Ignition phenomenon recognition processing system based on pulse feature fusion

Abstract

The present application relates to the technical field of electric parameter monitoring, and discloses a spark phenomenon identification processing system based on pulse feature fusion, comprising: a signal acquisition unit for acquiring controlled loop voltage and current signals; a feature preprocessing unit for extracting voltage and current change rates by using a hardware differential circuit; an arbitration processing unit for obtaining a change rate product by using a hardware multiplier to generate a correlation signal, and separating a high-frequency burr signal by using a high-pass filter circuit, and outputting a spark identification signal when the correlation signal crosses a negative determination threshold and the high-frequency burr signal pulse count exceeds a randomness threshold; and an output processing unit for executing loop locking according to the spark identification signal, wherein the present application utilizes the physical difference between random jump and smooth characteristics of a controlled switch in the discharge process, constructs time domain and topological dimension criteria, eliminates high-frequency power electronic commutation interference, reduces identification time delay, and guarantees the operation safety of the controlled loop.

Description

Technical Field

[0001] This invention belongs to the field of electrical parameter monitoring technology, and in particular relates to an arcing phenomenon identification and processing system based on pulse feature fusion. Background Technology

[0002] Current power supply circuit anomaly monitoring typically involves collecting voltage or current signals in the circuit and determining the operating status through preset amplitude thresholds. This serves as a common means of ensuring circuit safety. However, in process environments involving high-frequency inverter power supplies, such as semiconductor vapor deposition, the rapid switching action of power devices generates high-frequency electromagnetic noise, and the circuit impedance is in a dynamic evolution state. This causes the discharge pulse generated by the breakdown of the load dielectric to be submerged in the normal inverter switching noise.

[0003] This judgment logic based on the amplitude of a single variable faces a contradiction between detection sensitivity and operational stability. If the threshold is lowered to capture weak arcing, transient pulses caused by load switching or power grid fluctuations will lead to frequent false alarms. If the threshold is raised to maintain stability, the initial small arcing signal will be masked, resulting in the risk of missed detection and cumulative damage to high-precision loads caused by discharge energy. For example, Chinese invention patent CN107515386B discloses a control and protection module for microwave power modules. It uses FPGA to identify the input pulse width and repetition rate and uses pulse nesting generation technology to prevent arcing induced by abnormal operating ratio of traveling wave tube. Such solutions belong to macroscopic pre-inspection of control command compliance. The monitoring object is limited to time-domain shape parameters such as pulse width and repetition rate, without touching the physical essence of discharge.

[0004] Therefore, how to achieve real-time and accurate identification of arcing phenomena by utilizing the physical causal relationship of circuit parameters under strong noise and high-frequency switching backgrounds has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a sparking phenomenon identification and processing system based on pulse feature fusion, comprising:

[0006] The signal acquisition unit is used to synchronously acquire the voltage and current signals of the controlled circuit;

[0007] The feature preprocessing unit, connected to the signal acquisition unit, includes a hardware differential circuit for extracting the rate of change of the voltage signal and the rate of change of the current signal.

[0008] The arbitration processing unit is connected to the feature preprocessing unit and includes a hardware multiplier, a high-pass filter circuit, a comparator array, and a logic AND gate.

[0009] The hardware multiplier is used to multiply the rate of change of the voltage signal by the rate of change of the current signal to generate the associated intensity signal R. m ;in, dv / dt is the rate of change of the voltage signal, and di / dt is the rate of change of the current signal;

[0010] The high-pass filter circuit is connected to the output of the hardware multiplier and is used to filter the associated strength signal R. m High-frequency spike signals that characterize the discontinuous jump features of the discharge process are separated in the middle.

[0011] A comparator array is used to monitor the correlation strength signal R. m Whether the negative amplitude crosses the preset product judgment threshold, and whether the pulse count value of the high-frequency glitch signal within the preset time window exceeds the preset randomness threshold;

[0012] A logic AND gate is connected to a comparator array for use in the correlation strength signal R. m When the preset product determination threshold is crossed and the pulse count value exceeds the preset randomness threshold, an ignition identification signal is output.

[0013] The output processing unit, connected to the arbitration processing unit, is used to generate a circuit blocking command based on the ignition identification signal.

[0014] Preferably, the arbitration processing unit further includes a trajectory analysis module, which is used to map the rate of change of the voltage signal and the rate of change of the current signal extracted by the hardware differential circuit to a two-dimensional phase plane, and calculate the instantaneous curvature of the running trajectory in the two-dimensional phase plane; the arbitration processing unit is used to output an ignition identification signal when the instantaneous curvature has a discontinuous jump pole and the amplitude of the discontinuous jump pole deviates from the preset controlled switch envelope.

