Metal film deposition process control system based on PLC

Abstract

The invention relates to the technical field of physical vapor deposition and industrial automation control, in particular to a metal film deposition process control system based on a PLC (programmable logic controller), comprising an active disturbance injection module for generating a periodic micro-disturbance signal with a preset frequency and a preset amplitude, generating a composite driving instruction, and sending the composite driving instruction to the PLC; the composite driving instruction is sent to an execution mechanism of the deposition process; the dynamic response acquisition module is used for constructing a time sequence data stream containing process dynamic response; the impedance fingerprint demodulation module is used for resolving a local sensitivity index of the current process; the state mapping and closed-loop regulation module is used for generating a correction instruction aiming at the reference flow control signal so as to lock a working interval of a deposition process; according to the invention, production interruption or film layer quality accidents caused by excessive reaction are effectively avoided, and the operation safety of the system in a high deposition rate critical state is improved.

Description

Technical Field

[0001] This invention relates to the fields of physical vapor deposition and industrial automation control technology, specifically a PLC-based metal film deposition process control system. Background Technology

[0002] In the current metal film deposition environment, physical vapor deposition equipment such as magnetron sputtering needs to maintain a specific reaction state during long-term operation to ensure the macroscopic uniformity, physical properties and stoichiometry of the deposited film.

[0003] To control the deposition process, existing solutions generally employ a passive monitoring architecture, relying on static voltage and current signals or expensive external optical emission spectrometers and plasma emission monitors to acquire process status, and then adjusting gas flow through communication between the host computer and the controller. Although this approach has a certain adjustment capability under steady-state conditions, significant nonlinear hysteresis and physical transmission delays exist within the reaction chamber, and passive monitoring methods lack sensitivity in critical transition zones. Furthermore, communication delays and timing jitter between discrete hardware components prevent the control system from timely compensating for characteristic drift caused by target etching or environmental disturbances, making it difficult to maintain stable operating points at high deposition rates. Therefore, eliminating dependence on expensive external sensors and overcoming transmission hysteresis, improving the system's sensitivity to minute state changes and real-time response, and thus achieving precise locking of the operating point in unstable transition zones, has become a pressing technical problem to be solved. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides a PLC-based metal film deposition process control system. Specifically, the technical solution of the present invention includes:

[0005] Active perturbation injection module: used to generate periodic micro-perturbation signals with preset frequency and preset amplitude, superimpose the periodic micro-perturbation signals onto the reference flow control signal to generate composite drive commands, and send the composite drive commands to the actuator of the deposition process;

[0006] Dynamic response acquisition module: used to acquire electrical parameter signals output by the deposition power supply in real time and construct a time-series data stream containing the dynamic response of the process;

[0007] Impedance fingerprint demodulation module: used to perform frequency domain analysis on the time-series data stream, extract the response component in the electrical parameter signal that has the same frequency as the periodic micro-perturbation signal, and calculate the local sensitivity index of the current process based on the ratio relationship between the response component and the periodic micro-perturbation signal.

[0008] State mapping and closed-loop adjustment module: used to map the local sensitivity index to the working point state on the nonlinear hysteresis curve, calculate the deviation between the local sensitivity index and the preset target sensitivity threshold, and generate a correction command for the reference flow control signal based on the deviation, so as to lock the working range of the deposition process.

[0009] Optionally, the modules can be connected using the following method:

[0010] S1. The PLC controller generates the periodic micro-disturbance signal and linearly superimposes it with the reference flow control signal at the current moment, and outputs the composite drive command to the mass flow controller.

[0011] S2. Synchronously acquire the discharge voltage and discharge current data of the deposition power supply through the analog input interface to form the electrical parameter signal;

[0012] S3. Using digital signal processing algorithms, the dynamic response amplitude and phase information excited by the periodic micro-perturbation signal are separated from the electrical parameter signal, and the local sensitivity index characterizing the process gain is calculated.

[0013] S4. Input the local sensitivity index as a feedback variable to the sliding mode controller, and dynamically adjust the value of the reference flow control signal according to the deviation direction of the local sensitivity index relative to the target sensitivity threshold, so that the local sensitivity index is maintained within the range limited by the target sensitivity threshold.

[0014] Optionally, S1 specifically includes:

[0015] S11. A disturbance waveform table is preset in the memory of the PLC controller. The disturbance waveform table stores data points of a sine wave sequence or a pseudo-random binary sequence.

[0016] S12. Set the disturbance scanning period, read the data points in the disturbance waveform table sequentially according to the disturbance scanning period, and multiply the read data points by the preset disturbance intensity coefficient to generate the periodic micro-disturbance signal;

[0017] S13. Obtain the reference flow control signal of the previous control cycle, add the periodic micro-perturbation signal to the reference flow control signal, and generate the composite drive command, wherein the update frequency of the composite drive command is higher than the thermodynamic response frequency of the deposition process.

