Laser intelligent control and remote management method and system based on multi-mode driving

Abstract

The invention discloses a laser intelligent control and remote management method and system based on multi-mode driving, and belongs to the technical field of laser control, and the method comprises the steps: collecting the operation data of at least two different modes in the operation process of a laser, and forming a process feature vector; the operation data of the different modes at least comprise any two of laser output power data, retroreflective light data, spectral data, temperature data and current or voltage ripple data. Multi-modal operation data in the operation process of a laser is constructed into a cross-modal physical consistency residual error and a process credible evidence is formed, remote certification or integrity verification is performed on a local control end to form a platform credible evidence, and a network controllability index is calculated based on network time delay and jitter to form a network credible evidence. The three types of evidences can be fused into the credibility score, so that the problem that the unified constraint of the process state, the platform credibility and the network controllability is difficult to quantify at the same time in a remote management scene is solved.

Description

Technical Field

[0001] This invention relates to the field of laser control technology, and more specifically, to a method and system for intelligent control and remote management of lasers based on multi-mode driving. Background Technology

[0002] Existing lasers typically employ local closed-loop control in processes such as welding, cutting, and additive manufacturing. They rely on signals such as power, backlight reflection, spectrum, temperature, and power supply ripple to adjust the output power or pulse waveform in order to reduce melt depth fluctuations, avoid critical hole instability, or suppress heat accumulation.

[0003] With the increasing demand for production line digitalization and centralized equipment operation and maintenance, more and more solutions are introducing remote monitoring and remote parameter distribution, enabling remote terminals to update waveform libraries, adjust power limits, change power change rates or distribute process recipes, and perform quality traceability and fault diagnosis through multimodal data transmitted back via the network.

[0004] However, in remote control mode, existing technologies generally face the following key contradictions: On the one hand, network latency and jitter can lead to delayed closed-loop feedback or even disordered arrival sequence of control commands. The remote end has difficulty in timely determining whether a certain waveform has been accurately executed, which can easily lead to mismatch in energy boundary settings or failure of power change rate control, thereby causing process risks such as overshoot, burn-through, or expansion of the heat-affected zone. On the other hand, the remote control link introduces platform trust and execution consistency issues. For example, the software version, waveform library file, parameter table, or control logic of the local control end may be affected by abnormal upgrades, configuration drift, or malicious tampering. Off-baseline, even if the correct strategy is issued remotely, it may be replaced with a different waveform or the old command may be replayed locally, resulting in an unverifiable inconsistency between "remotely considered to have been executed" and "local actual output". Moreover, existing solutions mostly stay at the level of alarms, logs or simple access control, lacking a calculable evidence loop to quantify "process multimodal consistency", "platform trust status" and "network controllability" into a unified, auditable and reproducible control constraint variable, and further perform hierarchical convergence and hysteresis release on the allowed control modes, the set of callable waveforms, and the energy input boundary and power change rate boundary.

[0005] Therefore, we believe that existing technologies, especially in remote management scenarios, cannot simultaneously address the output risks caused by the coupling of network uncertainty, control end credibility, and unverifiable waveform execution consistency. There is a lack of a technical solution that can dynamically converge control permissions based on multimodal evidence and credible evidence and impose verifiable constraints on waveform execution consistency. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A multi-mode driven intelligent control and remote management method for lasers includes the following steps:

[0008] S1: During the operation of the laser, at least two types of operation data of different modes are collected and process feature vectors are formed. The operation data of different modes include at least two types of laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Network transmission status data is collected simultaneously. The network transmission status data includes at least time delay parameters and jitter parameters.

[0009] S2: Construct cross-modal physical consistency residuals based on the process feature vectors. The cross-modal physical consistency residuals are used to characterize the degree of deviation of different modal operation data under preset physical constraints and output process credibility evidence.

[0010] S3: Perform remote verification or integrity check on the local control terminal of the laser to output platform trust evidence, which is used to characterize the degree to which the hardware and software status of the local control terminal meets the preset trust baseline;

[0011] S4: Calculate the network controllability index based on the network transmission status data and output network trust evidence. The network trust evidence is used to characterize the degree of influence of the remote control link on the stability of closed-loop control.

