components obtained in step S2 and mid-to-high frequency current frequency domain components In this case, the loop harness impedance is no longer determined solely by the DC ohmic resistance of the conductor.

[0091] As the frequency increases, the skin effect causes the current to concentrate on the surface of the conductor, reducing the effective conductive cross-sectional area. This results in a nonlinear increase in AC resistance relative to DC resistance. At the same time, the distributed capacitance between conductors and between conductors and shielding layers, as well as the distributed inductance of the conductors themselves, begin to exhibit non-negligible capacitive and inductive reactance in this frequency band.

[0092] The aforementioned physical effects together cause the complex impedance measured in the mid-to-high frequency band to deviate from the simple actual resistance value. Both the real and imaginary parts are composed of multiple electrical parameters such as the skin resistance, parasitic capacitive reactance, parasitic inductive reactance of the circuit body, and the contact interface resistance of the terminal to be tested.

[0093] If the real part of the complex impedance at any single frequency point within the frequency band is used directly as an estimate of the terminal contact interface resistance, it will introduce serious coupling errors caused by parasitic reactance, thus missing or misjudging minute contact defects.

[0094] Therefore, it is necessary to perform multi-frequency impedance circle fitting on the mid-to-high frequency components to reconstruct the impedance trajectory caused by parasitic capacitive and inductive reactance on the complex impedance plane, and then extract the initial equivalent series resistance representing resistive loss from it.

[0095] For constituting a set Each frequency domain index Calculate the corresponding complex impedance :

[0096] In the formula, and These are the frequency domain components of the mid-to-high frequency voltage and current obtained in step S2, respectively, in volts (V) and amperes (A). The unit is ohm ( ).

[0097] This complex impedance can be decomposed into its real part. and the virtual part All in ohms ( The unit is ) and These represent the operations of extracting the real part and the imaginary part, respectively.

[0098] Under broadband excitation conditions, the equivalent circuit of the wire harness loop in the mid-to-high frequency range can be approximated by a simplified series RLC model, that is, the impedance of the circuit under test can be approximately expressed as: ,in For equivalent series resistance, For the parasitic series inductance of the circuit, For the parasitic parallel capacitance of the circuit, ω is the angular frequency.

[0099] When the frequency varies within a certain range, the parasitic capacitance... and parasitic resistance This will cause the imaginary part of the complex impedance to exhibit frequency dependence, while the real part will always be equal to... .

[0100] Therefore, under ideal conditions where high-frequency spurious effects and measurement noise are ignored, different frequencies Corresponding complex impedance data points Plotted on the real-imaginary plane, they will converge to the plane with respect to the real-imaginary part. On an impedance circle trajectory centered at a point.

[0101] In actual measurements, due to quantization noise, probe contact fluctuations, and minor non-ideal effects from high-frequency coupling of the wiring harness, the data points will be approximately distributed within a certain range. The center and radius of the fitted circle are determined by the equivalent reactance parameters.

[0102] Based on the above physical model, this step adopts the multi-frequency point least squares circle fitting method.

[0103] Take the mid-to-high frequency set All data points corresponding to the indices are used as fitting input, and the number of corresponding frequency points is denoted as . , a dimensionless integer.

[0104] Let the coordinates of the center of the fitted circle be... , radius is All units are ohms ( ).

[0105] The algebraic form of the equation of a circle is:

[0106] To find the center of the circle Expanding and linearizing the above equation, we can construct the objective function. This is the sum of the squares of the differences between the squares of the distances from all data points to the center of the circle and the squares of the radius:

[0107] In the formula, intermediate variables , and The relationship between the center and radius is as follows: , , Their units are ohms ( ),ohm( ) and Ohm square ( ).

[0108] Through the Seeking information about , and By taking the partial derivative of the given equation and setting it to zero, we obtain a set of linear equations. Solving this system of linear equations yields the solution. , and The optimal estimate.

[0109] Subsequently, the coordinates of the center are:

[0110] radius Obtained from the following formula:

[0111] Since the center of the impedance circle is theoretically located on the real axis, that is... The real part coordinates should approach zero, therefore the fitted real part coordinates are... As the initial equivalent series resistance :

[0112] In the formula, The unit is ohm ( ).

[0113] Should Characterized the stripping of parasitic capacitance under broadband mid-to-high frequency excitation. and parasitic resistance After the imaginary part frequency variation effect is caused, the actual series resistive loss component in the circuit is reduced.