[0015] Preferably, the arbitration processing unit is connected to a phase synchronization module for acquiring the modulated carrier phase of the controlled loop; the arbitration processing unit is used to divide the identification period into a sensitive time window and a shielded time window according to the modulated carrier phase; the arbitration processing unit maintains the initial sensitivity of the preset randomness threshold within the sensitive time window and increases the preset randomness threshold within the shielded time window.

[0016] Preferably, it also includes a reference calibration module, which is used to calculate the fluctuation ratio of the voltage signal and the current signal to update the loop impedance characteristic value during the steady-state operation cycle of the controlled loop; the arbitration processing unit dynamically corrects the preset product judgment threshold based on the loop impedance characteristic value.

[0017] Preferably, it also includes a waveform verification module, connected to the arbitration processing unit, used to obtain the second-order change D of the slope of the pulse recovery segment of the voltage signal after the ignition identification signal is triggered. r ;in, v r For the voltage component of the pulse recovery segment; when the second-order change D r When the plasma nonlinear composite model is satisfied, the arbitration processing unit executes the output action of the ignition identification signal.

[0018] Preferably, the hardware differential circuit in the feature preprocessing unit includes an analog differential circuit constructed from an operational amplifier, used to maintain the system's recognition response delay within 1 μs by extracting the real-time slope of the analog signal.

[0019] Preferably, the comparator array in the arbitration processing unit includes a multi-window comparator for processing the correlation strength signal R separately. m Threshold discrimination is performed based on the negative peak amplitude and the pulse density of the high-frequency glitch signal.

[0020] Preferably, it also includes a wavefront feature monitoring module for extracting the correlation intensity signal R. m Triggering instantaneous wavefront transition time T S When the transition time T of the wavefront S When the power device is within the preset inherent switching time range, the arbitration processing unit shields the arcing identification signal.

[0021] Preferably, it also includes a trend analysis module, connected to the arbitration processing unit, for continuously acquiring the correlation intensity signal R during periods when no ignition identification signal is output. m The low-amplitude residual sequence is obtained and the energy variance of the low-amplitude residual sequence is calculated; when the energy variance shows a non-linear growth trend, an early warning signal for insulation degradation of the output circuit is generated.

[0022] Preferably, the output processing unit is connected to a fast circuit breaker actuator for disconnecting the power input contacts of the controlled circuit within 500 ns after receiving a circuit blocking command.

[0023] Furthermore, compared to existing technologies, the spark detection and processing system based on pulse feature fusion of this invention has the following advantages:

[0024] 1. In the ignition phenomenon of pulse feature fusion, the physical correlation arbitration unit utilizes the polarity product feature of the transient voltage signal change rate and the transient current signal change rate to anchor the identification logic to the physical evolution process of loop impedance collapse. This enables the system to effectively distinguish between nonlinear breakdown discharge on the load side and normal power regulation at the source end, overcoming the contradiction between sensitivity and false alarm rate in traditional threshold determination methods. Without relying on complex digital signal processing, it improves the ability to capture weak initial ignition signals.

[0025] 2. The collaborative mechanism formed by the random feature extraction module and the phase synchronization module utilizes the deterministic trajectory of the switching action of the controlled power device and the random avalanche characteristics of the plasma discharge process to construct a dual filtering barrier in the time domain and topological dimension. This enables the system to identify the essential differences in physical evolution, eliminates the systematic interference generated by high-frequency wide-bandgap semiconductor devices during nanosecond-level switching processes from the bottom layer, and ensures the stability of the identification logic in harsh electromagnetic environments.

[0026] 3. The reference calibration module dynamically corrects the judgment threshold center point in the physical verification equation by real-time online evaluation of the circuit distributed impedance during the quasi-steady-state period. This eliminates the problem of threshold silent drift caused by cable aging, contact resistance fluctuations, and environmental temperature drift, ensuring that the identification system maintains consistent detection accuracy throughout the entire equipment life cycle and reducing the frequency of regular maintenance and manual calibration costs in industrial settings. The phase plane curvature monitoring mechanism introduced by the trajectory analysis module maps one-dimensional time-domain electrical variables to two-dimensional geometric topology space. By utilizing the continuity difference between the trajectory of linear damped oscillation and nonlinear breakdown pulse in phase space, it achieves physical-level decoupling of ultra-high frequency resonant noise, enabling the system to maintain accurate state judgment even under extreme conditions where the power conversion frequency and ignition frequency highly overlap. Attached Figure Description

[0027] Figure 1 This is a block diagram of the signal flow and hardware topology of the spark detection and processing system of the present invention;

[0028] Figure 2 This is a schematic diagram showing the multi-dimensional technical branches and functional architecture decomposition of the sparking phenomenon identification and processing system of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0030] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, low, lateral, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated.