[0018] Optionally, S3 specifically includes:

[0019] S31. Establish a sliding time window for the collected electrical parameter signals and extract the voltage timing data within the sliding time window;

[0020] S32. Perform discrete Fourier transform or digital phase-locked amplification on the voltage time-series data, and extract the spectral amplitude at the preset frequency, which is recorded as the response amplitude.

[0021] S33. Obtain the input amplitude of the periodic micro-perturbation signal at the preset frequency, calculate the ratio of the response amplitude to the input amplitude, and define the ratio as the local sensitivity index, wherein the local sensitivity index characterizes the transient response rate of the deposition voltage to changes in gas flow rate.

[0022] Optionally, S3 also includes a delay compensation step:

[0023] S34. Calculate the cross-correlation function between the periodic micro-perturbation signal and the electrical parameter signal;

[0024] S35. Search for the peak position of the cross-correlation function, determine the time lag corresponding to the peak position, and mark the time lag as the gas transmission delay time.

[0025] S36. Based on the gas transmission delay time, perform time alignment correction on the electrical parameter signal, and perform the amplitude extraction operation in S32 based on the corrected data to eliminate the influence of transmission delay on sensitivity calculation.

[0026] Optionally, S4 specifically includes:

[0027] S41. A hysteresis loop characteristic model of the deposition process is preset, in which the metal mode region, transition mode region and compound mode region are defined.

[0028] S42. Identify the region to which the current working point belongs based on the value of the local sensitivity index: if the local sensitivity index is lower than a preset low sensitivity threshold, determine that the current region is in the metal mode region or the compound mode region; if the local sensitivity index is higher than a preset high sensitivity threshold, determine that the current region is in the transition mode region.

[0029] S43. Set the high sensitivity threshold as the target sensitivity threshold, calculate the difference between the current local sensitivity index and the target sensitivity threshold, and generate a control error signal.

[0030] Optionally, S4 also includes a tuning step based on the sliding mode control law:

[0031] S44. Define a switching function and construct a sliding surface based on the control error signal and its rate of change;

[0032] S45. Based on the sign of the sliding surface, select the sign direction of the control gain and generate a discontinuous control correction value:

[0033] When the system state is on one side of the sliding surface, a positive flow correction step size is output, and the reference flow control signal is increased;

[0034] When the system state is on the other side of the sliding surface, a negative flow correction step size is output to reduce the reference flow control signal;

[0035] S46. Perform integral smoothing on the discontinuous control correction amount, update the reference flow control signal at the next moment, and drive the local sensitivity index to converge toward the high sensitivity threshold.

[0036] Optionally, the system also includes an abnormal drift blocking mechanism, which is implemented as follows:

[0037] Monitor the time change rate of the local sensitivity index; if the time change rate exceeds the preset avalanche threshold and the absolute value of the electrical parameter signal shows a monotonically decreasing trend, it is determined that the system is about to enter the target poisoning state.

[0038] The PLC controller immediately triggers a blocking command, forcibly resetting the reference flow control signal to a preset safe backoff value, and suspending the injection of the periodic micro-disturbance signal until the electrical parameter signal returns to the preset initial state range.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. This invention constructs an active perturbation injection and impedance fingerprint demodulation mechanism, transforming plasma into a dynamic system with self-sensing capabilities; by extracting the response component with the same frequency as the micro-perturbation to calculate the local sensitivity, it overcomes the problem of insufficient sensitivity in the transition region of traditional passive monitoring; this design does not rely on expensive optical sensors, but can accurately sense the process status within millisecond cycles using only electrical signals, significantly reducing hardware costs.

[0041] 2. This invention adopts a full-link closed-loop architecture based on a PLC real-time operating system, realizing tightly coupled execution from signal generation and acquisition to solution control; this integrated design eliminates the communication delay and timing jitter caused by the traditional separate architecture of host computer and controller, ensures strict synchronization between composite drive commands and thermodynamic response of deposition process, and greatly improves the response speed and determinism of control system to transient changes.

[0042] 3. This invention combines gas transport delay compensation and sliding mode control strategy to effectively solve the problems of nonlinear hysteresis and large time delay in physical vapor deposition; by correcting time alignment through cross-correlation function and using sliding mode variable structure characteristics to deal with model uncertainty, the system can still maintain strong locking and robust control of the extremely unstable transition zone operating point when the target material is etched and consumed, causing characteristic drift or environmental disturbance.

[0043] 4. This invention establishes an abnormal drift blocking mechanism as a safety defense. By monitoring the time change rate of local sensitivity and the monotonically decreasing trend of electrical parameter signals, it can accurately predict the avalanche characteristics before target poisoning. When it is determined that instability is imminent, the flow rate is forcibly reset to a safe value, which effectively avoids production interruption or film quality accidents caused by overreaction and improves the operational safety of the system under high deposition rate critical conditions. Attached Figure Description

[0044] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0045] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0047] Example 1:

[0048] Please see Figure 1 A PLC-based metal film deposition process control system includes:

[0049] Active perturbation injection module: used to generate periodic micro-perturbation signals with preset frequency and preset amplitude, superimpose the periodic micro-perturbation signals onto the reference flow control signal to generate composite drive commands, and send the composite drive commands to the actuators of the deposition process;

[0050] Dynamic response acquisition module: used to acquire electrical parameter signals output by the deposition power supply in real time and construct a time-series data stream containing the dynamic response of the process;

[0051] Impedance fingerprint demodulation module: used to perform frequency domain analysis on time-series data streams, extract response components in electrical parameter signals that have the same frequency as periodic micro-perturbation signals, and calculate the local sensitivity index of the current process based on the ratio relationship between the response components and periodic micro-perturbation signals.