[0114] In terms of physical composition, the equivalent series resistance includes the DC ohmic resistance of the wire harness body, the resistance increment caused by the skin effect, and the series contribution of the contact resistance at the terminal contact interface.

[0115] To prevent impedance circle fitting from getting stuck in a local minimum due to uneven frequency distribution or deep fading interference at individual frequency points, regularization constraints can be applied to the fitting process during implementation.

[0116] Taking the minimization of the absolute value of Xc as the regularization objective, a regularization penalty term λXc² is added to the original least squares objective function, resulting in the complete regularized objective function:

[0117] Where λ is the regularization strength, which is dimensionless, and the baseline empirical value is λ = 0.01; In another embodiment, to facilitate quick parameter tuning by those skilled in the art, a comparison example of λ values ​​under different wire harness diameter specifications is provided: Fine conductor (0.35mm², high high-frequency impedance scattering): λ=0.03; Standard automotive wiring harness (0.75mm²): λ=0.01; Thick conductor (2.0 mm², impedance scattering): λ = 0.001.

[0118] Explanation of the impact of parameter values: If the value of λ is too large, it will excessively suppress the degree of freedom of the imaginary part of the fitting, causing the equivalent series resistance Resr to shift positively; if λ approaches 0, the center convergence constraint will be lost, and the fitting result is prone to non-physical solutions in which the imaginary part deviates significantly from the real axis of the impedance.

[0119] After substituting the regularization term, the coefficient terms corresponding to parameter b in the linear equation system are increased by the regularization correction, which forces the fitting circle center to converge toward the real axis of impedance, thereby improving the physical rationality and numerical stability of the Resr estimate.

[0120] This step completes the multi-frequency impedance circle fitting process for mid-to-high frequency components, successfully eliminating the frequency-varying interference of parasitic capacitive and inductive reactance on the imaginary part of the impedance, and obtaining an initial equivalent series resistance independent of frequency. .

[0121] S5: Construct a dispersion compensation factor based on the nonlinear mapping relationship between the skin effect and the square root of the frequency.

[0122] The nominal resistance is jointly corrected for temperature drift and skin effect using the temperature drift coefficient and dispersion compensation factor, and the corrected nominal resistance is subtracted from the initial equivalent series resistance to obtain the terminal contact resistance.

[0123] The initial equivalent series resistance is obtained in step S4 by multi-frequency impedance circle fitting. It characterizes the total series resistive loss of the wire harness circuit. Physically, it includes the equivalent resistance of the wire harness conductor body in the mid-to-high frequency range and at the current temperature, as well as the contact resistance of the terminal contact interface.

[0124] Among them, the resistivity component of the conductor body of the wire harness is affected by both temperature drift and skin effect, and changes with frequency, making it impossible to determine from a single frequency. The contact resistance is obtained by directly subtracting a fixed value from the input.

[0125] Temperature drift causes the DC resistance of the conductor to deviate from the preset nominal resistance. This deviation is the temperature drift coefficient obtained in step S3. Quantitative description: The skin effect causes the AC resistance of a conductor in the mid-to-high frequency range to increase nonlinearly relative to its own DC resistance, and the degree of increase is related to the frequency.

[0126] Step S4 has eliminated the imaginary part frequency-varying effect caused by parasitic capacitive and inductive reactance through impedance circle fitting. Only the resistive component remains; therefore, this step requires relying on the physical laws of the skin effect and the obtained temperature drift coefficient, from... The resistive portion of the wire harness body is stripped out to obtain the contact resistance that simply reflects the terminal contact interface.

[0127] The classic theory of the skin effect states that for a cylindrical conductor, the AC resistance is approximately proportional to the square root of the frequency.

[0128] Under broadband excitation conditions, the frequency corresponding to the extremely low frequency component satisfies that the skin depth is much larger than the conductor radius, the current is uniformly distributed on the cross section, and the skin effect can be ignored. At this time, the AC resistance of the conductor is equal to the DC resistance. However, the skin depth corresponding to the frequency of the mid-to-high frequency component is smaller than the conductor radius, the current concentrates on the surface, and the AC resistance increases significantly.

[0129] To quantify this frequency-dependent resistive amplification effect, the center frequencies of the extremely low frequency band and the mid-to-high frequency band separated in step S2 are first determined.

[0130] For extremely low frequency sets For all frequency points, calculate the arithmetic mean of their frequencies, and use it as the center frequency of the extremely low frequency. For mid-to-high frequency sets Calculate the arithmetic mean of the frequencies of all points in the range, and use this as the center frequency for the mid-to-high frequency range. ,Right now

[0131]

[0132] In the formula, and The unit for all of them is Hertz (Hz).