[0012] S5: The process credibility evidence, the platform credibility evidence, and the network credibility evidence are fused together to obtain a credibility score, and the credibility score is mapped to a control authority set. The control authority set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary.

[0013] S6: Generate control waveform parameters and form a laser-driven control waveform under the constraints of the control authority set, and embed a physical watermark sequence in the laser-driven control waveform. The physical watermark sequence is achieved by perturbation of the pulse sequence in the time domain or frequency domain and satisfies the constraint of not exceeding the preset quality influence threshold.

[0014] S7: The local control terminal executes the laser drive control waveform in hard real-time mode to drive the laser output, and continuously monitors the confidence score and the network controllability index. When the confidence score is detected to be lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters the autonomous degradation mode and restricts the control waveform to the safe waveform in the waveform library subset, while converging the energy input boundary or the power change rate boundary.

[0015] S8: The remote management terminal receives the returned multimodal operation data and detects the physical watermark sequence based on the returned data to verify the execution consistency of the laser drive control waveform. At the same time, the confidence score, the control authority set mapping result, the autonomous degradation trigger event, and the physical watermark sequence verification result are recorded to form a traceable remote management audit record.

[0016] Furthermore, the cross-modal physical consistency residuals are obtained and used to form credible evidence of the process in the following manner:

[0017] After performing time alignment and sampling window constraint processing on the laser output power data and back-reflection data in the process feature vector, a back-reflection prediction value is generated based on a preset power-back-reflection coupling constraint relationship. The deviation between the back-reflection data and the back-reflection prediction value is used as the instantaneous residual. The instantaneous residual is further recursively accumulated to form a residual trend quantity. The residual trend quantity is then mapped to the process credibility evidence, such that the process credibility evidence monotonically decreases as the residual trend quantity increases. The process credibility evidence is used as a leading quantity for the convergence constraint of the control authority set in the fusion calculation of the credibility score, thereby triggering the entry condition of the autonomous degradation mode in advance.

[0018] Furthermore, the power back-reflection coupling function is an updatable coupling function, and the update of the coupling parameters is only allowed to be triggered under the condition that the platform's credible evidence meets the preset credible baseline and the verification result of the physical watermark sequence based on the back-transmitted multimodal operation data of the remote management terminal is consistent with the execution. After being allowed to be triggered, the spectral drift index is extracted from the spectral data within the same sampling window as the residual trend amount, and the spectral drift index is written into the update objective function of the coupling parameters to solve the updated coupling parameters. The back-reflection prediction value is corrected using the updated coupling parameters to obtain the corrected residual trend amount and update the process credible evidence accordingly, so that the sensitivity of the process credible evidence decreases when the spectral drift index increases and the control authority set is further converged.

[0019] When the platform's trusted evidence does not meet the preset trusted baseline or the physical watermark sequence verification result does not indicate consistent execution, the coupling parameters are locked and the process trusted evidence is updated according to the monotonically decreasing rule to force convergence of the control authority set and trigger the autonomous degradation mode.

[0020] Furthermore, the mapping from the confidence score to the control permission set is implemented using a step-by-step convergence segmentation rule. The segmentation rule sets a first confidence threshold and a second confidence threshold, with the first confidence threshold being less than the second confidence threshold. The permission status word is determined according to the comparison result between the confidence score and the first confidence threshold and the second confidence threshold. The permission status word is used to uniquely determine the set of allowed control modes and the corresponding set of available control waveforms, and simultaneously determine the energy input boundary and the power change rate boundary.

[0021] When the confidence score is not higher than the first confidence threshold, the permission status word is placed in a strong convergence state so that the available set of control waveforms converges to a safe waveform set and the energy input boundary and power change rate boundary converge to a preset minimum boundary and the autonomous degradation mode is forcibly maintained.

[0022] When the confidence score is higher than the first confidence threshold and not higher than the second confidence threshold, the permission status word is placed in the mid-convergence state to allow only limited step size updates to the control waveform parameters under the premise that the autonomous degradation mode is lifted, and to keep the power change rate boundary from exceeding the upper limit corresponding to the network controllability index.

[0023] When the confidence score is higher than the second confidence threshold, the permission status word is placed into a weak convergence state to unlock the call to the extended waveform set in the control waveform library and update the energy input boundary and power change rate boundary according to the monotonically relaxed rule, so that the control permission set is unlocked step by step as the confidence score increases and converges step by step as the confidence score decreases.