[0031] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0032] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0033] This invention provides a sparking phenomenon identification and processing system based on pulse feature fusion, comprising a signal acquisition unit, a feature preprocessing unit, an arbitration processing unit, and an output processing unit. The signal acquisition unit is connected to the feature preprocessing unit and is used to synchronously acquire voltage and current signals of the controlled circuit. The feature preprocessing unit is connected to the arbitration processing unit, which includes a hardware differential circuit for extracting the rate of change of the voltage signal and the rate of change of the current signal. The arbitration processing unit includes a hardware multiplier, a high-pass filter circuit, a comparator array, and an AND gate. The hardware multiplier multiplies the rate of change of the voltage signal and the rate of change of the current signal to generate a correlation intensity signal R. m The high-pass filter circuit is connected to the output of the hardware multiplier and is used to filter the associated strength signal R. m High-frequency glitch signals are separated in the middle; the comparator array monitors the correlation strength signal R. m Whether the negative amplitude crosses a preset product judgment threshold, and whether the pulse count value of the high-frequency glitches signal within a preset time window exceeds a preset randomness threshold; a logic AND gate is connected to a comparator array to be used in relation to the strength signal R. m When the preset product determination threshold is exceeded and the pulse count value exceeds the preset randomness threshold, an ignition identification signal is output; the output processing unit is connected to the arbitration processing unit to generate a loop blocking command based on the ignition identification signal.

[0034] The hardware differential circuit in the feature preprocessing unit includes an analog differential stage constructed from operational amplifiers. This stage extracts the slope of the front-end analog signal in real time, keeping the system's recognition response delay within 1μs. The hardware differential circuit uses a high-speed operational amplifier to construct an active differential topology. A high-frequency suppression capacitor in parallel with the feedback loop limits the self-oscillation caused by the overflow of the gain-bandwidth product. The time constant of the differential stage is set to 50ns, providing full-range gain for breakdown steep edges above 10MHz. The extraction delays of the analog signal dv / dt and di / dt are controlled within a physical noise floor of 150ns. The output signal is fed into the input of the hardware multiplier via a zero-drift compensation circuit, and the mapping relationship follows... The parameter configuration is based on the fidelity test of a 2MHz carrier environment signal. The hardware differential circuit converts the transient voltage signal V of the input circuit. t With transient current signal I t Converted to the corresponding rate of change of voltage signal dv / dt and rate of change of current signal di / dt; the hardware multiplier in the arbitration processing unit calculates the correlation strength signal R according to the physical conservation law. m The calculation relationship is Where dv / dt is the rate of change of the voltage signal, in V / μs, and di / dt is the rate of change of the current signal, in A / μs; during arcing and discharge, the voltage drop causes dv / dt to be negative, and the current surge causes di / dt to be positive, generating the associated intensity signal R. m It exhibits negative spike characteristics; the multi-window comparators in the comparator array respectively compare the correlation strength signal R m The negative peak amplitude and the pulse density of the high-frequency glitch signal are used for threshold discrimination; when the above dual threshold discrimination conditions are met, the logic AND gate triggers the output processing unit, and the fast circuit breaker connected to the output processing unit disconnects the power input contact of the controlled circuit within 500ns after receiving the circuit blocking command.

[0035] The arbitration processing unit also includes a trajectory analysis module. This module maps the rate of change of the voltage signal (dv / dt) and the rate of change of the current signal (di / dt) extracted by the hardware differential circuit to a two-dimensional phase plane and calculates the instantaneous curvature of the trajectory within the two-dimensional phase plane. The curvature calculation circuit in the arbitration processing unit uses the discrete-point slope change rate to calculate the instantaneous curvature. The rate of change sequences x1, x2, x3 of the voltage signal and y1, y2, y3 of the current signal output from three adjacent sampling points of the hardware differential circuit are used as coordinate inputs to the phase plane. The cosine of the angle between adjacent vectors is calculated to characterize the curvature of the trajectory. The decision logic limits the linear damped oscillation trajectory generated by the controlled switching process to a preset smooth evolution envelope. When the instantaneous curvature jumps across... When the resonant curvature exceeds the upper limit determined by the distributed inductance and capacitance of the circuit, the system identifies the discontinuous break transition as a characteristic pulse generated by nonlinear breakdown and outputs an arcing identification signal. For asynchronous electromagnetic interference pulses, the arbitration processing unit is connected to a phase synchronization module. The phase synchronization module acquires the power modulation carrier phase of the controlled circuit in real time. The arbitration processing unit divides the identification period into a sensitive time window and a shielded time window based on the power modulation carrier phase. Within the high-stress sensitive time window of the power system, the arbitration processing unit maintains the initial sensitivity of a preset randomness threshold. Within the shielded time window corresponding to the switching action of the power device, the arbitration processing unit increases the preset randomness threshold, thereby reducing false triggering caused by environmental random pulse interference. For the hardware group delay T generated by the sensor transmission link and analog conditioning circuitry... d The system executes a phase alignment procedure, and the phase synchronization module captures the reference carrier phase output by the phase-locked loop. Synchronously record the timestamp T of the voltage signal crossing zero. z Calculate the time deviation between the two to determine the phase deviation compensation factor. ; wherein, the The calculation formula is Where f is the modulation carrier frequency, in Hz, and T d The measured hardware group latency is expressed in seconds (s); based on this, the arbitration processing unit utilizes... The starting boundary of the shielding window is corrected in real time to cover the voltage oscillation peak generated during the commutation of power devices, ensuring that high-frequency switching noise is limited to the improved randomness filtering range. This eliminates logical misjudgments caused by sampling phase shift without affecting the system's sensitivity to weak arcing.