[0052] State mapping and closed-loop adjustment module: used to map the local sensitivity index to the working point state on the nonlinear hysteresis curve, calculate the deviation between the local sensitivity index and the preset target sensitivity threshold, and generate correction instructions for the reference flow control signal based on the deviation to lock the working range of the deposition process.

[0053] This embodiment provides a PLC-based metal film deposition process control system; the system aims to solve the problem of process operating point drift caused by reaction hysteresis and sensor delay in physical vapor deposition processes such as magnetron sputtering.

[0054] The core of this embodiment lies in constructing an active detection-instant response closed-loop control architecture, abandoning the traditional passive mode that relies solely on static voltage or expensive optical emission spectrum monitoring; the system mainly includes the following modules:

[0055] The active perturbation injection module aims to transform the plasma environment of the deposition process into a detectable dynamic system.

[0056] In this embodiment, the module does not output a constant control quantity like a traditional PID controller, but instead generates a control quantity with a preset frequency. and preset amplitude Periodic micro-perturbation signal This module will Linearly superimposed onto the reference flow control signal to maintain the basic process state. Above, generate composite driver instructions:

[0057] Here, the periodic micro-perturbation signal refers to an artificially introduced excitation waveform with a very small amplitude, usually 1%-5% of the reference signal, and a fixed frequency. Its function is to excite the dynamic characteristics of the system without destroying the macroscopic uniformity of the film.

[0058] This module sends composite drive commands to the actuators in the deposition process, such as the mass flow controller (MFC), through the high-speed analog output port of the PLC.

[0059] The dynamic response acquisition module aims to capture the system's transient feedback to the aforementioned micro-disturbances. In this embodiment, the module utilizes the PLC's high-speed analog input interface to acquire in real-time the electrical parameter signals output by the deposition power supply, primarily the target voltage. or discharge current This module discretizes the data at a sampling rate higher than the Nyquist frequency to construct a time-series data stream containing the dynamic response of the process.

[0060] The purpose of the impedance fingerprint demodulation module is to extract characteristic values ​​that characterize the system state from noisy electrical parameter signals.

[0061] In this embodiment, the module performs frequency domain analysis on the time-series data stream, such as Fast Fourier Transform (FFT) or digital phase-locked loop amplification, to extract the electrical parameter signal with the same frequency as the periodic micro-perturbation signal. The response components;

[0062] Based on the amplitude of the response component and the amplitude of the input disturbance The ratio relationship is used to calculate the local sensitivity index. ;

[0063] Local sensitivity index This refers to the approximate value of the partial derivative of the deposition voltage with respect to the change in gas flow rate at the current operating point. Its value directly reflects the position and slope of the current process point on the hysteresis curve;

[0064] The state mapping and closed-loop regulation module aims to achieve model-free adaptive control.

[0065] In this embodiment, the module will solve the Mapped to the operating point state on a nonlinear hysteresis curve; this module calculates Compared with the preset target sensitivity threshold The deviation between them is used to generate a reference flow control signal. Correction instructions;

[0066] Through the coordinated operation of the above modules, this embodiment effectively transforms the plasma itself into a sensor. By utilizing actively injected micro-perturbations and demodulation analysis, the system can accurately detect whether the deposition process is at the edge of an unstable transition region within a millisecond-level PLC scan cycle. This method eliminates the dependence on expensive external sensors such as PEM and OES, and by directly using electrical signals propagating at the speed of light for feedback, it significantly reduces the control lag caused by traditional gas transmission, achieving precise locking of the working range of the metal film deposition process.

[0067] This embodiment constructs an active control architecture that integrates excitation and detection. Its core innovation lies in using an active perturbation injection module to transform the static deposition environment into a dynamic and observable system, which solves the problem of insufficient sensitivity of traditional passive monitoring in the transition zone.

[0068] By extracting the response component with the same frequency as the micro-perturbation through the impedance fingerprint demodulation module, the system can accurately extract the process gain information from the high-noise plasma power supply signal. This process gain directly maps the tangent slope of the current operating point on the hysteresis curve. This control strategy based on local sensitivity index allows the system to lock the extremely unstable transition region operating point with only millisecond-level electrical response without relying on the hysteresis optical signal. This ensures a high deposition rate while avoiding target poisoning or film stoichiometry drift.

