A load state on-line identification and self-adaptive compensation system and method of a vacuum ion coating power supply

Abstract

The application provides a load state online identification and self-adaptive compensation system and method of a vacuum ion coating power supply, comprising: a parameter acquisition module, which carries out signal sampling, signal synchronization and signal denoising to obtain a denoised electric parameter set; a load characteristic module, which carries out characteristic extraction to obtain a load characteristic set; a load state module, which uses the load characteristic set to make an initial judgment to obtain an initial load state estimation result; a correction state module, which uses the load characteristic set and the initial load state estimation result to make data-driven error correction and multi-fusion decision to obtain a final load state; a hierarchical compensation and execution module, which uses a hierarchical compensation strategy to carry out compensation control to obtain a compensation control amount, and then carries out power supply adjustment to obtain a power supply output state. The application can finely identify the full state of discharge in the vacuum coating process, realizes the graded self-adaptive regulation and control from the premonition suppression to the hard arc shutdown, and ensures the reliability and continuity of the process.

Description

Technical Field

[0001] This invention relates to the field of vacuum ion plating power supplies, and in particular to an online load state identification and adaptive compensation system and method for vacuum ion plating power supplies. Background Technology

[0002] The load in vacuum ion plating is not a fixed resistor, but plasma, which has several characteristics: strong nonlinearity, rapid changes over time, and the potential for anomalies. Therefore, the power supply does not deal with a stable load, but a dynamic, complex, and even random load system. During equipment operation, the load model or state can be deduced from signals such as voltage, current, and waveform. Once the state is identified, the power supply needs to adjust the output voltage and current, change pulse parameters, and perform rapid arc extinguishing to maintain process stability and improve coating quality.

[0003] Existing methods and technologies for load state identification and compensation include voltage and current threshold judgment, equivalent circuit model identification, waveform feature analysis, and intelligent algorithms. However, these methods and technologies have the following drawbacks: 1) Perception level: They typically only collect average current and power used in conventional control, lacking dedicated hardware channels and metrics for diagnosing minute arcs and high-frequency plasma changes, resulting in the complete loss of precursor information; 2) Modeling level: Most methods treat the load as a black box or oversimplify it to a fixed RLC model, failing to reflect the rapid changes in nonlinear conductance and capacitance of the plasma during the charging, breakdown, glow, and arc initiation stages; 3) Identification level: State determination often uses simple threshold discrimination, which is difficult to characterize the gradual transition from stability to instability, easily generating false alarms, and failing to provide a direction for maintaining the process correction. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the purpose of this invention is to provide an online load state identification and adaptive compensation system and method for a vacuum ion plating power supply, which can accurately identify the entire discharge state of the vacuum plating process and realize hierarchical adaptive control from precursor suppression to hard arc shutdown, thus ensuring the reliability and continuity of the process.

[0005] To achieve the above objectives, the present invention provides the following solution: an online load state identification and adaptive compensation system for a vacuum ion plating power supply, comprising: The parameter acquisition module is used to perform signal sampling, signal synchronization, and signal denoising in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters. The load feature module is used to extract time-domain feature set, impedance-domain feature set, frequency-domain feature set, arc precursor features, and process and control parameters based on the denoised electrical parameter set to obtain the load feature set; The load status module is used to calculate the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model using the load feature set, so as to make an initial state judgment and obtain the initial load status estimation result. The corrected state module is used to perform data-driven error correction using the load feature set and the initial load state estimation result to obtain the corrected state result and the initial load state judgment result. Then, the corrected state result and the initial load state judgment result are used to perform multi-fusion decision to obtain the final load state. The hierarchical compensation and execution module is used to perform compensation control based on the final load state using a hierarchical compensation strategy to obtain a compensation control quantity, and then use the compensation control quantity to perform power regulation to obtain the power output state after real-time compensation. The parameter acquisition module, the load characteristic module, the load status module, the correction status module, and the hierarchical compensation and execution module are interconnected.

[0006] Optionally, the parameter acquisition module includes: The signal processing unit is used to sample signals in two dimensions, the main sampling channel and the arc detection channel, using multiple sensors to obtain the original electrical parameter set; among them, the multiple sensors include voltage sensors, current sensors, analog-to-digital converters and switch status acquisition modules; The signal synchronization unit is used to calculate the correlation between voltage and current based on the original electrical parameter set and the time offset, select the one with the highest correlation as the optimal alignment time, and compensate the current or voltage according to the optimal alignment time to complete the time alignment operation and obtain the synchronization electrical parameter set. The signal denoising unit is used to perform physical constraints, adaptive median filtering, and wavelet transform based on the synchronous electrical parameter set to obtain a filtered electrical parameter set. Then, based on the filtered electrical parameter set, it calculates the filtered voltage, filtered current, filtered power, equivalent resistance, equivalent conductance, current change rate, and voltage change rate to obtain a denoised electrical parameter set.

