Low-voltage wire harness short circuit diagnosis method based on electrical parameters

Abstract

The present application belongs to the technical field of electrical variable measurement and electrical fault diagnosis, and particularly relates to a low-voltage wire harness short circuit diagnosis method based on electrical parameters. The method obtains wire harness electrical parameters, determines distributed capacitance time constant and surge amplitude mapping coefficient, and constructs an exponential decay dynamic envelope curve to suppress false triggering caused by capacitive charging surges. After collecting step response current and generating current gradient sequence, the method locks mutation anchor points by detecting gradient breakthrough of continuous sampling points, intercepts local sequence around the anchor points for shallow wavelet packet decomposition, and calculates high-frequency energy concentration. At the same time, the method detrends and calculates steady-state noise variance in the end steady-state region of the current sequence. Finally, the method constructs a logarithmic domain divergence index combining high-frequency energy concentration and steady-state noise variance, and compares it with a dynamic early warning boundary to diagnose short circuit faults. The present application realizes high-precision, low-false-alarm and real-time diagnosis of micro short circuit faults.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement and electrical fault diagnosis technology. More specifically, this invention relates to a method for diagnosing short circuits in low-voltage wiring harnesses based on electrical parameters. Background Technology

[0002] In the online inspection stage of low-voltage wire harness automated production lines, a step voltage is typically used to excite the wire harness under test, and the response current sequence is collected. A differential threshold is then used to determine whether a micro-short circuit has occurred. However, the distributed capacitance of the wire harness itself generates a high-amplitude capacitive charging surge under the action of a step voltage. The initial value of the current gradient caused by this surge is extremely large and decays exponentially over time. Due to the different distributed capacitance parameters of wire harnesses of different specifications, the surge amplitude and decay constant vary significantly. Setting a uniform fixed differential threshold is difficult to balance: if the threshold is too high, small-amplitude short-circuit abrupt changes in the later stage of the surge or the steady-state stage may be missed; if the threshold is too low, the normal large gradient point in the early stage of the surge will frequently exceed the threshold, causing a large number of false triggers. To extract the high-frequency components introduced by the short-circuit action, some solutions use global frequency domain transformation (such as Fourier transform) to process the entire response sequence. However, such methods are computationally complex and resource-intensive, placing a significant burden on the embedded diagnostic controller. Furthermore, they are susceptible to interference from low-frequency power grid harmonics and broadband noise in the industrial environment, affecting the separation effect of high-frequency fault characteristics.

[0003] Furthermore, factors such as production line temperature drift, device aging, and gradual changes in power supply noise floor can cause a slow shift in the response current baseline. Existing methods mostly rely on fixed thresholds or static limits for diagnostic decisions, lacking the ability to adaptively adjust to environmental fluctuations and changes in noise levels. During continuous long-term operation, static criteria will gradually deviate from the true noise distribution, leading to an increased false alarm rate and making it difficult to meet the stringent requirements of production lines for real-time performance, high reliability, and low false alarm rates.

[0004] Therefore, how to dynamically constrain capacitive surges based on the electrical characteristics of different wire harnesses, efficiently extract local high-frequency discharge characteristics, and achieve real-time adaptation of diagnostic criteria to changes in background noise has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a short-circuit diagnosis method for low-voltage wiring harnesses based on electrical parameters, employing the following technical solution: determining the distributed capacitance time constant and surge amplitude mapping coefficient of the wiring harness under test; acquiring step response current and generating a current gradient sequence; constructing an exponentially decaying dynamic envelope curve based on the distributed capacitance time constant and surge amplitude mapping coefficient; when the gradient values ​​of multiple consecutive sampling points in the current gradient sequence exceed the envelope curve, taking the first exceeding point as a sudden change anchor point, including: setting the multiple consecutive sampling points as three consecutive sampling points; when the gradient values ​​of all three consecutive sampling points exceed the envelope curve, taking the first exceeding point as a sudden change anchor point. The first point exceeding the threshold is determined as the transient change time anchor point, serving as the change anchor point. A local sequence is extracted centered on the change anchor point, and the highest frequency energy is extracted by shallow wavelet packet decomposition. The high frequency energy concentration is calculated. The current gradient sequence is detrended at the end and the steady-state noise variance is calculated. Based on the high frequency energy concentration and the steady-state noise variance, a logarithmic domain divergence index containing a zero offset factor and a noise lower limit is constructed. The dynamic warning boundary is calibrated based on the steady-state noise variance and the distribution of historical qualified harness samples. The dynamic warning boundary is a critical abnormal threshold characterizing the harness without faults. The logarithmic domain divergence index is compared with the dynamic warning boundary to diagnose short-circuit faults.

[0006] Preferably, determining the distributed capacitance time constant and surge amplitude mapping coefficient of the tested wiring harness includes: applying a step excitation voltage to the tested wiring harness; obtaining the basic electrical parameters of the tested wiring harness, including wire diameter and length, by interacting with a manufacturing execution system or an RFID tag; calculating the distributed capacitance time constant based on the basic electrical parameters by looking up a table or by using an impedance model; and obtaining the surge amplitude mapping coefficient by reading the ratio of the step excitation voltage to the nominal characteristic impedance of the wiring harness, or by fitting the initial gradient of several preceding sampling points.

[0007] Preferably, the step response current is acquired by: acquiring the step response current according to the distributed capacitance time constant and adapting the sampling frequency to the range of 10 kHz to 100 kHz.