[0069] Example 2:

[0070] The modules are interconnected using the following method:

[0071] S1, the PLC controller generates periodic micro-disturbance signals and linearly superimposes them with the current reference flow control signal, and outputs composite drive commands to the mass flow controller;

[0072] S2. Synchronously acquire discharge voltage and discharge current data of the deposition power supply through analog input interface to form electrical parameter signals;

[0073] S3. Using digital signal processing algorithms, the amplitude and phase information of the dynamic response excited by the periodic micro-perturbation signal are separated from the electrical parameter signal, and the local sensitivity index characterizing the process gain is calculated.

[0074] S4. Input the local sensitivity index as a feedback variable to the sliding mode controller. Based on the deviation direction of the local sensitivity index from the target sensitivity threshold, dynamically adjust the value of the reference flow control signal to keep the local sensitivity index within the range limited by the target sensitivity threshold.

[0075] This embodiment further illustrates the specific methods and steps for achieving coordinated control between modules. This method strictly runs in the real-time operating system of the PLC to ensure the timing determinism of the control.

[0076] S1. Composite Instruction Generation Steps: Within each scan cycle, the PLC controller generates periodic micro-perturbation signals using its internal algorithm. ;

[0077] The PLC reads the reference flow control signal calculated by the control algorithm at the current moment. Perform a linear superposition operation:

[0078]

[0079] Through the PLC's D / A conversion module, It is converted into an analog voltage signal of 4-20mA or 0-10V and output to the mass flow controller MFC;

[0080] S2. Synchronous Acquisition Step: The system synchronously acquires the discharge voltage data of the deposition power supply through a high-precision analog input module with an A / D conversion accuracy of at least 16 bits. With discharge current data These data are stored in the PLC's RingBuffer, forming continuous electrical parameter signals;

[0081] S3. Sensitivity calculation steps: The system processes the electrical parameter signals in the buffer using the digital signal processing (DSP) instruction set integrated within the PLC.

[0082] The core lies in Separated from Excited dynamic response amplitude With phase information The system calculates the local sensitivity index that characterizes the process gain. The calculation formula is: ,in Given the known disturbance input amplitude;

[0083] S4. Sliding mode feedback adjustment steps: Calculate the... The input is used as a feedback variable to the sliding mode controller (SlidingModeController); the sliding mode controller then... Relative to target sensitivity threshold The deviation direction, i.e. the error sign, is used to dynamically adjust the reference flow control signal. The value;

[0084] For example, when Less than If this occurs, it indicates that the system may have slipped into a flat region of metallic or compound mode, and the controller will adjust the flow to bring the system back to the high-sensitivity transition region.

[0085] This embodiment completes the entire process from signal generation, acquisition to calculation and control within the PLC, avoiding time jitter caused by communication between the PC and PLC. Step S4 introduces sliding mode control, which utilizes its insensitivity to perturbations of nonlinear system parameters to effectively overcome the model uncertainty in the magnetron sputtering process, enabling the system to maintain extremely high robustness when facing characteristic drift caused by target etching and consumption.

[0086] This embodiment details the full-link control logic based on a PLC real-time operating system, aiming to solve the communication delay and timing jitter problems existing in traditional discrete control architectures. Through the tight coupling execution of S1 to S3, the system completes the closed-loop operation from signal injection to feature extraction within a single PLC scan cycle, ensuring strict synchronization between control commands and physical processes. In particular, the sliding mode controller introduced in S4 utilizes its variable structure characteristics to cope with the high nonlinearity of the magnetron sputtering process. In response to the deviation of local sensitivity indicators from the target threshold, the controller can quickly generate correction commands. This design enables the system to maintain long-term stability of the process operating point even when facing external disturbances such as changes in target etching depth or fluctuations in gas pipeline pressure.

[0087] Example 3:

[0088] S1 specifically includes:

[0089] S11. A disturbance waveform table is preset in the memory of the PLC controller. The disturbance waveform table stores data points of sine wave sequence or pseudo-random binary sequence.

[0090] S12. Set the disturbance scanning period, read the data points in the disturbance waveform table in sequence according to the disturbance scanning period, and multiply the read data points by the preset disturbance intensity coefficient to generate a periodic micro-disturbance signal.

[0091] S13. Obtain the reference flow control signal from the previous control cycle, add the periodic micro-disturbance signal to the reference flow control signal, and generate a composite drive command. The update frequency of the composite drive command is higher than the thermodynamic response frequency of the deposition process.

[0092] This embodiment provides a specific design for the generation of disturbance signals in step S1 to adapt to the operational characteristics of the PLC;

[0093] S11. Waveform preset: Preset a disturbance waveform table in the non-volatile memory of the PLC controller, such as DataBlock; the table stores discrete data points of a sine wave sequence or pseudo-random binary sequence PRBS with a complete cycle.

[0094] Disturbance waveform table: This is a pre-calculated array of floating-point numbers. Its purpose is to avoid the PLC performing time-consuming trigonometric function calculations such as SIN or COS instructions during real-time operation, thereby saving CPU computing power.