[0007] Optionally, the load characteristic module includes: The time-domain feature unit is used to calculate the current instantaneous state, current rate of change feature, voltage rate of change feature, fluctuation amplitude feature, current root mean square and voltage root mean square based on the denoised electrical parameter set, so as to obtain the time-domain feature set; The impedance feature unit is used to calculate the impedance change rate, conductance change rate, impedance variance, and conductance variance based on the denoised electrical parameter set, thereby obtaining the impedance domain feature set. The frequency domain feature unit is used to perform spectrum analysis, frequency band energy calculation and harmonic feature extraction based on the denoised electrical parameter set to obtain the frequency domain feature set; The arc precursor unit is used to obtain arc precursor characteristics based on the denoised electrical parameter set, by weighted fusion of current abrupt change degree, voltage collapse degree, high-frequency energy, and impedance; the calculation expression of the arc precursor characteristics is: ; in, This represents the arc risk value. The degree of sudden change in current, The degree of voltage collapse, High-frequency energy, For impedance, To prevent the minimum value of division by zero, These are the weighting coefficients; The comprehensive feature unit is used to combine the time-domain feature set, the impedance-domain feature set, the frequency-domain feature set, the arc precursor features, and process and control parameters to obtain the load feature set.

[0008] Optionally, the load status module includes: The input extraction unit is used to convert the load into a time-varying nonlinear admittance model, and to use the time-varying nonlinear admittance model to decompose the measured current into normal plasma conduction current, dynamic current caused by sheath capacitance, and additional current caused by abnormal discharge. The online estimation unit is used to discretize the time-varying nonlinear admittance model and write the discretized model into a parameter estimation form to obtain a regression model. Based on the regression model, the recursive least squares method is used to estimate the parameters online to calculate the estimated values ​​of the equivalent plasma conductance, the equivalent sheath capacitance, and the arc equivalent current, thereby obtaining the estimated values ​​of the sampling point parameters. The risk calculation unit is used to calculate the equivalent plasma impedance estimate using the equivalent plasma conductivity estimate, and to weight and fuse the equivalent plasma impedance estimate, the arc equivalent current estimate, and the absolute value of the current change rate to obtain the physical model arc risk quantity. The state estimation unit is used to make an initial state judgment based on the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model, according to physical rules, and obtain the initial load state estimation result.

[0009] Optionally, the initial load state estimation result includes the preliminary load state, the estimated values ​​of the sampling point parameters, and the arc risk quantity of the physical model. The preliminary load state includes no-load or non-breakdown state, breakdown establishment state, stable glow discharge state, weak disturbance state, arc precursor state, soft arc state, hard arc state, and recovery transition state.

[0010] Optionally, the correction state module includes: A standardized input unit is used to combine the load feature set and the initial load state estimation result to construct a unified input vector, and to normalize and encode the unified input vector to obtain a standardized input vector. The correction model unit is used to select a one-dimensional convolutional neural network as a data-driven correction model, input the standardized input vector into the data-driven correction model to extract local change patterns in voltage, current, impedance and frequency domain features, output the corrected state result including state probability and parameter correction amount, and then correct the estimated parameters in the initial load state estimation result according to the corrected state result to obtain the initial load state judgment result. The final load unit is used to weightedly fuse the initial load state judgment result, the corrected state result, the initial load state estimation result, and the historical state transition result to obtain the final load state.

[0011] Optionally, the corrected state results include the probability that the load is in an unloaded or non-breakdown state, the probability that the load is in a breakdown establishment state, the probability that the load is in a stable glow discharge state, the probability that the load is in a weak disturbance state, the probability that the load is in an arc precursor state, the probability that the load is in a soft arc state, the probability that the load is in a hard arc state, the probability that the load is in a recovery transition state, the equivalent plasma conductivity correction amount, the equivalent sheath capacitance correction amount, and the arc precursor index correction amount.

[0012] Optionally, the hierarchical compensation and execution module includes: The hierarchical compensation unit is used to perform compensation control based on different load states in the final load state using a hierarchical compensation strategy, and obtain the final voltage adjustment, final current adjustment, final power adjustment, final duty cycle adjustment, final pulse frequency adjustment, final turn-off time, final reverse pulse voltage amplitude, final reverse pulse duration and final current limit value, and obtain the compensation control quantity.

[0013] The adjustment execution unit is used to map the compensation control quantity into a control command that the power supply can actually execute, and apply the control command to the power conversion circuit to adjust the output voltage, current, power and pulse parameters in real time to obtain the power supply output state after real-time compensation.