[0133] Since the broadband composite excitation signal contains multiple spectral lines in both the extremely low and mid-high frequency ranges, the arithmetic mean frequency reflects the central position of the excitation energy in each frequency band, thus representing the typical frequency level of that band.

[0134] Based on the proportional relationship of the square root of the frequency, a dispersion compensation factor is constructed. :

[0135] In the formula, It is a dimensionless dispersion compensation factor, which characterizes the amplification factor of the conductor's AC resistance relative to its DC resistance when the excitation frequency changes from the extremely low frequency center to the mid-to-high frequency center.

[0136] index The skin nonlinear mapping exponent is taken as the exponent, based on the physical law that the AC resistance is proportional to the square root of the frequency due to the skin effect. , dimensionless.

[0137] It is a basic nonlinear gain weight used to compensate for deviations from the ideal square root law caused by non-ideal factors such as conductor stranding structure and non-uniform current distribution near terminals. It is dimensionless; the default benchmark value is empirical.

[0138] =1.0, the corresponding standardized and defect-free sample calibration process is as follows, including steps S50-S52, as detailed below: S50: Select three standard wire bundles of the same specification, free from oxidation and loose connection defects, and collect multiple sets of broadband excitation spectrum data using the complete detection process of this invention; S51: Fix the ambient temperature of the test environment at 25℃ to eliminate temperature drift interference. At this time, the contact resistance of the standard sample terminal is approximately 0. S52: Obtained by transforming the formula for calculating the body resistance. Solution formula:

[0139] S53: Take the arithmetic mean of the calculation results of 3 standard samples as the specific calibration for this wire harness specification. .

[0140] In one embodiment, this application provides a measured calibration example of multi-specification copper wire harnesses: (1) 0.35mm² multi-strand stranded wire: =1.08; (2) 0.75mm² automotive standard wiring harness: =1.00; (3) 2.0mm² thick single-strand wire: =0.94.

[0141] In addition to offline calibration, this parameter can also be iteratively optimized and adjusted using statistical learning methods based on qualified test data accumulated over a long period of time on the production line.

[0142] At the current ambient temperature, the DC ohmic resistance of the wire harness conductor has changed due to temperature drift. .

[0143] When the excitation frequency increases from extremely low to medium-high frequencies, the skin effect causes the DC resistance to amplify. Therefore, the equivalent series resistance of the wire harness conductor body in the mid-to-high frequency range and at the current temperature is times that of the wire harness conductor body. It is given by the following formula:

[0144] In the formula, The unit is ohm ( ).

[0145] Initial equivalent series resistance The physical composition is ,in The terminal contact resistance to be determined is expressed in ohms. ).

[0146] Therefore, minus That is, the contact resistance is obtained:

[0147] In the formula, The unit is ohm ( ).

[0148] Should The resistance components caused by temperature drift and skin effect of the wire harness conductor body have been deducted, and only the actual contact resistance of the terminal plug interface is reflected due to the metal contact area, oxide film and micro-roughness.

[0149] This step completes the joint compensation for temperature drift and skin effect, yielding the terminal contact resistance. .

[0150] S6: Establish a Gaussian distribution reliability scoring model based on historical statistical priors.

[0151] The model standard deviation is updated online and dynamically, and the confidence score is calculated based on the deviation between the terminal contact resistance and the target resistance.

[0152] Obtain the terminal contact resistance in step S5. Then, theoretically, this value should approach the pre-calibrated ideal target resistance.

[0153] However, even after extremely low-frequency unbiased filtering, the power frequency electromagnetic interference in the testing workshop may still retain a weak spectral leakage component. Sampling quantization noise, microscopic slippage at the probe-terminal contact surface, and the thermal noise of the wiring harness itself can all cause this. The random fluctuations cause a certain degree of dispersion in the results of repeated measurements of the same terminal circuit.

[0154] If a fixed hard threshold is used in this step, then... Compared to the target resistance, the deviation in a single measurement caused by dispersion is very likely to cause false alarms or missed detections.

[0155] Therefore, for the currently detected It is necessary to use historical connection quality data to build a statistical judgment model, convert the resistance deviation into a confidence index that can resist random disturbances in a statistical sense, and dynamically correct the standard deviation in the model that represents the level of random error online.

[0156] Pre-stored target resistance This is the standard contact resistance value obtained through offline calibration.