[0095] S12, Table Lookup Scan: Set the disturbance scan period The PLC uses pointer addressing to... Read the data points in the disturbance waveform table sequentially; multiply the read data points by the preset disturbance intensity coefficient. Generate the current periodic micro-perturbation signal ;

[0096] S13, High-frequency update: Obtain the reference flow control signal from the previous control cycle. to The composite driving command is generated by addition; the key is that the update frequency of the composite driving command is set to, for example, 1 kHz, higher than the thermodynamic response frequency of the deposition process, which is usually 1-10 Hz.

[0097] This embodiment uses a lookup table method to generate disturbance signals, which greatly reduces the computational load of the PLC and ensures the stability of the control cycle. At the same time, setting a high update frequency allows micro-disturbances to be regarded as a dither signal. This high-frequency dithering is not only used for detection, but can also eliminate the static friction of the MFC mechanical valve core to a certain extent, thereby improving the micro-linearity of the actuator response.

[0098] This embodiment cleverly solves the problem of limited computing power in industrial PLCs when performing complex mathematical operations by using a preset disturbance waveform table and a lookup table scanning mechanism. By discretizing the generation process of periodic micro-disturbance signals into memory pointer read operations, the system greatly frees up CPU resources for the calculation of core control algorithms. More importantly, by setting the update frequency of composite drive instructions to be higher than the thermodynamic response frequency of the deposition process, the micro-disturbance signals not only play the role of detecting system impedance, but also superimpose a high-frequency dither signal onto the actuator, effectively overcoming the static friction dead zone of the mass flow controller MFC valve core, and significantly improving the linearity and response speed during micro-flow regulation.

[0099] Example 4:

[0100] S3 specifically includes:

[0101] S31. Establish a sliding time window for the collected electrical parameter signals and extract the voltage timing data within the sliding time window;

[0102] S32. Perform discrete Fourier transform or digital phase-locked amplification on the voltage time series data, and extract the spectral amplitude at the preset frequency, which is recorded as the response amplitude.

[0103] S33. Obtain the input amplitude of the periodic micro-perturbation signal at a preset frequency, calculate the ratio of the response amplitude to the input amplitude, and define the ratio as the local sensitivity index, wherein the local sensitivity index characterizes the transient response rate of the deposition voltage to the change in gas flow rate.

[0104] S3 also includes a delay compensation step:

[0105] S34. Calculate the cross-correlation function between the periodic micro-perturbation signal and the electrical parameter signal;

[0106] S35. Search for the peak position of the cross-correlation function, determine the time lag corresponding to the peak position, and mark the time lag as the gas transmission delay time.

[0107] S36. Time alignment correction is performed on the electrical parameter signal based on the gas transmission delay time. The amplitude extraction operation in S32 is performed based on the corrected data to eliminate the influence of transmission delay on sensitivity calculation.

[0108] This embodiment details how to accurately extract the sensitivity index in step S3 and solve the gas transmission delay problem. To ensure the reproducibility of the algorithm in industrial settings and to correct the dimensional definitions in digital signal processing, the specific parameters and logic are as follows:

[0109] S31-S33, Frequency Domain Amplitude Extraction: A sliding time window (SlidingWindow) is established for the acquired electrical parameter signals; to avoid dimensional confusion and insufficient sampling points, the time span of the window is set. For the disturbance period Integer multiples, for example ;

[0110] Based on the sampling frequency of the PLC analog input module Determine the number of sampling points within the window;

[0111]

[0112] Extract voltage timing data within the window Before performing the transformation, to prevent sidelobe effects caused by aperiodic truncation, the transformation is first performed on... Multiply by HanningWindow:

[0113]

[0114] in Perform Discrete Fourier Transform (DFT); calculate the preset frequency. The corresponding index value in the DFT output array The calculation formula is: This embodiment defines a local sensitivity index. The calculation logic is as follows: Before performing amplitude extraction, delay compensation steps S34-S36 are performed first to determine the direction factor. And align the time series:

[0115]

[0116] in, The amplitude constant of the preset periodic micro-perturbation signal;

[0117] Voltage timing data within the sliding window;

[0118] The preset perturbation frequency is, for example, 10Hz;

[0119] : Represents the magnitude of the spectral amplitude at a specific frequency;

[0120] This index characterizes the transient response rate of deposition voltage to changes in gas flow rate;

[0121] S34-S36, Delay Compensation: Due to the physical delay caused by gas transport, the system needs to perform the following steps:

[0122] Cross-correlation calculation S34: Calculate the periodic micro-perturbation signal sequence correspond With electrical parameter signal sequence correspond Cross-correlation function between :

[0123]

[0124] in, The number of discrete lag points is set to the search range as follows: , , To estimate the maximum physical latency, such as 500ms; note that this is for support offset, sequence The actual buffer length must be at least ;

[0125] Extreme Value Search and Phase Determination S35: Considering that voltage and flow rate may be negatively correlated in physical vapor deposition, the location of the maximum absolute value is searched:

[0126]

[0127] Convert the discrete lag points into physical time to determine the gas transport delay time. At the same time, record The sign of the peak value; a negative value indicates that the system is in the negative impedance response region; the direction factor is defined. ,like The peak value is positive. ;like The peak value is negative. ;

[0128] Time alignment S36: based on Time alignment correction is applied to the electrical parameter signal, meaning that the voltage at the current moment is used when calculating sensitivity. Flow commands at historical moments Perform pairing operations;

[0129] This embodiment corrects the definition that may lead to dimensional errors in the original description by clearly distinguishing the relationship between the window time span and the number of sampling points, thus ensuring the correctness of the FFT frequency resolution; S34 supplements the requirements for sequence buffering, ensuring the boundary safety of cross-correlation operations; combined with the extreme value search based on absolute values, it improves the support for negative differential impedance characteristics in the PVD process, ensuring the causal consistency of closed-loop regulation.