[0014] Optionally, the hierarchical compensation strategy includes: The first compensation subunit is used to determine whether the load is unloaded or not in a breakdown state. If so, the output voltage is gradually increased to the preset breakdown voltage range, the current rise rate is limited, and the power is gradually increased using short pulse or soft-start pulse. The second compensation subunit is used to determine whether the load is in a breakdown state. If so, it reduces the voltage rise slope, limits the current change rate, shortens the high voltage maintenance time, and slowly increases the duty cycle in a step manner. The third compensation subunit is used to determine whether the load is in a stable glow discharge state. If so, constant power or constant current closed-loop control is used to maintain the duty cycle and frequency stability, and to compensate for small deviations in power, current or voltage. The fourth compensation subunit is used to determine whether the load is in a weak disturbance state. If so, it reduces the output power change slope and duty cycle, adjusts the pulse frequency, increases the damping of the closed-loop controller, and improves the sensitivity of arc precursor detection. The fifth compensation subunit is used to determine whether the load is in an arcing precursor state. If so, it reduces the output power, duty cycle and current limit value, shortens the positive pulse width, and applies a short-time reverse clearing pulse. The sixth compensation subunit is used to determine whether the load is in a soft arc state. If so, it reduces the output voltage and the current limit value, pauses one or more pulse cycles, and applies a short-term reverse pulse. The seventh compensation subunit is used to determine whether the load is in a hard arc state. If so, it blocks the main power switch drive signal, shuts off the positive output pulse, releases or transfers the residual energy of the load, applies a strong reverse pulse or a short-term reverse bias, sets a long turn-off time, and continuously samples to determine whether the arc light is extinguished during the turn-off period. When restarting, it adopts a low power, low duty cycle, and low current limiting mode. The eighth compensation subunit is used to determine whether the load is in a recovery transition state. If so, a stepped power recovery is adopted, gradually increasing the duty cycle and gradually restoring the current limit value.

[0015] This invention also provides a method for online load state identification and adaptive compensation of a vacuum ion plating power supply, comprising: Signal sampling, signal synchronization, and signal denoising are performed in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters. Based on the aforementioned denoised electrical parameter set, time-domain feature set, impedance-domain feature set, frequency-domain feature set, arc precursor features, and process and control parameters are extracted to obtain the load feature set; Using the load feature set, the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model are calculated to make an initial state judgment and obtain the initial load state estimation result; Using the load feature set and the initial load state estimation result, data-driven error correction is performed to obtain the corrected state result and the initial load state judgment result. Then, the corrected state result and the initial load state judgment result are used to perform multi-fusion decision to obtain the final load state. Based on the final load state, a hierarchical compensation strategy is used for compensation control to obtain a compensation control quantity. Then, the compensation control quantity is used for power supply regulation to obtain the power output state after real-time compensation.

[0016] This invention discloses the following technical advantages by providing an online load state identification and adaptive compensation system and method for a vacuum ion plating power supply: 1. By constructing a dual-dimensional sensing architecture of the main sampling channel and the arc detection channel, precise synchronization is achieved based on the optimal alignment of voltage and current correlation. A filtering strategy with physical constraints is introduced to preserve the physical and electrical characteristics of plasma conductivity during noise reduction. It can simultaneously acquire conventional control electrical parameters and micro-arc detection characteristic parameters for the strong pulse and strong interference of ion plating, achieving high-fidelity and phase-distortion-free accurate acquisition of raw signals.

[0017] 2. Through multi-dimensional feature engineering and quantification of arc precursor characteristics, integrating current surges, voltage collapses, high-frequency energy, and floating impedance, the extremely difficult-to-capture transient precursors are mathematically and explicitly represented. This transforms a single electrical signal into a multi-dimensional profile of plasma behavior with physical meaning, and the arc precursor indicators can detect impending discharge anomalies in advance, providing a time window for predictive protection.

[0018] 3. By establishing a time-varying nonlinear admittance model, the measured current is physically decomposed into three components: plasma conduction current, sheath capacitance dynamic current, and abnormal discharge additional current. The equivalent conductance, capacitance, and arc current are estimated online using the recursive least squares method, and the arc risk quantity of the physical model is derived. This solves the problem that the pure black box model is physically disconnected from the process and difficult to interpret. It can output clear physical stages such as no-load, glow, soft arc, and hard arc, and provide parameter estimates.

[0019] 4. By combining physical model residual estimation and data-driven error correction, the generalization ability of the physical model under unknown working conditions and the identification accuracy of deep learning for complex nonlinear modes are integrated. Data is used to correct the errors caused by idealized assumptions in physical modeling, which greatly improves the accuracy and robustness of state determination.