[0036] The system includes a reference calibration module, used to acquire the fluctuation ratio of voltage and current signals during the steady-state operation cycle of the controlled circuit; the algorithm updates the circuit impedance characteristic value reflecting the current line state in real time based on this ratio; the arbitration processing unit dynamically corrects the center point of the preset product judgment threshold based on the circuit impedance characteristic value; this process can periodically correct measurement deviations caused by line aging or environmental temperature drift; the system includes a waveform verification module, connected to the arbitration processing unit, used to acquire the voltage signal after the ignition identification signal is triggered; the waveform verification module calculates the second-order change Dr of the voltage signal slope in the pulse recovery segment, the calculation relationship is as follows: Where vr is the voltage component of the pulse recovery segment, in units of V; the actual ignition discharge process involves the extinction of the plasma channel, and its voltage recovery exhibits a nonlinear logarithmic decay characteristic; when the second-order change Dr conforms to the preset plasma nonlinear composite model, the arbitration processing unit maintains the output action of the ignition identification signal, realizing closed-loop verification of the entire discharge evolution process; the system includes a wavefront feature monitoring module, used to extract the correlation intensity signal R m Triggering instantaneous wavefront transition time T S The controlled switching edge of the power device is controlled by the drive circuit, and its wavefront transition time T S It exhibits consistency; when the wavefront transition time T S When the pulse is within the preset inherent switching time range of the power device, the arbitration processing unit identifies the pulse as the expected hardware action and blocks the ignition identification signal; the system includes a trend analysis module connected to the arbitration processing unit; during the period when no ignition identification signal is output, the trend analysis module continuously acquires the correlation intensity signal R. m The system generates a low-amplitude residual sequence and calculates the energy variance of the low-amplitude residual sequence using a statistical unit. When the energy variance exhibits a non-linear growth trend, the system outputs a circuit insulation degradation early warning signal, enabling predictive monitoring of the circuit status.

[0037] Example 1: In the semiconductor plasma-enhanced vapor deposition (PEVDC) production process, a silicon carbide high-frequency pulse power supply with a switching frequency of 2MHz is used. Due to the high voltage signal change rate dv / dt and high current signal change rate di / dt generated by the switching of silicon carbide devices, the controlled electromagnetic pulse overlaps with the arcing signal generated inside the cavity in both the time and frequency domains. This causes the conventional amplitude triggering system to falsely block out when maintaining high-sensitivity detection. When dielectric breakdown occurs under this condition, the signal acquisition unit synchronously acquires the voltage signal V of the controlled circuit. t With current signal I tThe feature preprocessing unit extracts the rate of change of the voltage signal dv / dt and the rate of change of the current signal di / dt through a hardware differential circuit; the hardware multiplier in the arbitration processing unit generates the correlation strength signal R based on the product of the rate of change of the voltage signal dv / dt and the rate of change of the current signal di / dt. m The calculation relationship is R. m = (dv / dt) × (di / dt), where dv / dt is the rate of change of the voltage signal in V / μs, and di / dt is the rate of change of the current signal in A / μs.

[0038] During the power commutation cycle of a silicon carbide device, the rate of change of the voltage signal dv / dt and the rate of change of the current signal di / dt change in the same direction, and the correlated intensity signal R m It is a positive value; when plasma arcing occurs inside the cavity, causing the load impedance to collapse, the voltage drops and the current rises, and the associated intensity signal R... m Reversing to a negative value and crossing the preset negative product determination threshold Th m Simultaneously, the high-pass filter circuit obtains the associated strength signal R m The high-frequency glitches reflecting the randomness of discharge are separated. The comparator array, combined with the power modulation carrier phase information provided by the phase synchronization module, counts the pulses within a preset time window Tw to determine whether they reach a preset randomness threshold N. rand ; AND gates in the correlation strength signal R m Crossing the preset product determination threshold Th m And the pulse count value exceeds the preset randomness threshold N rand When the ignition is detected, an ignition identification signal is output. The ignition identification signal drives the output processing unit to generate a loop blocking command. The fast circuit breaker actuator disconnects the power input contact of the controlled circuit within 500ns after receiving the command. Since the closed-loop delay of identification and blocking is lower than the destructive threshold of plasma energy accumulation, the discharge energy is prevented from causing physical damage to the wafer surface. The system utilizes the physical difference between the polarity determinism of the controlled switch action and the random fluctuation of the ignition process to resolve the contradiction between sensitivity and adaptability.