[0008] Preferably, generating the current gradient sequence includes: performing a first-order forward difference operation on the acquired step response current sequence to generate a current transient gradient sequence.

[0009] Preferably, constructing an exponentially decaying dynamic envelope curve based on the distributed capacitance time constant and the surge amplitude mapping coefficient includes: constructing an exponentially decaying function with the surge amplitude mapping coefficient as the initial amplitude and the distributed capacitance time constant as the decay time constant, and superimposing it with the system steady-state noise floor tolerance to form the exponentially decaying dynamic envelope curve.

[0010] Preferably, the local sequence is truncated centered on the mutation anchor point, including: extending a preset short-time compensation interval to both sides of the mutation anchor point, truncating the local feature projection time region to obtain the local sequence; wherein, the length of the short-time compensation interval is a positive integer multiple proportional to the distributed capacitance time constant; during truncation, boundary protection logic is executed: when the mutation anchor point is close to the beginning or end of the sequence, an adaptive window adjustment mechanism is triggered, and by reducing the window on the out-of-bounds side or using a mirror filling method, it is ensured that the truncated length is consistent and no memory addressing error occurs.

[0011] Preferably, the shallow wavelet packet decomposition to extract the highest frequency energy includes: performing three-layer wavelet packet decomposition on the local sequence to extract the wavelet packet coefficients of the highest frequency sub-band; calculating the sum of squares of the coefficients of the highest frequency sub-band to obtain the transient high-frequency radio frequency energy, which is used as the highest frequency energy.

[0012] Preferably, calculating the high-frequency energy concentration degree includes: calculating the sum of squares of all sampling points in the local sequence as the total response energy; setting a zero offset factor; and determining the high-frequency energy concentration degree as the ratio of the highest frequency energy to the sum of the total response energy and the zero offset factor.

[0013] Preferably, the process involves detrending the current gradient sequence at its end and calculating the steady-state noise variance. A logarithmic domain divergence index, including a zero-offset factor and a noise lower limit, is constructed based on the high-frequency energy concentration and the steady-state noise variance. A dynamic warning boundary is then determined. The logarithmic domain divergence index is compared with the dynamic warning boundary to diagnose short-circuit faults. This includes: identifying the steady-state tail section at the end of the current gradient sequence; performing mean-free preprocessing or first-order difference high-pass filtering preprocessing on its current data; calculating the steady-state noise variance; setting an empirical scaling factor and a floor noise lower limit parameter to construct the logarithmic domain divergence index; determining the dynamic warning boundary using a mean-multiple standard deviation method based on the logarithmic domain divergence index distribution of historical qualified samples; and determining the presence of a micro-short-circuit fault and generating a diagnostic alarm command when the logarithmic domain divergence index exceeds the dynamic warning boundary.

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

[0015] 1. Due to the high-amplitude capacitive charging surge introduced by step excitation in the low-voltage harness, a dynamic differential envelope constraint curve with an exponential decay pattern is constructed based on the time constant of the distributed capacitance of the harness under test and the surge amplitude mapping coefficient. This makes the sudden change judgment threshold follow the natural decay law of surge energy, suppressing false triggering caused by normal gradient jumps in the initial charging stage. The anti-jitter criterion logic of gradient values ​​breaking through the envelope at three consecutive sampling points, while tolerating single high-frequency interference spikes, helps to reliably capture the real short-circuit arc with the minimum maintenance pulse width, thereby accurately locking the transient change time anchor point.

[0016] 2. By extracting local current subsequences centered on transient change time anchor points and performing fixed three-layer wavelet packet decomposition in conjunction with out-of-bounds protection logic, the computational scale of frequency domain analysis is reduced from the entire sequence to the local window length, keeping the computational complexity at the linear level and adapting to the real-time processing and memory limitations of embedded controllers. After isolating low-frequency power frequency interference, the three-layer decomposition structure extracts the highest frequency sub-band coefficients and calculates the sum of squares to obtain high-frequency energy. By calculating the high-frequency energy concentration, the proportion of micro-discharge characteristic energy in the total response is directly quantified, enhancing the characterization of short-circuit faults.

[0017] 3. The steady-state tail section is automatically identified from the end of the response sequence. By performing mean removal or first-order difference preprocessing on the current in this section, low-frequency baseline interference such as slow temperature drift is eliminated to obtain the background noise variance. This variance works in conjunction with the anti-zero offset factor and the lower limit of the base noise in the divergence index, so that the diagnostic index can still output an effective relative value in low signal-to-noise ratio scenarios and avoid being overwhelmed by the background noise. The dynamic warning boundary, which is statistically calibrated by historical qualified batch samples, replaces the fixed threshold, so that the short-circuit discrimination criterion adapts to the gradual change of the environmental background noise, reducing the probability of false alarms caused by temperature drift or component aging. Attached Figure Description

[0018] Figure 1 The flowchart illustrating the steps of the low-voltage wiring harness short-circuit diagnosis method based on electrical parameters in this invention is shown in the schematic diagram. 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] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings. (Refer to...) Figure 1 The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters includes steps S1-S7, as follows: S1: Determine the distributed capacitance time constant and surge amplitude mapping coefficient of the tested wiring harness.

[0021] To determine the distributed capacitance time constant and surge amplitude mapping coefficient of the wire harness under test, the production line continuously transmits low-voltage wire harnesses of different specifications. Before executing step excitation and response acquisition, the embedded controller at the test station first interacts with the data interface of the Manufacturing Execution System (MES) or with the reader of the radio frequency identification (RFID) tag on the wire harness tray to obtain the basic electrical parameters of the current wire harness under test.