[0024] Furthermore, the update of the permission status word follows a state transition rule with hysteresis constraints. The state transition rule includes maintaining a state holding register corresponding to the permission status word on the local control terminal and setting a minimum holding window number for each state transition. The transition of the permission status word from a weakly converged state to a moderately converged state and from a moderately converged state to a strongly converged state is configured as an immediate-effective transition. The immediate-effective transition is triggered by the platform's trusted evidence not meeting a preset trusted baseline, the physical watermark sequence verification result not representing execution consistency, or the trust score not exceeding the first trust threshold. After triggering, the state holding register is locked.

[0025] Furthermore, the migration of the permission status word from a strongly converged state to a moderately converged state is only allowed to be triggered when the platform's trusted evidence continuously meets the preset trusted baseline and the physical watermark sequence verification results are consistently performed within a continuous first minimum hold window. The migration of the permission status word from a moderately converged state to a weakly converged state is only allowed to be triggered after the aforementioned triggering is completed, and further when the network controllability index does not deteriorate within a continuous second minimum hold window and the trust score is continuously higher than the second trust threshold. Thus, permission unlocking must go through a step-by-step release path from strongly converged to moderately converged and then to weakly converged, and is subject to the prior gating constraints of platform trust and watermark consistency.

[0026] Furthermore, the physical watermark sequence is determined by the remote management terminal based on the proof random number generation seed bound to the platform's trusted evidence and synchronously forms a reference watermark sequence. The reference watermark sequence is mapped at the local control terminal to a perturbation coding amount of the laser drive control waveform to perform window-allocated micro-modulation on the pulse width or pulse repetition frequency of the pulse sequence, so that the energy deviation introduced by the perturbation coding amount is always limited to the preset quality influence threshold.

[0027] After receiving the reflected light data, the remote management terminal performs synchronous alignment on the reflected light data under the same time reference as the sampling window, and extracts watermark features after alignment. Then, it obtains the watermark confidence by correlation matching with the reference watermark sequence, and uses the watermark confidence as a quantitative representation of the verification result of the physical watermark sequence. When the watermark confidence is lower than a preset threshold, the process credible evidence is updated according to the monotonically decreasing rule, and the permission status word is forcibly migrated to a strong convergence state and the state is locked to maintain the register. Only when the watermark confidence continuously meets the preset threshold and meets the minimum holding window number is the lock allowed to be released and enter the step-by-step unlocking path, so that the detectability of the physical watermark sequence becomes a necessary condition for controlling permission unlocking and forms a closed-loop constraint relationship with the confidence score.

[0028] A laser intelligent control and remote management system based on multi-mode driving further includes a laser, a local control terminal, a remote management terminal, and a communication link for connecting the local control terminal and the remote management terminal.

[0029] The local control terminal is connected to the laser and configured to collect at least two types of operating data in different modes during laser operation to form a process feature vector. The operating data in different modes includes at least two of the following: laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Simultaneously, the network transmission status data of the communication link is acquired. The network transmission status data includes at least delay parameters and jitter parameters.

[0030] The local control terminal is further configured to construct cross-modal physical consistency residuals based on the process feature vector to output process credible evidence, and to perform remote proof or integrity verification on the local control terminal to output platform credible evidence, while calculating network controllability indicators based on the network transmission status data to output network credible evidence.

[0031] The local control terminal is further configured to fuse the process credible evidence, the platform credible evidence, and the network credible evidence to calculate a credibility score and generate a control permission set accordingly. The control permission set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary.

[0032] The local control terminal is further configured to generate a laser-driven control waveform under the constraints of the control permission set and embed a physical watermark sequence in the laser-driven control waveform, and execute the laser-driven control waveform in hard real-time mode to drive the laser output. At the same time, when the credibility score is lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters an autonomous degradation mode and restricts the control waveform to a safe waveform in the waveform library subset and converges the energy input boundary or the power change rate boundary.

[0033] The remote management terminal is configured to receive multimodal operation data transmitted back from the local control terminal, and detect the physical watermark sequence based on the transmitted data to verify the execution consistency of the laser drive control waveform and form a traceable remote management audit record.