[0157] During calibration, a group of terminal wire harnesses of the same model, whose contact condition has been confirmed to be excellent through metallographic sectioning, X-ray fluoroscopy, or multiple insertion and removal verifications, are selected. Measurements are repeated under an excitation and calculation process completely consistent with production line testing, and the values ​​of each circuit are recorded. The arithmetic mean of the results is used as And store it in the parameter library of the detection system.

[0158] The unit is ohm ( ), which represents the expected resistance level of the contact interface of this type of terminal in a near-ideal plugging state.

[0159] First, calculate the absolute deviation between the terminal contact resistance and the target resistance:

[0160] In the formula, The terminal contact resistance output in step S5 is in ohms. ); The target resistance is pre-stored, in ohms ( ); The absolute deviation is expressed in ohms. ).

[0161] A confidence scoring model is established based on the probability density function form of the Gaussian distribution.

[0162] Standard deviation (Unit is ohms) This characterizes the effect caused by random interference under normal plugging conditions. The degree of dispersion.

[0163] Under the Gaussian distribution assumption, when exist When the dimensions are small, the terminals are more likely to be in a normal state; when Significantly exceeding If so, there is a high probability of contact defects.

[0164] Based on this, a normalized confidence score is defined. :

[0165] In the formula, The absolute deviation of the current terminal under test, in ohms ( ); This is the current effective standard deviation, in ohms ( ). ); The confidence score is dimensionless and its value range is [range missing]. . This represents an exponential function with base e.

[0166] when Exactly equal to ,Right now hour, , indicating the highest confidence level; as Increase The exponential decay and convergence towards 0 reflects an increased likelihood that the deviation exceeds the normal range of random fluctuations.

[0167] Model standard deviation It is not fixed once set, but rather the strategy is dynamically updated online, adjusting in real time based on continuously accumulating historical testing data, so that... It can adaptively track the random error level under the current production line environment.

[0168] When the testing system is put into use for the first time or when the model of the wire harness being tested is changed, a small sample initialization is performed first.

[0169] Select Given a known qualified reference sample, perform all steps S1 to S5 on its circuit to obtain... Contact resistance value Use these values ​​to calculate the sample standard deviation. As Initial value:

[0170] In the formula, To initialize the number of samples, take an empirical value. , dimensionless integer, this parameter is a preset parameter, but can also be adjusted by the implementer according to the acceptable training sample size at the beginning of the production line; For the first The contact resistance measured in this measurement is in ohms. ); For this The arithmetic mean of the measurements, in ohms ( ).

[0171] Calculated Directly assigned As the initial standard deviation of the system, the unit is ohms ( ).

[0172] At the same time, this Absolute deviation for each sample Fill in the sliding window This serves as the initial content of the historical data queue.

[0173] In the subsequent batch testing process, the sliding window method was used. Perform online dynamic updates.

[0174] Predefine a fixed-length first-in-first-out window queue. Window length Get experience points This is a dimensionless integer. This parameter is a preset value, but its specific value can be adjusted by the implementer based on the production line cycle speed and the rate of disturbance change. Larger values... make The updates are smoother but the response is slower, and the smaller... It is more sensitive but more susceptible to interference from individual outliers.

[0175] For each new test completed, the absolute deviation is calculated. The wire harness sample is first judged. Does it meet the requirements? ,in The tolerance coefficient is an empirical value. Dimensionless, this parameter is a preset parameter, but can also be adjusted by the implementer according to the production line's sensitivity to anomalies.

[0176] If the conditions are met, the deviation is considered to fall within the current normal random error range and is caused by routine production fluctuations. Add window The tail of the queue; if the condition is not met, it indicates that the deviation may mainly come from contact defects rather than random interference, and should not be included in the window to avoid contaminating normal fluctuation statistics.

[0177] If the number of data items in the window has reached... If no new data is added, the earliest data entered by the leader of the queue will be removed.

[0178] Then, using the current window Recalculate the standard deviation for each absolute deviation value:

[0179] In the formula, For sliding windows The Middle One absolute deviation value, in ohms ( ); For all within the window The arithmetic mean, in ohms ( ); This is a robustness expansion factor used to compensate for potential underestimation of the standard deviation due to limited window sample size or continuous abnormal disturbances. Get experience points Dimensionless, this parameter is a preset parameter, but can also be adjusted by the implementer according to the frequency and magnitude of extreme deviations in historical data.

[0180] Updated Immediately apply confidence scoring to current and subsequent terminals under inspection calculate.