[0022] Basic electrical parameters include at least wire diameter and harness length , The unit is millimeters (mm). The unit is meter (m).

[0023] The acquisition method is related to the coupling state of the production line information system: when the MES has issued a work order sequence, the controller directly reads the wire harness specification data of the corresponding workstation; if the tag stores specification information, the wire diameter is obtained by decoding through the card reader. and harness length .

[0024] Based on the obtained basic electrical parameters, the controller calculates the distributed capacitance time constant of the measured wire harness. .

[0025] The controller internally stores an empirical table of distributed capacitance per unit length for different wire diameters and insulation structures. This table is obtained through offline measurement and calibration of wire harness samples. Preferably, an LCR tester can be used to measure the distributed capacitance per unit length of wire harnesses of different specifications, and a database of distributed capacitance per unit length can be established based on the measurement results. The corresponding distributed capacitance per unit length can be obtained by looking up the table according to the wire diameter, insulation structure, or wire harness model. The unit is F / m.

[0026] Then calculate the total distributed capacitance of the current harness. , The unit is the farad (F); The length of the wire harness; The capacitance per unit length is the distributed capacitance.

[0027] Equivalent resistance of the test circuit The parameters are composed of the sampling resistor, the fixture contact resistance, and the DC resistance of the wire harness conductor. These are known fixed parameters of the test station, measured in ohms (Ω), and are preset in the controller firmware.

[0028] The distributed capacitance time constant is calculated according to... We obtain the following formula: Equivalent resistance; Total distributed capacitance, in farads. , The unit is microseconds (s).

[0029] Surge amplitude mapping coefficient It is determined by one of the following two methods, depending on whether the controller can currently obtain the accurate nominal characteristic impedance of the harness.

[0030] If the system configuration stores the nominal characteristic impedance of this batch of wire harnesses Then read the amplitude of the step excitation voltage. , The unit is volt (V). The unit is ohms (Ω), directly expressed as the amplitude of the step excitation voltage. With nominal characteristic impedance The ratio is used as a surge amplitude mapping coefficient. ,Right now , The unit is ampere (A), which represents the estimated initial amplitude of the capacitive charging surge current.

[0031] If a reliable nominal characteristic impedance is lacking The data, after the step voltage injection, is compared with the initially acquired data. First-order polynomial least squares fitting is performed on each response current sampling point to obtain the initial current change rate. , The unit is amperes per second (A / s); The empirical value of 5 is taken; it is dimensionless and can be adjusted by the implementer based on the sampling frequency and surge duration.

[0032] The rate of change of the initial current With sampling period Multiply by (in seconds) to obtain an initial surge current amplitude estimate, which is used as... ,Right now .

[0033] Therefore, the distributed capacitance time constant of the current tested wire harness is obtained. Mapping coefficient of surge amplitude .

[0034] S2: Acquire the step response current and generate a current gradient sequence.

[0035] After the step excitation voltage is injected into the test harness through the test fixture, a corresponding response current signal is generated on the conductor of the harness.

[0036] The analog-to-digital converter of the embedded controller synchronously samples the response current, with the sampling frequency based on the distributed capacitance time constant obtained in S1. Adaptation is required.

[0037] time constant The smaller the value, the more drastic the transient changes in the current surge, requiring a higher sampling rate to capture the details of possible micro-short circuit abrupt changes.

[0038] In engineering, to ensure complete acquisition of surge attenuation processes and micro-short-circuit transient characteristics, a preset number of data points must be collected within each distributed capacitance time constant. Therefore, the sampling frequency is set inversely proportional to the distributed capacitance time constant. Determine by the following formula:

[0039] In the formula, The sampling frequency is expressed in Hertz (Hz). The distributed capacitance time constant is expressed in seconds (s). The sampling density factor is taken as an empirical value of 10, which is dimensionless. The lower limit protection term for the time constant is determined by taking an empirical value. (i.e., 1ns), the unit is seconds (s), which can also be adjusted by the implementer according to the highest response speed of the analog-to-digital converter. This item is introduced as a preset parameter to ensure the denominator Even in extreme micro-short circuits or measurement noise causing It will not disappear even when the denominator approaches zero, thus preventing the calculated sampling frequency from overflowing due to the denominator approaching zero, and also preventing the sampling frequency from exceeding the hardware physical limits due to an excessively small time constant; this setting enables the sampling frequency to... It falls within the range of 10 kHz to 100 kHz, and can also be adjusted by the implementer based on the highest sampling capability of the analog-to-digital converter and the pulse width of the micro-short-circuit target signal.

[0040] After the controller outputs the step voltage synchronous trigger signal, it continuously samples the response current at that frequency to obtain a set of discrete response current sequences, denoted as... ,in , This represents the total number of sampling points in this sampling period. The unit is ampere (A).

[0041] Since the raw sampled values ​​obtained after analog-to-digital conversion usually contain common-mode noise and quantization steps, the controller first performs a three-point moving average preprocessing on the sequence, calculated as follows:

[0042] In the formula, This is a smoothed current sequence, in amperes (A).

[0043] Since the first and last points of the sequence lack neighboring points, their original values ​​are directly retained. and .

[0044] This preprocessing operation can suppress the interference of single-point high-frequency glitches on subsequent gradient calculations without significantly changing the transient edge position of the current step.

[0045] Based on the smoothed response current sequence This generates a transient gradient sequence of current.