[0034] Furthermore, the local control terminal is configured to solidify the mapping result of the confidence score into an authorization status word and use the authorization status word to uniquely define the waveform library subset and the energy input boundary or the power change rate boundary. The authorization status word is determined by a first confidence threshold and a second confidence threshold, with the first confidence threshold being less than the second confidence threshold. A step-by-step unlocking path with hysteresis constraints is formed by setting a minimum hold window number, so that the authorization status word takes effect immediately when it migrates to a stronger convergence direction, while it is only allowed to be triggered when it migrates to a weaker convergence direction, provided that the platform's credible evidence continuously meets the preset credible baseline, the physical watermark sequence verification result is continuously characterized as consistent within the continuous minimum hold window number, and the network controllability index does not deteriorate within the corresponding window.

[0035] The remote management terminal is further configured to determine a reference watermark sequence based on a proof random number seed bound to the platform's trusted evidence, and to determine the physical watermark sequence in consistency with the local control terminal. Furthermore, it obtains the watermark confidence level based on the correlation matching between the returned reflected light data and the reference watermark sequence, and uses the watermark confidence level to quantify and characterize the verification result of the physical watermark sequence. This ensures that when the watermark confidence level is lower than a preset threshold, the local control terminal forcibly migrates the permission status word to a strong convergence state and locks the step-by-step unlocking path until the watermark confidence level continuously meets the minimum hold window number.

[0036] In summary, the present invention has the following beneficial effects:

[0037] By constructing multimodal operational data during laser operation into cross-modal physical consistency residuals to form process credibility evidence, and simultaneously performing remote verification or integrity checks on the local control terminal to form platform credibility evidence, and calculating network controllability indicators based on network latency and jitter to form network credibility evidence, these three types of evidence can be integrated into a credibility score. This solves the problem of simultaneously quantifying the unified constraints of process status, platform credibility, and network controllability in remote management scenarios.

[0038] By mapping the confidence score to a set of control permissions and solidifying it into permission status words, and adopting a step-by-step convergence mechanism with segmented thresholds, energy input boundaries, and power change rate boundaries, the system can automatically converge the allowed control modes and waveform library subsets when the network is uncertain or the platform status is abnormal. This solves the problem that energy overshoot, power change rate exceeding limits, and waveform abnormalities are difficult to control in a timely manner in remote control links.

[0039] By embedding a physical watermark sequence into the laser-driven control waveform and detecting the watermark based on the back-transmitted data on the remote side to quantify execution consistency, and using the watermark confidence as a necessary condition for unlocking permissions, and forming a step-by-step release path with the hysteresis holding window, the remote end can verify whether the waveform is executed as expected and force a downgrade lock when the consistency is insufficient. This solves the problem that the remote distribution strategy may be replaced, replayed or deviated and cannot be reliably detected and constrained.

[0040] By creating traceable audit records and documenting and solidifying the autonomous downgrade and unlocking processes, remote operation and maintenance and quality traceability have a consistent chain of evidence, thus solving the problem of lack of verifiable evidence for responsibility definition and post-event traceability under remote management conditions. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0042] Figure 1 This is a schematic diagram of the system architecture of the present invention;

[0043] Figure 2 This is a schematic diagram providing an overview of the method steps of the present invention;

[0044] Figure 3 This is a schematic diagram illustrating the permission status transition of the present invention;

[0045] Figure 4 This is a schematic diagram of the physical watermark embedding and verification process of the present invention;

[0046] Figure 5 This is a schematic diagram of the audit record structure of the present invention. Detailed Implementation

[0047] 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.

[0048] Example:

[0049] The following is in conjunction with the appendix Figure 1-5 The present invention will be described in further detail below.

[0050] Please see Figure 1-5 This invention provides a technical solution: a laser intelligent control and remote management method based on multi-mode driving, such as... Figure 1-5 As shown, it includes the following steps:

[0051] S1: During the operation of the laser, at least two types of operation data of different modes are collected and process feature vectors are formed. The operation data of different modes include at least two types of laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Network transmission status data is collected simultaneously. The network transmission status data includes at least time delay parameters and jitter parameters.