[0130] Example 5:

[0131] S4 specifically includes:

[0132] S41. Preset a hysteresis loop characteristic model of the deposition process, and define the metal mode region, transition mode region and compound mode region in the hysteresis loop characteristic model.

[0133] S42. Identify the region to which the current working point belongs based on the value of the local sensitivity index: If the local sensitivity index is lower than the preset low sensitivity threshold, determine that the current region is in the metal mode region or the compound mode region; if the local sensitivity index is higher than the preset high sensitivity threshold, determine that the current region is in the transition mode region.

[0134] S43. Set the high sensitivity threshold as the target sensitivity threshold, calculate the difference between the current local sensitivity index and the target sensitivity threshold, and generate a control error signal;

[0135] S4 also includes adjustment steps based on sliding mode control laws:

[0136] S44. Define the switching function and construct the sliding surface based on the control error signal and its rate of change.

[0137] S45. Based on the sign of the sliding surface, select the sign direction of the control gain and generate discontinuous control correction quantities: when the system state is on one side of the sliding surface, output a positive flow correction step size to increase the reference flow control signal; when the system state is on the other side of the sliding surface, output a negative flow correction step size to decrease the reference flow control signal.

[0138] S46. Perform integral smoothing on the discontinuous control correction, update the reference flow control signal at the next moment, and drive the local sensitivity index to converge toward the high sensitivity threshold.

[0139] This embodiment describes in detail the sliding mode control strategy based on the hysteresis loop feature in step S4, especially the algorithm implementation details and parameter acquisition sources for the PLC discrete-time system;

[0140] S41-S42, Pattern Recognition and Feature Model Construction: The preset hysteresis loop feature model is not static data, but dynamically constructed through the InitializationScan process.

[0141] Parameter acquisition scan: Before the process begins, the PLC controls the MFC to perform a triangular wave flow scan from a preset minimum sustaining flow rate, such as 20% to 100%, and then back to the preset minimum sustaining flow rate. The scan cycle is set to 30 to 60 seconds to ensure that the thermodynamic hysteresis response of the process can be fully excited and recorded, and the voltage is recorded in real time. Traffic The response curve;

[0142] Feature definition: Calculating the discrete derivative of the scanned data Identify the maximum value of the absolute value of the derivative. ;

[0143] Set a high sensitivity threshold Corresponding to the edge of the transition zone;

[0144] Set low sensitivity threshold The corresponding flat section of the metal / compound region;

[0145] Region identification:

[0146] like It was determined to be a metal / compound mode region;

[0147] like It is determined to be a transition mode region;

[0148] S43-S46, Discretization of sliding mode control law: The system will Set as the target sensitivity threshold In order to achieve the discrete scan cycle time of the PLC Internal sliding mode control is implemented using a dual-rate processing mechanism:

[0149] Signal injection and acquisition in S1 to S2, and control law update in S4, are executed in a high-priority interrupt OB within 1ms. The time-consuming FFT and cross-correlation operations in S3 are executed in a low-priority loop task. The cross-correlation operation uses a fast algorithm with a limited search range, calculating only within a local window near the estimated delay time to ensure that the operation time is less than 10ms, for example, updating once every 10ms. The controller calls the latest version within each Ts cycle. The value is used to perform the following operations:

[0150] Errors and Sliding Surface Construction S43-S44: In the first Calculate the control error for each control cycle. ; Calculate the rate of change of error Constructing discrete sliding surfaces ;

[0151] Parameter c is the time constant coefficient of the sliding surface, in seconds, and its value range is... s, this value determines the convergence rate of the system in the phase plane;

[0152] Generate control correction quantity S45: Based on the constant-rate approaching law, generate discontinuous control correction quantity. Dead zone threshold is introduced here. Pick 2% to eliminate high-frequency jitter:

[0153]

[0154] The preset flow correction step size is set to 0.05% of the MFC full scale. For example, for a 100 sccm MFC... sccm ensures that the step size is greater than the mechanical resolution of MFC;

[0155] Integral Smoothing and Instruction Update S46: To convert the switch quantity into a smoothing instruction, perform a discrete integral update:

[0156]

[0157] Integral gain coefficient , used to adjust control flexibility;

[0158] In this embodiment, through the initial scanning step in S41, the system automatically obtains the sensitivity threshold adapted to the current target state, whether it is a new or old target, thus solving the problem that a fixed threshold cannot cope with the evolution of target etching; the dead zone threshold is clearly defined in S45. and physical step size Combined with the integral smoothing of S46, it effectively solves the chattering problem that is common in sliding mode control in digital systems. It not only ensures strong locking of the operating point in the transition zone, but also avoids wear of the MFC valve core due to high-frequency oscillation.