[0020] 5. Through an 8-level differentiated layered compensation strategy, perception and identification are transformed into direct process maintenance and equipment protection, achieving long life and high reliability of high-end coating power supplies.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the system architecture provided in an embodiment of the present invention; Figure 2 This is a diagram illustrating the hierarchical compensation strategy architecture provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the method flow provided in an embodiment of the present invention. Detailed Implementation

[0024] 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 embodiments of the present invention, and not all embodiments. 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.

[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown, this invention provides an online load state identification and adaptive compensation system for a vacuum ion plating power supply, comprising: 1. Parameter Acquisition Module Signal sampling, synchronization, and denoising are performed in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters; the parameter acquisition module includes: 1.1 Operation Signal Unit Multiple sensors are used to sample signals in two dimensions: the main sampling channel and the arc detection channel, to obtain the original electrical parameter set. These sensors include voltage sensors, current sensors, analog-to-digital converters, and switch status acquisition modules. These sensors are located at the power output, power converter bridge arm, DC bus side, and process interfaces.

[0027] Original set of electrical parameters: Output voltage: The instantaneous voltage output from the power supply to the plasma load; Output current: The instantaneous current flowing into the plasma; Pulse voltage: Voltage waveform in pulse mode; Pulse current: Current waveform in pulse mode; DC bus voltage: The internal supply voltage of the power supply; Sampling period: The time interval between adjacent sampling points; Duty cycle: The percentage of pulse conduction time; Pulse frequency: power switching frequency; Output power: Power supply output power; Switching status: The state of the power device being turned on or off; Gas signals: process parameters such as gas pressure and gas flow rate.

[0028] Main channel: Used for overall status identification; High-frequency channel: used to capture arc light transients, at the microsecond level.

[0029] 1.2 Signal Synchronization Unit Based on the original electrical parameter set, the correlation between voltage and current is calculated by combining the time offset. The time with the highest correlation is selected as the optimal alignment time. Based on the optimal alignment time, the current or voltage is compensated to complete the time alignment operation and obtain the synchronous electrical parameter set.

[0030] 1.3 Signal Denoising Unit Based on the synchronous electrical parameter set, physical constraints, adaptive median filtering, and wavelet transform are performed to obtain the filtered electrical parameter set. Then, based on the filtered electrical parameter set, the filtered voltage, filtered current, filtered power, equivalent resistance, equivalent conductance, current change rate, and voltage change rate are calculated to obtain the denoised electrical parameter set.

[0031] 2. A load feature module, used to extract time-domain feature sets, impedance-domain feature sets, frequency-domain feature sets, arc precursor features, and process and control parameters based on the denoised electrical parameter set, to obtain a load feature set; the load feature module includes: 2.1 Temporal Feature Unit Based on the denoised electrical parameter set, the current instantaneous state, current rate of change characteristics, voltage rate of change characteristics, fluctuation amplitude characteristics, current root mean square and voltage root mean square are calculated to obtain the time-domain feature set.

[0032] Fluctuation amplitude characteristics: characterizes discharge stability, with fluctuations increasing significantly during arcing.

[0033] Root mean square current: describes the energy level and is used to distinguish between stable discharge and abnormal conditions.

[0034] 2.2 Impedance Characteristic Unit Based on the denoised electrical parameter set, the impedance change rate, conductance change rate, impedance variance, and conductance variance are calculated to obtain the impedance domain feature set.

[0035] 2.3 Frequency Domain Feature Units Based on the denoised electrical parameter set, spectrum analysis, frequency band energy calculation, and harmonic feature extraction are performed to obtain a frequency domain feature set.

[0036] 2.4 Arc Light Index Unit Based on the aforementioned denoised electrical parameter set, the arc precursor characteristics are obtained by weighted fusion of current abrupt change degree, voltage collapse degree, high-frequency energy, and impedance; the calculation expression for the arc precursor characteristics is as follows: ; in, This represents the arc risk value. The degree of sudden change in current, The degree of voltage collapse, High-frequency energy, For impedance, To prevent the minimum value of division by zero, These are the weighting coefficients.

[0037] 2.5 Comprehensive Feature Unit By combining the time-domain feature set, the impedance-domain feature set, the frequency-domain feature set, the arc precursor features, and the process and control parameters, a load feature set is obtained. The process and control parameters include power, duty cycle, frequency, and air pressure.

[0038] 3. A load state module, used to calculate the estimated values ​​of sampling point parameters and the arc risk quantity of the physical model using the load feature set, in order to make an initial state judgment and obtain the initial load state estimation result; the load state module includes: 3.1 Input Extraction Unit In the vacuum ion plating process, the load faced by the power supply is not a simple fixed resistor, but a dynamic load formed by the plasma, the target surface, the sheath capacitance, the gas discharge channel, and the local arc channel. Therefore, the load needs to be equivalent to a time-varying nonlinear admittance model.