[0046] Controller responds to current sequence Perform a first-order forward difference operation, that is, calculate the current change between adjacent sampling points:

[0047] In the formula, This is a sequence of current gradients, measured in amperes (A).

[0048] Each point in this sequence directly reflects the response current at the corresponding sampling interval. The amplitude and direction of the jump within.

[0049] For the initial capacitive charging surge region of the step response, the current gradient sequence It exhibits a high-amplitude positive sequence, whose envelope varies with the distributed capacitance time constant. Exponential decay; for the normal flat-top steady-state region, the current gradient sequence The current fluctuates around zero, with the amplitude determined by the system's floor noise. However, when a micro-short circuit occurs, the localized transient breakdown of the conductor introduces an additional steep current rise time within the current gradient sequence. It appears as an isolated positive impact peak superimposed on the decay envelope or steady-state substrate.

[0050] Therefore, the current gradient sequence With the corresponding time index This combination serves as the direct input for subsequent dynamic envelope determination.

[0051] Thus, this step has yielded the response current sequence of the tested harness under the current step excitation. and its corresponding current gradient sequence .

[0052] S3: Construct an exponentially decaying dynamic envelope curve based on the time constant and mapping coefficient.

[0053] Under step voltage excitation, the tested wire harness generates an initial high-amplitude capacitive surge due to the charging of distributed capacitance. The current gradient amplitude corresponding to this surge exhibits an exponential decay law over time.

[0054] The transient current jump caused by a micro-short circuit fault is superimposed on this decay trend in the form of a steep peak.

[0055] If a fixed threshold is used for the current gradient sequence When performing abrupt change detection, if the threshold is too low, the high gradient value in the early stage of the surge will lead to a large number of false alarms; if the threshold is too high, the small-amplitude short-circuit changes in the later stage of the surge and the steady state will not be effectively detected.

[0056] Therefore, it is necessary to construct a dynamic envelope curve that changes synchronously with the surge attenuation law to reflect the current gradient sequence under normal operating conditions. The upper boundary of the decay.

[0057] The envelope curve is constructed directly using the distributed capacitance time constant already determined in S1. Mapping coefficient of surge amplitude .

[0058] The surge decay time constant is expressed in seconds (s). This is the estimated amplitude of the initial surge current, in amperes (A).

[0059] Surge amplitude mapping coefficient Initial amplitude, surge decay time constant An exponential decay term is constructed for the decay time constant, and then the system's steady-state noise floor tolerance is superimposed. This yields the exponentially decaying dynamic envelope constraint function:

[0060] In the formula, For a moment The corresponding envelope constraint value, in amperes (A); A dimensionless factor describing the exponential decay process of surge; The system's steady-state noise floor tolerance is expressed in amperes (A). This represents an exponential function with base e; The surge decay time constant mentioned above is expressed in seconds (s). This is the estimated amplitude of the initial surge current, in amperes (A).

[0061] For practical discrete sampling systems, continuous time... Replace with the specific time corresponding to each sampling point. ,in The sampling interval is... Based on the surge decay time constant in S2 The sampling frequency is determined and is measured in Hertz (Hz).

[0062] Therefore, for the current gradient sequence Each index Calculate the discretized envelope sequence :

[0063] In the formula, The system's steady-state noise floor tolerance is expressed in amperes (A). This represents an exponential function with base e; This is the estimated amplitude of the initial surge current, in amperes (A). , To calculate the discretized envelope sequence in response to the total number of sampling points of the current sequence. With current gradient sequence They have the same length and a one-to-one temporal relationship.

[0064] System steady-state noise floor tolerance The setting is used to absorb the small current gradient fluctuations introduced by electronic noise, quantization error, etc. under normal operating conditions, and to prevent these fluctuations from breaking through the envelope.

[0065] One method for determining this is as follows: During the production line commissioning phase, a step response test is performed on several known defect-free standard wire harnesses, and the current gradient sequence of each wire harness is collected. Then, the steady-state interval after the surge basically ends is selected, and the standard deviation of the gradient value within this interval is calculated. , The unit is ampere (A); take , The coverage factor, taken as an empirical value of 3.0, is dimensionless. This value ensures a sufficiently high probability that normal gradient fluctuations lie below the envelope. It can also be adjusted by the implementer according to the noise level at the site.

[0066] When the production line noise statistics are relatively stable, the steady-state noise floor tolerance of the system can also be set directly. It is a fixed empirical amplitude, but it must be ensured that it is not less than 3 times the standard deviation of the measured steady-state gradient noise.

[0067] Based on the above construction, the discrete envelope sequence At the initial moment The value at that location is This is comparable to the high amplitude gradient level at the beginning of a surge; with index Increase, factor The envelope value gradually approaches zero and eventually converges to the system's steady-state noise floor tolerance. .

[0068] This sequence fully describes the current gradient sequence during the normal capacitive charging surge decay process. The upper boundary of the amplitude variation provides a dynamic constraint that matches the time-varying characteristics of the surge to identify additional current jumps superimposed on the decay trend.

[0069] S4: When the current gradient sequence When the gradient values ​​of multiple consecutive sampling points exceed the envelope curve, the first point of excess is taken as the abrupt change anchor point.

[0070] Obtaining the discretized exponentially decaying envelope sequence and current gradient sequence Afterwards, the controller needs to index from arrive Within the range, point by point, determine whether there is a transient current jump caused by a micro short circuit.