[0159] Example 6:

[0160] The system also includes an abnormal drift blocking mechanism, which is implemented as follows:

[0161] The rate of change of local sensitivity indicators over time;

[0162] If the rate of change over time exceeds the preset avalanche threshold and the absolute value of the electrical parameter signal shows a monotonically decreasing trend, it is determined that the system is about to enter the target poisoning state.

[0163] The PLC controller immediately triggers a blocking command, forcibly resetting the reference flow control signal to a preset safe backoff value, and suspending the injection of periodic micro-disturbance signals until the electrical parameter signals return to the preset initial state range.

[0164] This embodiment describes the abnormal drift blocking mechanism of the system as a safety defense line. The key is how to accurately determine avalanche and monotonic descent mathematically and ensure the consistency of physical dimensions.

[0165] The specific implementation of the anomaly monitoring logic: The system runs the following monitoring algorithm in the PLC:

[0166] Sensitivity Change Rate Monitoring Avalanche Determination: To achieve the monitoring of the time change rate of the local sensitivity index in this embodiment, the second-order change rate of the local sensitivity index, i.e., acceleration, is calculated:

[0167]

[0168] in, Indicates the current control cycle. Indicates the previous control cycle. Indicates the control cycle preceding the previous one;

[0169] like Exceeding the preset avalanche threshold Then, a potential instability indicator is set.

[0170] The method of obtaining the data is as follows: During the system debugging phase, the target material is artificially poisoned once, and the peak value of the second-order change rate of sensitivity at the moment of poisoning is recorded. 80% of this peak value is taken as the threshold.

[0171] Electrical parameter signal trend determination and monotonicity confirmation: To accurately determine a monotonically decreasing trend in a noisy environment, the system uses a sliding window trend algorithm to process the data modulus: establishing a length of... For example A voltage sampling sliding window is used to uniformly perform absolute value operations on the original sampled data within the window, constructing an absolute value sequence; the slope of the linear regression of this sequence is then calculated. With intercept :

[0172]

[0173]

[0174] Key definition: Wherein, For the first in the sliding window The relative time value corresponding to each data point Unit: seconds;

[0175] : Magnitude, the absolute value of the electrical parameter sample at the corresponding moment, in volts;

[0176] : All within the sliding window The arithmetic mean;

[0177] This definition ensures The physical meaning is the voltage change rate V / s; simultaneously, the determination coefficient is calculated. To verify the reliability of the linear fit:

[0178]

[0179] in, These are the regression fitted values. ; This is the arithmetic mean of the data within the sliding window;

[0180] like and Minimum descent slope threshold, where The preset threshold for the absolute value of the descent slope is, for example, 50V / s, and If , then it is determined to be a valid monotonically decreasing condition;

[0181] Blocking instruction trigger: The AND gate logic outputs a high level only when the potential instability flag is TRUE and the valid monotonic descent judgment is true, and the system determines that it is about to enter the target poisoning state.

[0182] Execute blocking action: The PLC controller immediately triggers the highest priority blocking instruction:

[0183] Forced reset: Reset the reference flow control signal Immediately reset to the preset safe rollback value Defined as 90% of the upper limit of flow rate in the metal mode zone;

[0184] Pause Disturbance: Stop The injection will Locked to pure DC component;

[0185] Recovery mechanism: Maintain this state until an electrical parameter signal is detected, specifically the DC average value of the deposited voltage, which rises back to the preset initial state range and remains stable for more than 3 seconds before the closed-loop control is allowed to be reactivated.

[0186] This embodiment clarifies the time variable in regression calculations. The physical unit makes the slope threshold It has a clear engineering significance V / s, avoiding threshold failure caused by different sampling rates; combined with absolute value-based trend analysis logic, it effectively filters out the inherent high-frequency random noise in plasma glow discharge, prevents false blocking caused by instantaneous voltage fluctuations, and significantly improves the safety of the system in critical states.

[0187] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A PLC-based control system for metal film deposition process, characterized in that, include: Active perturbation injection module: used to generate periodic micro-perturbation signals with preset frequency and preset amplitude, superimpose the periodic micro-perturbation signals onto the reference flow control signal to generate composite drive commands, and send the composite drive commands to the actuator of the deposition process; Dynamic response acquisition module: used to acquire electrical parameter signals output by the deposition power supply in real time and construct a time-series data stream containing the dynamic response of the process; Impedance fingerprint demodulation module: used to perform frequency domain analysis on the time-series data stream, extract the response component in the electrical parameter signal that has the same frequency as the periodic micro-perturbation signal, and calculate the local sensitivity index of the current process based on the ratio relationship between the response component and the periodic micro-perturbation signal. State mapping and closed-loop adjustment module: used to map the local sensitivity index to the working point state on the nonlinear hysteresis curve, calculate the deviation between the local sensitivity index and the preset target sensitivity threshold, and generate a correction command for the reference flow control signal based on the deviation, so as to lock the working range of the deposition process.