[0039] Example 2: In a test environment simulating a semiconductor vacuum plasma load, a 50kW RF power supply circuit was used. The test data came from real-time sequences collected by the physical test bench. The measuring instrument had a bandwidth of no less than 500MHz, a sampling rate of no less than 1GS / s, and a resolution of 12bit, capturing sub-microsecond transient pulse details. The sampling period Tw was set by the controlled switch frequency and the duration of the ignition pulse. The setting of the sampling period Tw balanced the real-time performance of data acquisition with the system processing load. When the controlled switch frequency was around 2MHz, the sampling period Tw was set to 400ns to avoid the influence of switch ringing. The sample group of this invention applied an ignition phenomenon recognition and processing system based on pulse feature fusion, while the control group used a voltage amplitude threshold determination method.

[0040] The test environment was actively injected with Gaussian white noise with a signal-to-noise ratio of 20dB, and a damped oscillation interference with a frequency of 10MHz caused by the distributed inductance of the feeder was simulated. The peak amplitude of the voltage signal change rate dv / dt was set to 1.2 times that of normal switching operation. When the system was in a quasi-steady-state period, the reference calibration module calculated the characteristic value K of the loop impedance under steady-state conditions. z The impedance is 4.15Ω; the arbitration processing unit, based on the impedance characteristic value, determines the preset product threshold Th. m The system is calibrated to the order of -6500 units. Before performing dual threshold discrimination, the system constructs a baseline for the randomness distribution characteristics. Within the shielded time window of the controlled power device under high-frequency carrier modulation, the high-pass filter circuit removes high-frequency features from the background noise. The statistical unit accumulates the count of weak pulses generated by the charging and discharging of stray capacitance within a preset time window Tw and constructs a Poisson distribution model. The system sets the preset randomness threshold N... rand The upper edge trigger value of the Poisson distribution model at 99.9% confidence level is set. The count value of environmental interference pulses is limited to the non-trigger area using statistical principles, so that the logic AND gate only produces a decision output when facing a real discharge pulse with avalanche multiplication effect. Table 1 records the key physical characteristic data under background noise interference, normal switching and different gradient ignition conditions. See Table 1.

[0041] Table 1: Key Physical Characteristic Data Table

[0042]

[0043] According to the measured values ​​recorded in Table 1, when the sample group of this invention is under typical ignition condition B, the rate of change of the voltage signal dv / dt is -215.6V / μs, the rate of change of the current signal di / dt is 48.3A / μs, and the generated correlation intensity signal R m The value is -10413, which exceeds the preset product determination threshold Th. mAt this time, the high-frequency glitches separated by the high-pass filter circuit generate a pulse count of 42 within the sampling period Tw, which exceeds the preset randomness threshold N. rand The logic AND gate outputs an ignition recognition signal, with a total recognition response time of 785ns. When controlled actions such as power transistor turn-off occur, the voltage change rate amplitude is large, but because the voltage and current changes are in the same direction, the associated strength signal R... m If the value is positive, the judgment logic is not triggered; during the static calibration phase at the initial stage of system launch, the residual noise sequence of the controlled circuit under no-discharge steady state is obtained through the signal acquisition unit, and the correlation intensity signal R within the corresponding time period is extracted using the feature preprocessing unit. m The baseline fluctuation distribution is used to calculate the correlation strength signal R by the statistical unit. m The mean μ and standard deviation σ over 1000 consecutive sampling periods; subsequently, the preset product judgment threshold Th... m The value is set to μ-5σ to cover 99.99% of the system thermal noise range. At the same time, the power modulation carrier is turned on under zero load conditions. The high-frequency component of the signal output by the hardware multiplier is extracted by the high-pass filter circuit. The random pulse count value generated by the charging and discharging of the circuit stray capacitor within the preset time window is counted. The integer value of the count value at the upper limit of 95% confidence is selected as the preset randomness threshold according to the Poisson distribution model, thereby establishing a quantitative judgment benchmark based on the current hardware noise floor environment.