[0071] Due to discrete envelope sequence The upper boundary of gradient value decay during normal capacitive charging surge has been fully characterized, and the current gradient sequence has been established. The component contributed solely by the surge will not exceed this envelope; when the index at a certain moment... A micro-short circuit occurs at a certain point, and the local breakdown of the conductor insulation introduces an additional current rising edge superimposed on the surge, causing the gradient values ​​at that moment and the immediately following moment to increase significantly and exceed the discrete envelope sequence. The defined normal range.

[0072] However, the quantization noise and spatial electromagnetic coupling of the analog-to-digital converter can also affect the current gradient sequence. Isolated high-frequency spikes may be generated. If only a single point is compared, such random spikes may be misjudged as short-circuit events.

[0073] The minimum duration of a micro-short-circuit arc breakdown process is typically on the order of tens to hundreds of microseconds, and is determined by the distributed capacitance time constant. Determined sampling frequency Make the sampling interval The timeframe is between 10 and 100 microseconds, so a real short-circuit transient will span at least three consecutive sampling points.

[0074] Based on this causal characteristic, the controller is set with a number of consecutive breakthrough points. The empirical value of 3 is taken. It is dimensionless and can be adjusted by the implementer based on the relationship between the pulse width and sampling rate of the target short-circuit signal.

[0075] For each index The controller checks simultaneously , and Three consecutive gradient values, if each of them is greater than the envelope value at the corresponding time step. , and Then determine the index. There is a real transient change caused by a short circuit.

[0076] At this point, the controller immediately terminates the traversal and changes the index. The corresponding sampling point is recorded as the transient change time anchor point, which is the change anchor point.

[0077] S5: Extract the local sequence centered on the anchor point, extract the highest frequency energy through shallow wavelet packet decomposition, and calculate the high frequency energy concentration.

[0078] The highest frequency energy is extracted by shallow wavelet packet decomposition, and the high frequency energy concentration is calculated.

[0079] The local sequence is extracted centered on the anchor point, and the highest frequency energy is extracted by shallow wavelet packet decomposition. The high frequency energy concentration is calculated, and the abrupt anchor point is locked by continuous multi-point criteria. The controller then needs to further extract the high-frequency energy characteristics triggered when the transient change occurs.

[0080] Local insulation breakdown caused by a micro-short circuit will superimpose a steep transient oscillatory component on the current response around the anchor point. The energy of this component is mainly concentrated in the high-frequency range, and its duration is approximately several times the distributed capacitance time constant. In regions far from the anchor point, the current response is only a low-frequency decay of capacitive charging surges or a steady-state base, containing almost no effective high-frequency short-circuit information.

[0081] If the current sequence after smoothing the entire sequence Performing global frequency domain analysis not only drastically increases computational overhead, but also dilutes the expression of high-frequency short-circuit characteristics by a large number of signal segments that are irrelevant to the current transient.

[0082] Therefore, with anchor points A local current sequence is extracted from the center, and subsequent frequency domain energy extraction is performed only on this local region.

[0083] Number of half-width sampling points in a local time window From the distributed capacitance time constant and sampling period The decision is made jointly and determined according to the following formula:

[0084] In the formula, The number of points representing the half-width of the window is dimensionless. The distributed capacitance time constant is expressed in seconds (s). The sampling period is expressed in seconds (s). The sampling frequency that has been adapted is in Hertz (Hz). The window coverage factor is taken as an empirical value of 2.5, which is dimensionless. It can also be adjusted by the implementer based on the relationship between the actual duration of the micro-short circuit transient and the sampling rate. This indicates rounding up to the nearest integer.

[0085] Based on this, the theoretical length of the local sequence is obtained. This length is sufficient to cover the main current variation range related to the short-circuit discharge process before and after the anchor point.

[0086] Anchor points may exist during the interception process. Proximity to response current sequence In the case of the beginning or end of the interval, the theoretical cutoff interval is... Partially exceeding the response current sequence Valid index range .

[0087] To avoid array addressing out of bounds and to keep the length of the local sequence constant at the theoretical length. The controller performs boundary protection and uses the mirror filling method to extend the out-of-bounds area.

[0088] Specifically, for the expected position index ,when At that time, by the left boundary As a mirror axis, take As the current value at that location; when At that time, by the right boundary As a mirror axis, take As the current value.

[0089] After mirror filling, the resulting length is the theoretical length. Local current sequence ,in Each point and anchor point The temporal relationships remain clearly corresponding.

[0090] For the truncated local sequence Perform shallow wavelet packet decomposition with a fixed number of layers to separate low-frequency surge components and extract the high-frequency subband where micro-discharges are located.

[0091] This step selects the Daubechies4 (db4) wavelet, which has compact support properties, as the mother wavelet for the local sequence. Perform a three-level wavelet packet transform.

[0092] Wavelet packet transform is a well-known technique that constructs a binary tree-like subband decomposition structure through layer-by-layer high-pass and low-pass filtering and decimation.

[0093] After the three-layer decomposition was completed, a total of [number] were obtained. Each terminal sub-band corresponds to a different frequency band, with nodes... This corresponds to the highest frequency sub-band within the entire analysis frequency band.

[0094] The local current pulse generated by the micro-short circuit has an extremely steep rising edge, and its energy is mainly concentrated in the highest frequency sub-band.

[0095] Controller Extraction Node wavelet packet coefficient sequence , , The number of coefficients in this subband is approximately equal to .

[0096] Based on the highest frequency subband coefficient, the transient high-frequency energy is calculated. :

[0097] In the formula, This is the highest frequency energy, measured in amperes squared (A / m²). ); For the first The highest frequency subband wavelet packet coefficients, in amperes (A).