[0039] The load is equivalent to a time-varying nonlinear admittance model. The measured current is decomposed into normal plasma conduction current, dynamic current caused by sheath capacitance, and additional current caused by abnormal discharge using the time-varying nonlinear admittance model. This can avoid judging arcing based solely on the current magnitude and improve the reliability of state identification.

[0040] 3.2 Online Estimation Unit Since the controller actually processes discrete sampled data, the time-varying nonlinear admittance model is discretized, and the discretized model is written in the form of parameter estimation to obtain a regression model. Based on the regression model, the parameters are estimated online using the recursive least squares method to calculate the estimated values ​​of equivalent plasma conductance, equivalent sheath capacitance, and arc equivalent current, thereby obtaining the estimated values ​​of the sampling point parameters.

[0041] 3.3 Risk Calculation Unit The equivalent plasma impedance estimate is calculated using the equivalent plasma conductivity estimate, and the equivalent plasma impedance estimate, the arc equivalent current estimate, and the absolute value of the current change rate are weighted and fused to obtain the physical model arc risk quantity.

[0042] 3.4 State Estimation Unit Based on the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model, the initial state is determined according to the physical rules to obtain the initial load state estimation result.

[0043] The initial load state estimation results include the preliminary load state, the estimated values ​​of the sampling point parameters, and the arc risk quantity of the physical model. The preliminary load state includes no-load or non-breakdown state, breakdown establishment state, stable glow discharge state, weak disturbance state, arc precursor state, soft arc state, hard arc state, and recovery transition state.

[0044] 4. A state correction module, used to perform data-driven error correction using the load feature set and the initial load state estimation result, to obtain a corrected state result and a preliminary load state judgment result, and then to perform multi-fusion decision using the corrected state result and the preliminary load state judgment result to obtain the final load state; the state correction module includes: 4.1 Standardized Input Unit By combining the load feature set and the initial load state estimation results, a unified input vector is constructed. The unified input vector is then normalized and encoded to obtain a standardized input vector.

[0045] Since the input vector contains data with different dimensions such as voltage, current, power, impedance, frequency, probability, and state number, directly inputting the data to drive the model can easily lead to features with larger dimensions dominating the model output. Therefore, normalization is performed before inputting the data into the model.

[0046] 4.2 Modified Model Unit A one-dimensional convolutional neural network is selected as the data-driven correction model. The standardized input vector is input into the data-driven correction model to extract local change patterns in voltage, current, impedance and frequency domain features. The corrected state result, including state probability and parameter correction amount, is output. Then, based on the corrected state result, the estimated parameters in the initial load state estimation result are corrected to obtain the initial load state judgment result.

[0047] 4.3 Final Load Unit The final load state is obtained by weighted fusion of the initial load state judgment result, the corrected state result, the initial load state estimation result, and the historical state transition result.

[0048] The corrected state results include the probability that the load is in an unloaded or non-breakdown state, the probability that the load is in a breakdown establishment state, the probability that the load is in a stable glow discharge state, the probability that the load is in a weak disturbance state, the probability that the load is in an arc precursor state, the probability that the load is in a soft arc state, the probability that the load is in a hard arc state, the probability that the load is in a recovery transition state, the equivalent plasma conductivity correction amount, the equivalent sheath capacitance correction amount, and the arc precursor index correction amount.

[0049] 5. A hierarchical compensation and execution module, used to perform compensation control based on the final load state using a hierarchical compensation strategy, obtain a compensation control quantity, and then use the compensation control quantity to adjust the power supply to obtain the real-time compensated power output state; the hierarchical compensation and execution module includes: 5.1 Hierarchical Compensation Unit Based on the different load states in the final load state, a hierarchical compensation strategy is used for compensation control to obtain the final voltage adjustment, final current adjustment, final power adjustment, final duty cycle adjustment, final pulse frequency adjustment, final turn-off time, final reverse pulse voltage amplitude, final reverse pulse duration, and final current limit value, thus obtaining the compensation control quantity.

[0050] like Figure 2 As shown, the hierarchical compensation strategy includes: 5.1.1 First Compensation Subunit Determine if the load is unloaded or not in a breakdown state. If so, then: 1) Gradually increase the output voltage to the preset breakdown voltage range; 2) Limit the rate of current rise to avoid excessive inrush current at the moment of breakdown; 3) Gradually increase power using short pulses or soft-start pulses; 4) Maintain a low current limiting value to prevent excessive current during abnormal short circuits; 5) If the breakdown does not occur after several consecutive cycles, slightly increase the pulse duty cycle or pulse frequency.

[0051] 5.1.2 Second Compensation Subunit Determine if the load is in a breakdown-establishment state. If so, then: 1) Reduce the voltage rise slope; 2) Limit the rate of change of current; 3) Shorten the high-voltage maintenance time; 4) Gradually increase the duty cycle in a step-like manner; 5) Monitor whether the corrected arc precursor index rises rapidly. If it rises too quickly, initiate the arc precursor compensation strategy ahead of schedule.