[0044] Correlation strength signal R m The negative amplitude is lower than the preset product determination threshold Th m In situations like weak ignition (condition A), the system detects signal reversal, but the energy level does not reach the judgment threshold, and the pulse count is small, thus determining it as normal. This process prevents unnecessary circuit blocking induced by extremely small discharges. When the product of the rate of change of the voltage signal (dv / dt) or the rate of change of the current signal (di / dt) crosses the judgment threshold, the randomness count increases with the increase of ignition intensity, reflecting the system's quantitative monitoring capability of discharge evolution intensity. In the partial missing feature control group, when only the rate of change of the single voltage signal is used for judgment, the system cannot distinguish between resonant ringing pulses and ignition signals, generating false alarms at 1550 ns. The system utilizes the polarity characteristics of controlled switch action and the random characteristics of plasma discharge to identify uncontrolled discharges in a background noise environment. Test results show that the preset product judgment threshold Th... m With preset randomness threshold N rand The dual criteria filter out high-frequency commutation interference at the megahertz level, maintaining the identification accuracy above 99%, and driving the fast circuit breaker to disconnect the power input contacts of the controlled circuit within 500ns.

[0045] Example 3: This example combines Figures 1 to 2 The following describes a system for identifying and processing sparking phenomena based on pulse feature fusion, such as... Figure 1 As shown, the overall topology of the system begins with the controlled loop. The signal acquisition unit is responsible for synchronously acquiring the voltage and current signals of the controlled loop and transmitting them to the feature preprocessing unit. The feature preprocessing unit uses a hardware differential circuit to extract the rate of change and then sends the signal to the arbitration processing unit. Inside the arbitration processing unit, the voltage / current rate of change is multiplied by a hardware multiplier to generate a correlation strength signal. At the same time, the high-pass filter circuit is responsible for separating high-frequency glitches. The comparator array monitors the correlation strength signal and the high-frequency glitches to determine the threshold of amplitude and pulse count. When the determination condition is met, the logic AND gate outputs an arc identification signal. This signal drives the output processing unit to generate a loop blocking command, which ultimately controls the fast circuit breaker actuator to protect the circuit.

[0046] like Figure 2 As shown, the main architecture of the arcing phenomenon identification and processing system is divided into four technical branches: feature extraction, logic arbitration, physical verification, and system assurance. The feature extraction branch includes synchronous voltage and current acquisition, hardware differential extraction of change rate, and analog differential low-latency processing elements. The logic arbitration branch consists of hardware multiplication to generate associated signals, high-frequency glitch signal separation, and dual thresholds for amplitude and count. The physical verification branch covers waveform second-order change verification, sensitivity and shielding time window synchronization, and phase plane trajectory curvature analysis mechanism. The system assurance branch integrates nanosecond-level fast circuit breaking, insulation degradation trend early warning, and impedance reference dynamic calibration functions.

[0047] Example 4: When the distributed impedance of the controlled circuit drifts due to changes in the length of the transmission cable or oxidation of the connection terminals over a long period of operation, the system initiates a reference calibration procedure to recalibrate the physical verification parameters; the controlled circuit is controlled to enter a no-discharge operation cycle, during which the reference calibration module continuously acquires voltage and current signals for 50 power frequency cycles and calculates the effective value V of the voltage signal. rms With the effective value of the current signal I rms The ratio, which is defined as the characteristic value K of the loop impedance. z The arbitration processing unit uses a calibration procedure to determine the preset product judgment threshold Th. m The calculation relationship is Th m =-1500×K z , where K z This is the characteristic value of the loop impedance, in Ω; when the reference calibration module measures the characteristic value of the loop impedance K... z When the Ω is 5.2Ω, the system will determine the preset product threshold Th according to the above calibration procedure. mThe value has been updated to the -7800 unit level. This process ensures that the judgment benchmark is anchored on the static impedance base of the current circuit, eliminating silent deviations caused by hardware differences. For periodic interference generated by commutation of silicon carbide power devices, the power modulation carrier phase is extracted via a phase-locked loop. The execution time window division method is as follows: the phase within a single carrier cycle is divided into... In as well as The interval is divided into a shielding time window, and the remaining phase intervals are divided into sensitive time windows; for sensor group delay compensation, the system starts standard sinusoidal excitation to extract the carrier zero-crossing point. Phase-locked loop reference phase Time bias Calculate the phase deviation compensation factor Where f is the modulation carrier frequency, and the real-time correction shielding window boundary is... and After compensation, the time window covers the voltage oscillation peak of the SiC device switching edge, filters out the synchronization error caused by the difference in hardware transmission path, and maintains a shielding efficiency of over 99.8%.