[0098] This sum of squares measures the total intensity of the high-frequency transient current components excited by micro-short circuits within a local time window.

[0099] At the same time, the total response energy within the time window is calculated directly using the local sequence. :

[0100] In the formula, Total energy, expressed in amperes squared (A / m²). ); For the local sequence of the first Each current sample value is in amperes (A).

[0101] The total energy reflects the overall intensity of all frequency components within that time window, including both high-frequency micro-discharge components and low-frequency residuals and floor noise from capacitive surges.

[0102] Because the total response energy is obtained when the noise is extremely low or the local sequence amplitude is close to zero. It may tend to zero, and using it directly as the denominator will lead to numerical instability.

[0103] Therefore, a zero offset prevention factor is introduced. Take experience points The unit is ampere square (A / m²). This bias value is much smaller than the typical order of magnitude of the sum of squares of local currents under normal operating conditions, and its impact on the accuracy of energy proportion is negligible. It can also be finely adjusted by the implementer based on the system's quantified noise level.

[0104] Based on the above quantities, calculate the local high-frequency energy concentration. :

[0105] In the formula, High-frequency energy concentration, dimensionless; This is the highest frequency energy, measured in amperes squared (A / m²). ); Total energy, expressed in amperes squared (A / m²). ).

[0106] This indicator specifically describes the proportion of the highest frequency subband energy in the total local energy, and the degree of high-frequency energy concentration. The higher the value, the more pronounced the high-frequency transient characteristics of micro-short-circuit discharge within that local region.

[0107] At this point, the controller has obtained the current mutation anchor point. Corresponding local high-frequency energy concentration .

[0108] S6: Detrend the current sequence at the end and calculate the steady-state noise variance.

[0109] In the middle and later stages of the step response, the capacitive charging surge has increased with the distributed capacitance time constant. The current sequence has basically decayed to almost nothing. It enters a quasi-steady-state stage, which is mainly determined by the DC impedance of the test circuit.

[0110] However, in actual production line environments, there are temperature drifts and gradual changes in equipment bias. These low-frequency disturbances are superimposed on the steady-state current, resulting in a slow baseline shift.

[0111] If the variance is calculated directly from the current data of this stage, the baseline offset will significantly increase the variance estimate, so that the variance no longer mainly reflects the transient noise intensity of the system, but is dominated by low-frequency drift energy, thus causing distortion in the noise level measurement.

[0112] Therefore, before calculating the steady-state noise variance, it is necessary to determine the accurate segment in which the current sequence has truly entered the steady state, and to perform detrending preprocessing on the current data in that segment to eliminate baseline drift before calculating the variance.

[0113] The controller is based on the generated current gradient sequence. Automatic identification of the steady-state tail region is achieved.

[0114] For a normal surge decay process, the current gradient magnitude It reaches a maximum in the initial stage, then monotonically decreases and eventually fluctuates slightly randomly around zero; when the current gradient magnitude... It remains at an extremely low level and no longer changes with the index. When the value increases but then decreases in a trend, it indicates that the surge process has ended and the test circuit has entered a steady state.

[0115] The controller uses the current gradient sequence index Starting from point A, check each point one by one until a consecutive number of points are detected. All sampling points satisfy When the condition is met, the first index that satisfies the condition is recorded as the starting index of the steady-state interval. And immediately terminate the search.

[0116] At this point, it is determined that the response current sequence originates from the sampling point. It then enters the steady-state tail region.

[0117] Among them, consecutive points Get experience points Dimensionless, and can be adjusted by the implementer according to the sampling frequency and production line cycle time; minimum judgment threshold. Based on the system steady-state noise floor tolerance determined in step S3 and coverage factor The settings must satisfy all three conditions. .

[0118] System steady-state noise floor tolerance To cover the steady-state gradient under normal operating conditions A tolerance value of multiple standard deviations, covering steady-state gradients. Experience points have been collected. Therefore, the minimum decision threshold is... It essentially reflects the standard deviation level of the steady-state gradient sequence, and its unit is ampere (A).

[0119] This dynamic determination method minimizes the judgment threshold. It can adaptively match the system under test and the noise environment without the need for manual adjustment.

[0120] Determining the starting index Then, the controller extracts the current data within the steady-state tail section, defining it as... , ,in The length of the steady-state interval. This represents the total number of sampling points in the original response current sequence. The unit is ampere (A).

[0121] Perform detrending preprocessing on the data in this interval to eliminate slow, low-frequency baseline drift.

[0122] The preferred method for removing the mean is to calculate the arithmetic mean of the currents over the steady-state range.

[0123] In the formula, The current is the average value over the steady-state interval, and the unit is ampere (A). The unit is ampere (A); is the length of the steady-state interval.

[0124] This leads to the zero-mean current sequence after removing baseline offset.

[0125] The unit is amperes (A). The sequence has been stripped of slow-drift components, retaining only transient noise fluctuating around zero and any possible weak ripples.

[0126] As an equivalent detrending method, the controller can also perform first-order differential high-pass filtering preprocessing on the current in this range, i.e., calculate... , , differential sequence It directly reflects the high-frequency variation components of the current, while its low-frequency spectral components have been suppressed. Therefore, subsequent processing can be directly based on the difference sequence. Variance calculations were performed; the two detrending methods were equivalent in eliminating baseline shift.

[0127] Zero-mean current sequence with baseline offset removed The controller calculates the steady-state noise variance.

[0128] In the formula, The steady-state noise variance is expressed in ampere squares (A / m²). ); is the length of the steady-state interval.