2. The PLC-based metal film deposition process control system according to claim 1, characterized in that, The modules are interconnected using the following method: S1. The PLC controller generates the periodic micro-disturbance signal and linearly superimposes it with the reference flow control signal at the current moment, and outputs the composite drive command to the mass flow controller. S2. Synchronously acquire the discharge voltage and discharge current data of the deposition power supply through the analog input interface to form the electrical parameter signal; S3. Using digital signal processing algorithms, the dynamic response amplitude and phase information excited by the periodic micro-perturbation signal are separated from the electrical parameter signal, and the local sensitivity index characterizing the process gain is calculated. S4. Input the local sensitivity index as a feedback variable to the sliding mode controller, and dynamically adjust the value of the reference flow control signal according to the deviation direction of the local sensitivity index relative to the target sensitivity threshold, so that the local sensitivity index is maintained within the range limited by the target sensitivity threshold.

3. The PLC-based metal film deposition process control system according to claim 2, characterized in that, S1 specifically includes: S11. A disturbance waveform table is preset in the memory of the PLC controller. The disturbance waveform table stores data points of a sine wave sequence or a pseudo-random binary sequence. S12. Set the disturbance scanning period, read the data points in the disturbance waveform table sequentially according to the disturbance scanning period, and multiply the read data points by the preset disturbance intensity coefficient to generate the periodic micro-disturbance signal; S13. Obtain the reference flow control signal of the previous control cycle, add the periodic micro-perturbation signal to the reference flow control signal, and generate the composite drive command, wherein the update frequency of the composite drive command is higher than the thermodynamic response frequency of the deposition process.

4. The PLC-based metal film deposition process control system according to claim 3, characterized in that, S3 specifically includes: S31. Establish a sliding time window for the collected electrical parameter signals and extract the voltage timing data within the sliding time window; S32. Perform discrete Fourier transform or digital phase-locked amplification on the voltage time-series data, and extract the spectral amplitude at the preset frequency, which is recorded as the response amplitude. S33. Obtain the input amplitude of the periodic micro-perturbation signal at the preset frequency, calculate the ratio of the response amplitude to the input amplitude, and define the ratio as the local sensitivity index, wherein the local sensitivity index characterizes the transient response rate of the deposition voltage to changes in gas flow rate.

5. A PLC-based metal film deposition process control system according to claim 4, characterized in that, S3 further includes a delay compensation step: S34. Calculate the cross-correlation function between the periodic micro-perturbation signal and the electrical parameter signal; S35. Search for the peak position of the cross-correlation function, determine the time lag corresponding to the peak position, and mark the time lag as the gas transmission delay time. S36. Based on the gas transmission delay time, perform time alignment correction on the electrical parameter signal, and perform the amplitude extraction operation in S32 based on the corrected data to eliminate the influence of transmission delay on sensitivity calculation.

6. A PLC-based metal film deposition process control system according to claim 2, characterized in that, S4 specifically includes: S41. A hysteresis loop characteristic model of the deposition process is preset, in which the metal mode region, transition mode region and compound mode region are defined. S42. Identify the region to which the current working point belongs based on the value of the local sensitivity index: if the local sensitivity index is lower than a preset low sensitivity threshold, determine that the current region is in the metal mode region or the compound mode region; if the local sensitivity index is higher than a preset high sensitivity threshold, determine that the current region is in the transition mode region. S43. Set the high sensitivity threshold as the target sensitivity threshold, calculate the difference between the current local sensitivity index and the target sensitivity threshold, and generate a control error signal.

7. A PLC-based metal film deposition process control system according to claim 6, characterized in that, The S4 also includes an adjustment step based on a sliding mode control law: S44. Define a switching function and construct a sliding surface based on the control error signal and its rate of change; S45. Based on the sign of the sliding surface, select the sign direction of the control gain and generate a discontinuous control correction value: When the system state is on one side of the sliding surface, a positive flow correction step size is output, and the reference flow control signal is increased; When the system state is on the other side of the sliding surface, a negative flow correction step size is output to reduce the reference flow control signal; S46. Perform integral smoothing on the discontinuous control correction amount, update the reference flow control signal at the next moment, and drive the local sensitivity index to converge toward the high sensitivity threshold.

8. A PLC-based metal film deposition process control system according to claim 6, characterized in that, The system also includes an abnormal drift blocking mechanism, which is implemented as follows: Monitor the time change rate of the local sensitivity index; if the time change rate exceeds the preset avalanche threshold and the absolute value of the electrical parameter signal shows a monotonically decreasing trend, it is determined that the system is about to enter the target poisoning state. The PLC controller immediately triggers a blocking command, forcibly resetting the reference flow control signal to a preset safe backoff value, and suspending the injection of the periodic micro-disturbance signal until the electrical parameter signal returns to the preset initial state range.

Images