[0048] The arbitration processing unit performs adaptive sensitivity control based on the current carrier phase. When entering the shielding window, the logic circuit will preset the randomness threshold N. rand The sensitivity is increased to three times the initial value while maintaining the initial sensitivity value within the sensitive time window. This phase resource allocation method enables the system to suppress systematic false triggering during commutation moments when the power device stress is high by increasing the randomness filtering requirements, while maintaining high capture sensitivity in the voltage peak phase region where the discharge probability is high, thus achieving a balance between identification accuracy and anti-interference capability in the electromagnetic environment. When verifying the physical properties of the trigger pulse, after the waveform verification module outputs the ignition identification signal from the logic AND gate, it extracts the instantaneous value sequence of the voltage signal in the pulse recovery segment, performs second-order difference operation, and calculates the second-order change of the slope of the pulse recovery segment, Dr. When the second-order change Dr satisfies Dr > 10 and shows a monotonically decaying trend, it is determined that the voltage recovery trajectory conforms to the physical characteristics of the plasma channel extinguishing process. After confirming that Dr meets the above quantitative criteria, the system issues an ignition identification signal. Using the secondary closed-loop verification of the physical evolution process, the system eliminates the symmetrical oscillation interference caused by external electromagnetic coupling from the logic level, ensuring the determinism of the loop blocking command generated by the output processing unit.

[0049] Example 5: In scenarios where a controlled power circuit is first put into service or a power module is replaced, a phase alignment calibration procedure is performed to compensate for the hardware group delay T caused by the sensor transmission cable and the analog conditioning circuit. d The system controls the output of a standard sinusoidal excitation signal from the controlled loop, and uses a phase synchronization module to capture the reference carrier phase output from the phase-locked loop. Simultaneously, the signal acquisition unit synchronously acquires the voltage signal V. t Zero-timestamp T z The arbitration processing unit calculates T. d and The time deviation between the zero-crossing moments is defined as the inherent delay parameter of the system, and then the phase deviation compensation factor in the time window division method is adjusted. To achieve time-domain alignment of physical quantities, the computational relationship is as follows: ,in, The phase deviation compensation factor is expressed in rad, f is the modulation carrier frequency expressed in Hz, and T is the phase deviation compensation factor. d The hardware group delay is expressed in seconds. This calibration procedure ensures that the shielding window covers the switching noise spikes during commutation of power devices, preventing the failure of the decision logic due to sampling phase offset.

[0050] When the system is deployed in an environment with electromagnetic noise background, the statistical unit executes the background randomness characteristic characterization procedure to determine the preset randomness threshold N. rand The reference value, wherein during the zero-load operation cycle of the controlled loop, the high-pass filter circuit continuously monitors the associated strength signal R. m The high-frequency residual components are statistically analyzed by the statistical unit, which accumulates the total number of random pulses induced by electromagnetic stray coupling over 100 consecutive sampling periods and calculates the average value of these random pulse counts. with standard deviation The system sets a preset randomness threshold N based on the noise distribution pattern. rand Set to the following values, i.e. , where N rand The preset randomness threshold is a dimensionless parameter. The average value of the random pulse count. The standard deviation of the random pulse count is used; this calibration procedure, by determining the physical judgment boundary, maintains the sensitivity to weak discharge pulses while suppressing the risk of misidentification caused by environmental fluctuations.

[0051] Example 6: In deployment scenarios where there are differences in the hardware gain of power modules, the system calibrates the preset product judgment threshold Th through a hardware gain normalization procedure. m The physical mapping relationship; the control front-end circuit receives a step reference signal with an amplitude of 1V and a rise time of 10ns. The signal acquisition unit captures the response peak value after processing by the hardware differential circuit, and the arbitration processing unit calculates the hardware gain coefficient G of the hardware path. h The arbitration processing unit, based on the hardware gain coefficient G, h and the characteristic value of the loop impedance K z Determine the preset product determination threshold Th unique to this channel.m The calculation relationship is Th m =-k×G h ×K z , among which, Th m The preset product judgment threshold is a dimensionless parameter, k is the discharge sensitivity coefficient with a value of 1200 and a unit of 1 / Ω, and G h K is the hardware gain coefficient, which is a dimensionless parameter. z The characteristic value of the loop impedance is expressed in Ω; this specification ensures that different hardware components exhibit consistent sensitivity when faced with physical breakdown signals of the same strength.

[0052] When the system faces a situation where the electromagnetic environment of the distributed load drifts, the trend analysis module performs a dynamic threshold calibration procedure for residual variance to quantify the risk of insulation degradation; during normal operation, the system captures the associated intensity signal R with a length of 2048 sampling points every 60 seconds. m The sequence, statistical unit calculates its real-time energy variance Var t And extract the mean background variance Var over the past 24 hours. avg The system limits the triggering condition of the circuit insulation degradation early warning signal to the real-time energy variance Var. t Satisfy Var t >3>Var avg The quantitative criteria utilize long-term monitoring of signal statistical characteristics to transform trend perception into logic gating based on real-time energy fluctuations for quantitative monitoring of the physical insulation state of controlled circuits. In the initial deployment of controlled circuits using SiCMOSFET power devices, a wavefront feature extraction procedure is executed to calibrate the interception window of the wavefront feature monitoring module. Under no-load conditions, the system controls the drive circuit to trigger the power device to perform 100 switching actions at rated voltage, and the wavefront feature monitoring module simultaneously extracts the associated intensity signal R. m The statistical unit calculates the fluctuation range of the sequence at the transition time Ts of the wavefront at the moment of each switching trigger. This fluctuation range, after adding a 20% redundancy margin, is defined as the inherent switching time range of the power device. The calculation relationship is as follows: Where Ts is the wavefront transition time, in ns. avg This is the statistical average of the switching edge transition times, in nanoseconds (ns).