[0129] This variance measures the transient noise intensity of the test system itself after the capacitive surge effect has disappeared, excluding low-frequency drift energy.

[0130] Thus, this step has yielded the steady-state noise variance of the tested wiring harness under current operating conditions. .

[0131] S7: Construct a logarithmic domain divergence index with a zero-offset factor and a noise lower limit based on high-frequency energy concentration and noise variance, relying on steady-state noise variance. The dynamic early warning boundary is calibrated by combining the distribution of qualified samples of historical wire harnesses. The dynamic early warning boundary is the critical abnormal threshold that characterizes the fault-free state of the wire harness. The logarithmic domain divergence index is compared with the dynamic early warning boundary to diagnose short-circuit faults.

[0132] In step S5, obtain the current mutation anchor point. Corresponding local high-frequency energy concentration The steady-state noise variance is obtained in step S6. Then, the controller needs to integrate the two to form a unified diagnostic indicator.

[0133] If we directly consider the high-frequency energy concentration With steady-state noise variance Constructing divergence features using ratios presents two practical problems.

[0134] First, when the tested wiring harness is completely normal and the electromagnetic environment of the workstation is extremely clean, the steady-state noise variance... It may approach a very small value, making First, it becomes a huge quantity, causing numerical overflow or drastic fluctuations in diagnostic indicators; second, high-frequency energy concentration. It is itself composed of steps S5 The calculation is based on the fact that the denominator already includes a zero offset factor. However, the high-frequency energy ratio alone cannot directly reflect the significance of short-circuit characteristics relative to the background noise. The steady-state noise level must be introduced into the denominator as a reference to highlight the outlier degree of abnormal high-frequency energy caused by micro-short circuits in the background noise.

[0135] To construct a diagnostic index that can simultaneously avoid the risk of division by zero and maintain the ability to distinguish between low signal-to-noise ratio conditions, the controller constructs the following logarithmic domain divergence index. :

[0136] In the formula, The divergence index of the logarithmic field is dimensionless; The local high-frequency energy concentration output of step S5 is dimensionless. The steady-state noise variance of the S6 step output is expressed in ampere squares. ); The lower limit of the floor noise is determined by taking an empirical value. The unit is ampere square (A / m²). This parameter can also be adjusted by the implementer based on the lowest measurable noise level on site. It is introduced as a preset parameter to ensure... It will not disappear even in extremely low noise scenarios, thus avoiding the denominator from approaching zero; This is an empirical scaling factor, taken as an empirical value. Units are set as follows Alternatively, the implementer can decide based on the actual situation. The dynamic range requirement is adjusted, and this item is introduced as a preset parameter to adjust the dynamic range requirement. The magnitude is adjusted to the sensitive range of the logarithmic function to prevent logarithmic saturation due to an excessively large ratio or loss of diagnostic sensitivity due to an excessively small ratio.

[0137] Logarithmic operations use natural constants With base, the natural logarithm function Mapping the ratio to a compressed domain keeps the exponent compact over a wide noise range, while the addition of 1 ensures that the logarithmic term is zero as the ratio approaches zero, i.e., the logarithmic domain divergence exponent. The minimum value is zero.

[0138] The logarithmic field divergence index The exponent quantifies the degree of anomaly of the micro-short-circuit high-frequency energy relative to the background noise level in the current step test, and is a logarithmic domain divergence exponent. The higher the value, the greater the concentration of high-frequency energy. The more prominent it is against a noisy background, the higher the likelihood of a micro-short-circuit defect.

[0139] To make the logarithmic field divergence index To transform this into a reliable diagnostic conclusion that adapts to the long-term operation of the production line, the controller does not use a fixed hard threshold, but instead introduces a dynamic early warning boundary based on historical statistics. .

[0140] During production line commissioning or batch verification, a batch of standard wire harnesses, confirmed to be defect-free by manual visual inspection and electrical testing, undergoes the complete process from S1 to S7 consecutively. Diagnostic index sequence of a normal sample .

[0141] Calculate its sample mean based on this normal sample set. and sample standard deviation :

[0142]

[0143] In the formula, and All are dimensionless; is the divergence index in the logarithmic field.

[0144] The lower side of the dynamic warning boundary is given by subtracting a certain number of standard deviations from the mean, but due to the logarithmic domain divergence exponent when micro-short circuits occur... The value will increase, therefore the warning boundary is set above the normal distribution, that is:

[0145] In the formula, For dynamic early warning boundaries; The warning boundary width coefficient is determined by taking an empirical value. It is dimensionless and can be adjusted by the implementer based on the balance between the allowable false alarm rate and the false alarm rate. This item is introduced as a preset parameter.

[0146] This boundary is updated periodically by the controller as the production line environment changes: the controller recalculates the sample mean after each maintenance cycle or whenever a certain number of new qualified samples are accumulated. and sample standard deviation And refresh the dynamic early warning boundary. This allows diagnostic decisions to adaptively track long-term drift caused by environmental background noise, temperature, and equipment aging.

[0147] Within the current test cycle, the controller will use the logarithmic domain divergence index calculated in this step. With the current effective dynamic early warning boundary Compare them.

[0148] like If the test result is positive, the tested wiring harness is determined to have a micro-short circuit fault; otherwise, it is considered normal.