[0052] 5.1.3 Third Compensation Subunit Determine if the load is in a stable glow discharge state. If so, then: 1) Use constant power or constant current closed-loop control; 2) Maintain a basically stable duty cycle and frequency; 3) Only minor compensation is provided for small deviations in power, current, or voltage; 4) Continue monitoring of the corrected arc precursor index, but do not implement strong compensation; 5) If the load fluctuation increases, switch to the weak disturbance compensation strategy.

[0053] 5.1.4 Fourth Compensation Subunit Determine if the load is in a weak disturbance state. If so, then: 1) Reduce the slope of the output power change; 2) Slightly reduce the duty cycle; 3) Adjust the pulse frequency appropriately to prevent the system from oscillating in frequency bands; 4) Increase the damping of the closed-loop controller; 5) Improve the sensitivity of arc precursor detection; 6) If the corrected arc precursor index continues to rise, then the arc precursor compensation strategy will be activated.

[0054] 5.1.5 Fifth Compensation Subunit Determine if the load is in an arcing precursor state. If so, then: 1) Reduce output power within one or more control cycles; 2) Shorten the positive pulse width; 3) Reduce the duty cycle; 4) Reduce the current limit value; 5) Apply a short-duration reverse clearance pulse to clear local charge accumulation on the target surface; 6) Improve the sensitivity of the hard arc fast protection channel; 7) If the corrected arc precursor index decreases after compensation, then enter a stable or weak disturbance strategy; 8) If the current continues to increase suddenly after compensation, immediately switch to a soft arc or hard arc strategy.

[0055] 5.1.6 Sixth Compensation Subunit Determine if the load is in a soft arc state; if so, then: 1) Quickly reduce output voltage or output power; 2) Pause one or more pulse cycles; 3) Apply a short-duration reverse pulse; 4) Reduce the current rate limit; 5) After the voltage and current return to normal, restore the power at a smaller slope; 6) During the recovery process, continuously monitor the compensated and corrected arc precursor indicators to prevent secondary arcing.

[0056] 5.1.7 Seventh Compensation Subunit Determine if the load is in a hard arc state; if so, then: 1) Immediately block the main power switch drive signal; 2) Turn off the positive output pulse; 3) Release or transfer residual energy from the load; 4) Apply a strong reverse pulse or a short-term reverse bias; 5) Set a longer shutdown time; 6) During the shutdown period, continuous sampling is performed to determine whether the arc light has been extinguished; 7) Low power, low duty cycle, and low current limiting are used during restart; 8) If hard arcs occur continuously within a short period of time, the system will enter the lockout protection or manual inspection mode.

[0057] 5.1.8 Eighth Compensation Subunit Determine if the load is in a recovery transition state; if so, then: 1) Employ a stepped power recovery method, rather than restoring the power to the target level all at once; 2) The duty cycle gradually increases; 3) The current limit value gradually recovers; 4) Continuously monitor the corrected arc precursor indicators, voltage recovery speed, and current fluctuations; 5) If the risk of arcing increases again during the recovery process, stop the recovery and return to the arcing precursor state strategy, soft arc state, or hard arc state; 6) If the system is stable for several consecutive cycles, then return to the stable glow discharge strategy.

[0058] 5.2 Adjustment Execution Unit The compensation control quantity is mapped to the actual executable control command of the power supply. The control command is applied to the power conversion circuit to adjust the output voltage, current, power and pulse parameters in real time, so as to obtain the power supply output state after real-time compensation.

[0059] like Figure 3 As shown, the present invention also provides a method for online identification and adaptive compensation of the load state of a vacuum ion plating power supply, comprising: Signal sampling, signal synchronization, and signal denoising are performed in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters. Based on the aforementioned denoised electrical parameter set, time-domain feature set, impedance-domain feature set, frequency-domain feature set, arc precursor features, and process and control parameters are extracted to obtain the load feature set; Using the load feature set, the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model are calculated to make an initial state judgment and obtain the initial load state estimation result; Using the load feature set and the initial load state estimation result, data-driven error correction is performed to obtain the corrected state result and the initial load state judgment result. Then, the corrected state result and the initial load state judgment result are used to perform multi-fusion decision to obtain the final load state. Based on the final load state, a hierarchical compensation strategy is used for compensation control to obtain a compensation control quantity. Then, the compensation control quantity is used for power supply regulation to obtain the power output state after real-time compensation.