[0053] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A sparking phenomenon identification and processing system based on pulse feature fusion, characterized in that, include: The signal acquisition unit is used to synchronously acquire the voltage and current signals of the controlled circuit; The feature preprocessing unit, connected to the signal acquisition unit, includes a hardware differential circuit for extracting the rate of change of the voltage signal and the rate of change of the current signal. The arbitration processing unit is connected to the feature preprocessing unit and includes a hardware multiplier, a high-pass filter circuit, a comparator array, and a logic AND gate. The hardware multiplier is used to multiply the rate of change of the voltage signal by the rate of change of the current signal to generate the associated intensity signal R. m ;in, dv / dt is the rate of change of the voltage signal, and di / dt is the rate of change of the current signal; The high-pass filter circuit is connected to the output of the hardware multiplier and is used to filter the associated strength signal R. m High-frequency spike signals that characterize the discontinuous jump features of the discharge process are separated in the middle. A comparator array is used to monitor the correlation strength signal R. m Whether the negative amplitude crosses the preset product judgment threshold, and whether the pulse count value of the high-frequency glitch signal within the preset time window exceeds the preset randomness threshold; A logic AND gate is connected to a comparator array for use in the correlation strength signal R. m When the preset product determination threshold is crossed and the pulse count value exceeds the preset randomness threshold, an ignition identification signal is output. The output processing unit, connected to the arbitration processing unit, is used to generate a circuit blocking command based on the ignition identification signal.

2. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, The arbitration processing unit also includes a trajectory analysis module, which maps the rate of change of the voltage signal and the rate of change of the current signal extracted by the hardware differential circuit to a two-dimensional phase plane, and calculates the instantaneous curvature of the trajectory in the two-dimensional phase plane. The arbitration processing unit is used to output an ignition identification signal when a discontinuous jump pole appears in the instantaneous curvature and the amplitude of the discontinuous jump pole deviates from the preset controlled switch envelope.

3. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, The arbitration processing unit is connected to a phase synchronization module, which is used to obtain the modulation carrier phase of the controlled loop; The arbitration processing unit is used to divide the identification period into a sensitive time window and a shielding time window according to the phase of the modulated carrier. The arbitration processing unit maintains the initial sensitivity of the preset randomness threshold within the sensitive time window, and increases the preset randomness threshold within the shielding time window.

4. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, It also includes a reference calibration module, which is used to calculate the fluctuation ratio of the voltage signal to the current signal to update the loop impedance characteristic value during the steady-state operation cycle of the controlled loop; The arbitration processing unit dynamically adjusts the preset product judgment threshold based on the loop impedance characteristic value.

5. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, It also includes a waveform verification module, connected to the arbitration processing unit, used to obtain the second-order change D of the slope of the pulse recovery segment of the voltage signal after the ignition identification signal is triggered. r ;in, v r For the voltage component of the pulse recovery segment; when the second-order change D r When the plasma nonlinear composite model is satisfied, the arbitration processing unit executes the output action of the ignition identification signal.

6. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, The hardware differential circuit in the feature preprocessing unit includes an analog differential circuit built from an operational amplifier, which is used to maintain the system’s recognition response delay within 1 μs by extracting the real-time slope of the analog signal.

7. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, The comparator array in the arbitration processing unit includes a multi-window comparator for evaluating the correlation strength signal R. m Threshold discrimination is performed based on the negative peak amplitude and the pulse density of the high-frequency glitch signal.

8. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, It also includes a wavefront feature monitoring module for extracting the correlation intensity signal R. m Triggering instantaneous wavefront transition time T S When the transition time T of the wavefront S When the power device is within the preset inherent switching time range, the arbitration processing unit shields the arcing identification signal.

9. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, It also includes a trend analysis module, connected to the arbitration processing unit, used to continuously acquire the correlation intensity signal R during periods when no ignition identification signal is output. m The low-amplitude residual sequence is obtained and the energy variance of the low-amplitude residual sequence is calculated; when the energy variance shows a non-linear growth trend, an early warning signal for insulation degradation of the output circuit is generated.

10. The spark detection and processing system based on pulse feature fusion according to claim 1, characterized in that, The output processing unit is connected to a fast circuit breaker actuator, which is used to disconnect the power input contacts of the controlled circuit within 500 ns after receiving a circuit blocking command.

Images