[0149] Once a fault is detected, the controller immediately generates a micro short-circuit diagnostic alarm command. This command drives the audible and visual alarm associated with the production line via the industrial bus to execute the audible and visual alarm. Through the communication interface with the Manufacturing Execution System (MES), the fault status and the time anchor point of the corresponding transient change are highlighted on the MES terminal interface. Simultaneously, the controller sends a logic signal to the automated rejection mechanism, triggering it to move the currently tested wire harness into the defective product isolation area at the sorting station on the production line, completing the closed-loop control from detection and diagnosis to rejection.

[0150] This concludes the final short-circuit fault diagnosis and handling procedure for the currently tested wiring harness.

[0151] 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 short-circuit diagnosis of low-voltage wiring harnesses based on electrical parameters, characterized in that, include: Determine the distributed capacitance time constant and surge amplitude mapping coefficient of the tested wiring harness; Collect the step response current and generate a current gradient sequence; An exponentially decaying dynamic envelope curve is constructed based on the distributed capacitance time constant and surge amplitude mapping coefficient. When the gradient values ​​of multiple consecutive sampling points in the current gradient sequence exceed the envelope curve, the first exceeding point is taken as the abrupt change anchor point, including: setting the multiple consecutive sampling points as three consecutive sampling points; When the gradient values ​​of the three consecutive sampling points all exceed the envelope curve, the first exceeding point among the three sampling points is determined as the transient change time anchor point, and is used as the change anchor point. Using the mutation anchor point as the center, a local sequence is extracted, and the highest frequency energy is extracted by shallow wavelet packet decomposition. The high frequency energy concentration is then calculated. The current gradient sequence is detrended at the end and the steady-state noise variance is calculated. Based on the high-frequency energy concentration and steady-state noise variance, a logarithmic domain divergence index containing a zero offset factor and a noise lower limit is constructed. The dynamic early warning boundary is calibrated by combining the steady-state noise variance with the distribution of qualified samples of historical harnesses. The dynamic early warning boundary is a critical abnormal threshold characterizing the harness without faults. The logarithmic domain divergence index is compared with the dynamic early warning boundary to diagnose short-circuit faults.

2. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, Determine the distributed capacitance time constant and surge amplitude mapping coefficient of the tested wiring harness, including: A step excitation voltage is applied to the wire harness under test; The basic electrical parameters of the wire harness under test are obtained by interacting with the manufacturing execution system or radio frequency identification tag. The basic electrical parameters include wire diameter and length. Based on the aforementioned basic electrical parameters, the distributed capacitance time constant is calculated either by looking up a table or by relying on an impedance model. The surge amplitude mapping coefficient is obtained by reading the ratio of the step excitation voltage to the nominal characteristic impedance of the harness, or by fitting the initial gradient of several preceding sampling points.

3. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The acquisition of the step response current includes: acquiring the step response current by adapting the sampling frequency from 10 kHz to 100 kHz according to the distributed capacitance time constant.

4. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The process of generating the current gradient sequence includes: performing a first-order forward difference operation on the acquired step response current sequence to generate a transient current gradient sequence.

5. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The construction of the exponentially decaying dynamic envelope curve based on the distributed capacitance time constant and the surge amplitude mapping coefficient includes: constructing an exponentially decaying function with the surge amplitude mapping coefficient as the initial amplitude and the distributed capacitance time constant as the decay time constant, and superimposing it with the system steady-state noise floor tolerance to form the exponentially decaying dynamic envelope curve.

6. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The step of extracting a local sequence centered on the mutation anchor point includes: extending a preset short-time compensation interval to both sides of the mutation anchor point, extracting a local feature projection time region, and obtaining the local sequence. Wherein, the length of the short-time compensation interval is a positive integer multiple that is directly proportional to the distributed capacitance time constant; Boundary protection logic is executed during truncation: when the mutation anchor point is close to the beginning or end of the sequence, an adaptive window adjustment mechanism is triggered to ensure consistent truncation length and prevent memory addressing errors by reducing the window on the out-of-bounds side or using mirror filling.

7. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The shallow wavelet packet decomposition to extract the highest frequency energy includes: performing a three-layer wavelet packet decomposition on the local sequence to extract the wavelet packet coefficients of the highest frequency subband; The sum of squares of the highest frequency subband coefficients is calculated to obtain the transient high-frequency radio frequency energy, which is used as the highest frequency energy.

8. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The calculation of high-frequency energy concentration includes: calculating the sum of squares of all sampling points in the local sequence as the total response energy; Set the anti-zero offset factor; The ratio of the highest frequency energy to the sum of the total response energy and the anti-zero offset factor is determined as the high frequency energy concentration degree.

9. The low-voltage wiring harness short-circuit diagnosis method based on electrical parameters according to claim 1, characterized in that, The process of detrending the current gradient sequence at its end and calculating the steady-state noise variance, constructing a logarithmic domain divergence index with a zero-offset factor and a noise lower limit based on the high-frequency energy concentration and the steady-state noise variance, determining the dynamic warning boundary, and comparing the logarithmic domain divergence index with the dynamic warning boundary to diagnose short-circuit faults includes: identifying the steady-state tail section at the end of the current gradient sequence, performing mean-free preprocessing or first-order difference high-pass filtering preprocessing on its current data, and calculating the steady-state noise variance; setting an empirical scaling factor and a floor noise lower limit parameter to construct a logarithmic domain divergence index; determining the dynamic warning boundary based on the logarithmic domain divergence index distribution of historical qualified samples using a mean plus standard deviation method; and determining the existence of a micro-short-circuit fault and generating a diagnostic alarm command when the logarithmic domain divergence index exceeds the dynamic warning boundary.

Images