[0060] Therefore, by providing an online load state identification and adaptive compensation system and method for a vacuum ion plating power supply, the present invention can achieve: The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0061] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A load state online identification and adaptive compensation system for a vacuum ion plating power supply, characterized in that, include: The parameter acquisition module is used to perform signal sampling, signal synchronization, and signal denoising in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters. The load feature module is used to extract time-domain feature set, impedance-domain feature set, frequency-domain feature set, arc precursor features, and process and control parameters based on the denoised electrical parameter set to obtain the load feature set; The load status module is used to calculate the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model using the load feature set, so as to make an initial state judgment and obtain the initial load status estimation result. The corrected state module is used to perform data-driven error correction using the load feature set and the initial load state estimation result to obtain the corrected state result and the initial load state judgment result. Then, the corrected state result and the initial load state judgment result are used to perform multi-fusion decision to obtain the final load state. The hierarchical compensation and execution module is used to perform compensation control based on the final load state using a hierarchical compensation strategy to obtain a compensation control quantity, and then use the compensation control quantity to perform power regulation to obtain the power output state after real-time compensation. The parameter acquisition module, the load characteristic module, the load status module, the correction status module, and the hierarchical compensation and execution module are interconnected.

2. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 1, characterized in that, The parameter acquisition module includes: The signal processing unit is used to sample signals in two dimensions, the main sampling channel and the arc detection channel, using multiple sensors to obtain the original electrical parameter set; among them, the multiple sensors include voltage sensors, current sensors, analog-to-digital converters and switch status acquisition modules; The signal synchronization unit is used to calculate the correlation between voltage and current based on the original electrical parameter set and the time offset, select the one with the highest correlation as the optimal alignment time, and compensate the current or voltage according to the optimal alignment time to complete the time alignment operation and obtain the synchronization electrical parameter set. The signal denoising unit is used to perform physical constraints, adaptive median filtering, and wavelet transform based on the synchronous electrical parameter set to obtain a filtered electrical parameter set. Then, based on the filtered electrical parameter set, it calculates the filtered voltage, filtered current, filtered power, equivalent resistance, equivalent conductance, current change rate, and voltage change rate to obtain a denoised electrical parameter set.

3. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 2, characterized in that, The load characteristic module includes: The time-domain feature unit is used to calculate the current instantaneous state, current rate of change feature, voltage rate of change feature, fluctuation amplitude feature, current root mean square and voltage root mean square based on the denoised electrical parameter set, so as to obtain the time-domain feature set; The impedance feature unit is used to calculate the impedance change rate, conductance change rate, impedance variance, and conductance variance based on the denoised electrical parameter set, thereby obtaining the impedance domain feature set. The frequency domain feature unit is used to perform spectrum analysis, frequency band energy calculation and harmonic feature extraction based on the denoised electrical parameter set to obtain the frequency domain feature set; The arc precursor unit is used to obtain arc precursor characteristics based on the denoised electrical parameter set, by weighted fusion of current abrupt change degree, voltage collapse degree, high-frequency energy, and impedance; the calculation expression of the arc precursor characteristics is: ； in, This represents the arc risk value. The degree of sudden change in current, The degree of voltage collapse, High-frequency energy, For impedance, To prevent the minimum value of division by zero, These are the weighting coefficients; The comprehensive feature unit is used to combine the time-domain feature set, the impedance-domain feature set, the frequency-domain feature set, the arc precursor features, and process and control parameters to obtain the load feature set.

4. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 3, characterized in that, The load status module includes: The input extraction unit is used to convert the load into a time-varying nonlinear admittance model, and to use the time-varying nonlinear admittance model to decompose the measured current into normal plasma conduction current, dynamic current caused by sheath capacitance, and additional current caused by abnormal discharge. The online estimation unit is used to discretize the time-varying nonlinear admittance model and write the discretized model into a parameter estimation form to obtain a regression model. Based on the regression model, the recursive least squares method is used to estimate the parameters online to calculate the estimated values ​​of the equivalent plasma conductance, the equivalent sheath capacitance, and the arc equivalent current, thereby obtaining the estimated values ​​of the sampling point parameters. The risk calculation unit is used to calculate the equivalent plasma impedance estimate using the equivalent plasma conductivity estimate, and to weight and fuse the equivalent plasma impedance estimate, the arc equivalent current estimate, and the absolute value of the current change rate to obtain the physical model arc risk quantity. The state estimation unit is used to make an initial state judgment based on the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model, according to physical rules, and obtain the initial load state estimation result.

5. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 4, characterized in that, The initial load state estimation results include the preliminary load state, the estimated values ​​of the sampling point parameters, and the arc risk quantity of the physical model. The preliminary load state includes no-load or non-breakdown state, breakdown establishment state, stable glow discharge state, weak disturbance state, arc precursor state, soft arc state, hard arc state, and recovery transition state.

6. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 5, characterized in that, The corrected state module includes: A standardized input unit is used to combine the load feature set and the initial load state estimation result to construct a unified input vector, and to normalize and encode the unified input vector to obtain a standardized input vector. The correction model unit is used to select a one-dimensional convolutional neural network as a data-driven correction model, input the standardized input vector into the data-driven correction model to extract local change patterns in voltage, current, impedance and frequency domain features, output the corrected state result including state probability and parameter correction amount, and then correct the estimated parameters in the initial load state estimation result according to the corrected state result to obtain the initial load state judgment result. The final load unit is used to weightedly fuse the initial load state judgment result, the corrected state result, the initial load state estimation result, and the historical state transition result to obtain the final load state.

7. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 6, characterized in that, The corrected state results include the probability that the load is in an unloaded or non-breakdown state, the probability that the load is in a breakdown establishment state, the probability that the load is in a stable glow discharge state, the probability that the load is in a weak disturbance state, the probability that the load is in an arc precursor state, the probability that the load is in a soft arc state, the probability that the load is in a hard arc state, the probability that the load is in a recovery transition state, the equivalent plasma conductivity correction amount, the equivalent sheath capacitance correction amount, and the arc precursor index correction amount.

8. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 7, characterized in that, The hierarchical compensation and execution module includes: The hierarchical compensation unit is used to perform compensation control based on different load states in the final load state using a hierarchical compensation strategy, and obtain the final voltage adjustment, final current adjustment, final power adjustment, final duty cycle adjustment, final pulse frequency adjustment, final turn-off time, final reverse pulse voltage amplitude, final reverse pulse duration and final current limit value, and obtain the compensation control quantity. The adjustment execution unit is used to map the compensation control quantity into a control command that the power supply can actually execute, and apply the control command to the power conversion circuit to adjust the output voltage, current, power and pulse parameters in real time to obtain the power supply output state after real-time compensation.

9. The online load state identification and adaptive compensation system for a vacuum ion plating power supply according to claim 8, characterized in that, The hierarchical compensation strategy includes: The first compensation subunit is used to determine whether the load is unloaded or not in a breakdown state. If so, the output voltage is gradually increased to the preset breakdown voltage range, the current rise rate is limited, and the power is gradually increased using short pulse or soft-start pulse. The second compensation subunit is used to determine whether the load is in a breakdown state. If so, it reduces the voltage rise slope, limits the current change rate, shortens the high voltage maintenance time, and slowly increases the duty cycle in a step manner. The third compensation subunit is used to determine whether the load is in a stable glow discharge state. If so, constant power or constant current closed-loop control is used to maintain the duty cycle and frequency stability, and to compensate for small deviations in power, current or voltage. The fourth compensation subunit is used to determine whether the load is in a weak disturbance state. If so, it reduces the output power change slope and duty cycle, adjusts the pulse frequency, increases the damping of the closed-loop controller, and improves the sensitivity of arc precursor detection. The fifth compensation subunit is used to determine whether the load is in an arcing precursor state. If so, it reduces the output power, duty cycle and current limit value, shortens the positive pulse width, and applies a short-time reverse clearing pulse. The sixth compensation subunit is used to determine whether the load is in a soft arc state. If so, it reduces the output voltage and the current limit value, pauses one or more pulse cycles, and applies a short-term reverse pulse. The seventh compensation subunit is used to determine whether the load is in a hard arc state. If so, it blocks the main power switch drive signal, shuts off the positive output pulse, releases or transfers the residual energy of the load, applies a strong reverse pulse or a short-term reverse bias, sets a long turn-off time, and continuously samples to determine whether the arc light is extinguished during the turn-off period. When restarting, it adopts a low power, low duty cycle, and low current limiting mode. The eighth compensation subunit is used to determine whether the load is in a recovery transition state. If so, a stepped power recovery is adopted, gradually increasing the duty cycle and gradually restoring the current limit value.

10. A method for online identification and adaptive compensation of the load state of a vacuum ion plating power supply, characterized in that, include: Signal sampling, signal synchronization, and signal denoising are performed in two dimensions: the main sampling channel and the arc detection channel, to obtain a set of denoised electrical parameters. Based on the aforementioned denoised electrical parameter set, time-domain feature set, impedance-domain feature set, frequency-domain feature set, arc precursor features, and process and control parameters are extracted to obtain the load feature set; Using the load feature set, the estimated values ​​of the sampling point parameters and the arc risk quantity of the physical model are calculated to make an initial state judgment and obtain the initial load state estimation result; Using the load feature set and the initial load state estimation result, data-driven error correction is performed to obtain the corrected state result and the initial load state judgment result. Then, the corrected state result and the initial load state judgment result are used to perform multi-fusion decision to obtain the final load state. Based on the final load state, a hierarchical compensation strategy is used for compensation control to obtain a compensation control quantity. Then, the compensation control quantity is used for power supply regulation to obtain the power output state after real-time compensation.

Images