[0181] thus, It can reflect the overall random error level of the current workstation's testing environment in real time, enabling confidence scoring. It has adaptive anti-interference capabilities.

[0182] Besides the sliding window method, the Bayesian update strategy can also be used to dynamically update the standard deviation of the model online. The complete mathematical implementation process is as follows: S60: Distribution Hypothesis: Detection Bias Follows a normal distribution Standard deviation Conjugate priors are chosen from the inverse Gamma distribution. ; S61: Hyperparameter initialization: Setting baseline prior hyperparameters after small sample calibration =2.5, =0.0001; S62: Posterior update calculation formula: Let the number of valid observation samples in the current sliding window be... Sum of squared sample bias ; The posterior distribution is still the inverse Gamma distribution. ; in: , ; S63: Calculating the effective standard deviation: Take the mean of the inverse Gamma posterior distribution as the updated detection standard deviation: ; S64: Iterative Logic: After each qualified wire harness is inspected, update the deviation statistic S and refresh the synchronous update. , , Used for the next confidence calculation.

[0183] Hyperparameter adaptation adjustment rules: When the noise level is too high at the fine wire harness station, Adjusted to 0.0003; under low-noise conditions with thick conductors, The value was lowered to 0.00005. The posterior distribution of the current batch of ΔR observation data was calculated using Bayes' theorem, and the mean of the posterior distribution was then used as the updated value. .

[0184] This step ultimately outputs the confidence score corresponding to the current terminal circuit under test. The score is dimensionless and quantitatively characterizes the degree of confidence that the terminal contact resistance deviates from the target resistance under historical statistical priors and current noise levels.

[0185] S7: When confidence score If the value is below the preset threshold, it is considered a poor connection quality.

[0186] Step S6 has provided a confidence score for the current terminal circuit under test. The rating is at The interval, dimensionless, quantitatively characterizes the degree to which the terminal contact resistance deviates from the target resistance under historical statistical prior and current field random error levels.

[0187] In continuous production testing, the confidence score of terminals with qualified mating conditions is affected by unavoidable random factors such as residual power frequency interference, quantization noise, and micro-slippage of the probe contact. Slight fluctuations may occur, but statistically they still tend to cluster in the higher value range close to 1; when there are contact defects such as oxidation, incomplete connection, or incomplete insertion of the terminals, the contact resistance will increase. Significantly deviates from the target resistance This leads to absolute deviation. Increase confidence score According to the exponential pattern, it quickly falls to a low value range far below the normal fluctuation range.

[0188] Therefore, qualified products and defective products The values ​​form a natural separation on the number line, which can be achieved by setting a preset threshold. The two should be distinguished.

[0189] Preset threshold Take the empirical value of 0.6, which is dimensionless.

[0190] This threshold is based on a confidence score derived from a large number of known qualified terminal samples. The lower boundary of the statistical distribution is determined, and the standardization calibration process is as follows: S70: Before the system goes online, a calibration sample is collected: ncal=200 wire harnesses of the same model that have been verified by metallographic sectioning and multiple insertion and removal to have no contact defects. S71: Perform the S1~S6 testing process on all samples and generate confidence scores. Sample set {J1, J2, ..., J200}; S71: Sort all scores in ascending order from smallest to largest, and select the lower 5th percentile on one side as the threshold benchmark. The formula for calculating the quantile index is as follows: idx = rounded down (ncal × 0.05) S73: Take the score at the idx-th position after sorting as the initial threshold benchmark, and then fine-tune it according to the production line control standards: ① Highly stringent production lines (prioritizing the elimination of missed inspections): Baseline value lowered by 0.05; ② Mass production workstations (prioritize reducing false alarms): Increase the baseline value by 0.05.

[0191] Example of actual measurement and calibration: The 5th percentile of 200 qualified samples of conventional 0.75mm² automotive wiring harness is 0.61, and the calibration threshold Jth after rounding is 0.6; The quantile of the sample under high noise conditions for thin wires is 0.57, and the actual threshold can be taken as 0.55.

[0192] In this embodiment, a preset threshold is used. An empirical value of 0.6 was used to effectively distinguish between normal vibration and defective dropouts in the verification of various low-voltage harness specifications. This preset threshold... These are preset parameters, but can also be adjusted by the implementer based on the target contact resistance value dispersion of a specific type of terminal and the stringency of the final inspection standards.

[0193] The execution decision process is as follows: The confidence score output in step S6 is... With pre-stored in the detection system Perform numerical comparisons.

[0194] like If so, it is determined that the current terminal circuit under test has poor connection quality; if If the connection is successful, it is considered to be of acceptable quality.

[0195] Due to the model standard deviation Online adaptive adjustment has been achieved through sliding window or Bayesian update in step S6. The distribution has matched the actual random error level on site, therefore a fixed preset threshold is used. It can maintain the stability and consistency of the judgment rules and will not be accidentally triggered due to changes in environmental temperature and humidity or minor differences in workstation equipment.

[0196] The judgment result is output in the form of a logical state variable, with a bad state recorded as 1 and a good state recorded as 0.

[0197] The judgment result is the final quality conclusion of the terminal circuit, and it is recorded in the process database by the detection system.

[0198] This completes the step based on confidence scoring. With preset threshold The system determines the poor connection quality and outputs a conclusion indicating whether the current terminal circuit is defective or acceptable.

[0199] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the insertion quality of low-voltage wire harness terminals, characterized in that, include: A broadband composite voltage excitation signal is injected into the wire harness under test and the current response signal is acquired. The broadband composite voltage excitation signal and current response signal are subjected to frequency domain transformation to separate the extremely low frequency component and the mid-to-high frequency component. The extremely low frequency components are filtered in the unbiased frequency domain to remove power frequency interference, the real part of the extremely low frequency impedance is extracted, and the temperature drift coefficient is calculated based on the real part of the extremely low frequency impedance and the pre-stored nominal resistance. For the mid-to-high frequency components, parasitic capacitive and inductive reactances are removed by multi-frequency impedance circle fitting to obtain the initial equivalent series resistance. The complex impedance data points corresponding to different frequencies are plotted on the real-imaginary plane, and they will converge to the point where... On an impedance circle trajectory centered at a certain point, This is the initial equivalent series resistance; A dispersion compensation factor is constructed based on the nonlinear mapping relationship between the skin effect and the square root of frequency. The nominal resistance is then jointly corrected for temperature drift and skin effect using the temperature drift coefficient and the dispersion compensation factor. Finally, the corrected nominal resistance is subtracted from the initial equivalent series resistance to obtain the terminal contact resistance. The temperature drift coefficient is obtained by dividing the difference between the real part of the ultra-low frequency impedance and the nominal resistance by the nominal resistance. The method for constructing a dispersion compensation factor based on the nonlinear mapping relationship between the skin effect and the square root of frequency includes: obtaining the center frequency ratio of the mid-to-high frequency components to the extremely low frequency components; multiplying the power of the center frequency ratio by a pre-calibrated or statistically learned basic nonlinear gain weight to obtain the dispersion compensation factor; the method for jointly correcting the nominal resistance for temperature drift and skin effect using the temperature drift coefficient and the dispersion compensation factor, and subtracting the corrected nominal resistance from the initial equivalent series resistance to obtain the terminal contact resistance includes: multiplying the pre-stored nominal resistance by the dispersion compensation factor, then multiplying by a value plus the temperature drift coefficient to obtain the corrected nominal resistance; and subtracting the corrected nominal resistance from the initial equivalent series resistance to obtain the terminal contact resistance. A Gaussian distribution reliability scoring model is established based on historical statistical priors, and the model standard deviation is updated online. The confidence score is calculated based on the deviation between the terminal contact resistance and the target resistance. When the confidence score is lower than a preset threshold, it is determined to be a poor connection quality.

2. The method for detecting the insertion quality of low-voltage wire harness terminals according to claim 1, characterized in that, The frequency domain transformation employs the discrete Fourier transform.

3. The method for detecting the insertion quality of low-voltage wire harness terminals according to claim 1, characterized in that, The frequencies corresponding to the extremely low frequency components are selected based on the condition that the skin depth is greater than the conductor radius, and the frequencies corresponding to the mid-to-high frequency components are selected based on the condition that the skin depth is less than the conductor radius.

4. The method for detecting the insertion quality of low-voltage wire harness terminals according to claim 1, characterized in that, The unbiased frequency domain filtering is a windowed notch filtering process for power frequency, and the fixed attenuation factor is abolished to avoid systematic amplitude deviation.

5. The method for detecting the insertion quality of low-voltage wire harness terminals according to claim 1, characterized in that, The online dynamic update of the model standard deviation includes: using a sliding window method or a Bayesian update strategy to update the standard deviation in real time using current and historical detection data; The standard deviation is initialized using a small sample initialization method.

6. The method for detecting the insertion quality of low-voltage wire harness terminals according to claim 1, characterized in that, The target resistance is a pre-calibrated standard contact resistance value.

Images