[0052] S2: Construct cross-modal physical consistency residuals based on the process feature vectors. The cross-modal physical consistency residuals are used to characterize the degree of deviation of different modal operation data under preset physical constraints and output process credibility evidence.

[0053] S3: Perform remote verification or integrity check on the local control terminal of the laser to output platform trust evidence, which is used to characterize the degree to which the hardware and software status of the local control terminal meets the preset trust baseline;

[0054] S4: Calculate the network controllability index based on the network transmission status data and output network trust evidence. The network trust evidence is used to characterize the degree of influence of the remote control link on the stability of closed-loop control.

[0055] S5: The process credibility evidence, the platform credibility evidence, and the network credibility evidence are fused together to obtain a credibility score, and the credibility score is mapped to a control authority set. The control authority set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary.

[0056] S6: Generate control waveform parameters and form a laser-driven control waveform under the constraints of the control authority set, and embed a physical watermark sequence in the laser-driven control waveform. The physical watermark sequence is achieved by perturbation of the pulse sequence in the time domain or frequency domain and satisfies the constraint of not exceeding the preset quality influence threshold.

[0057] S7: The local control terminal executes the laser drive control waveform in hard real-time mode to drive the laser output, and continuously monitors the confidence score and the network controllability index. When the confidence score is detected to be lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters the autonomous degradation mode and restricts the control waveform to the safe waveform in the waveform library subset, while converging the energy input boundary or the power change rate boundary.

[0058] S8: The remote management terminal receives the returned multimodal operation data and detects the physical watermark sequence based on the returned data to verify the execution consistency of the laser drive control waveform. At the same time, the confidence score, the control authority set mapping result, the autonomous degradation trigger event, and the physical watermark sequence verification result are recorded to form a traceable remote management audit record.

[0059] In this embodiment, the system is applied to a fiber laser with a rated output power of 1000W. The local control unit consists of a real-time controller and a programmable logic device (PLD). The real-time controller is responsible for multi-modal data synchronization and reliability calculation, while the PLD is responsible for microsecond-level waveform generation and execution. The remote management unit is deployed in a private cloud within the factory and communicates with the local control unit via a dedicated Ethernet network. The local control unit is electrically or optically connected to the power sampling channel, reflective light sampling channel, spectral sampling channel, temperature sampling channel, and power ripple sampling channel, respectively, and synchronized through a unified time base. The unified time base can adopt IEEE 1588 clock synchronization or an equivalent synchronization mechanism, so that different modal data can be aligned and calculated within the same sampling window.

[0060] In step S1, the local control terminal acquires at least two different types of operating data in different modes during laser operation and forms a process feature vector. For example, laser output power data P is acquired at a sampling frequency of not less than 10 kHz, reflected light data R is acquired at a sampling frequency of not less than 50 kHz, spectral data S is acquired at a frame rate of not less than 200 Hz, temperature data T is acquired at a frame rate of not less than 100 Hz, and current or voltage ripple data U is acquired at a sampling frequency of not less than 20 kHz. Simultaneously, network transmission status data, including one-way delay d and jitter j, is acquired. The one-way delay d is obtained by carrying the sending timestamp in the control message and transmitting the receiving timestamp back at the receiving end, and the jitter j is obtained by statistically analyzing the absolute value of the delay difference between adjacent messages. In this embodiment, the sampling window length Tw is set to 250ms, and the data in the window is resampled to a unified time grid point fs of 1kHz to form a process feature vector xt. The xt includes at least the power mean and variance, the backlight mean and variance, the wavelength shift of the main peak of the spectrum, the temperature slope, the root mean square of the ripple, and the network delay statistics and jitter statistics.

[0061] In step S2, the local control terminal constructs cross-modal physical consistency residuals based on the process feature vectors and outputs process credibility evidence. For example, a power back-reflection coupling constraint relationship is established as the back-reflection prediction model fθ, such that the aligned power sequence P yields a back-reflection prediction sequence Rhat. The back-reflection prediction can be implemented using a quadratic coupling function, where Rhat equals a0 + a1 * P + a2 * P squared, and a0, a1, and a2 constitute the coupling parameter θ. The instantaneous residual sequence rt is calculated within the window, where rt is the absolute value of the difference between the measured back-reflection value R and the predicted back-reflection value Rhat. The instantaneous residuals within the window are averaged to obtain the window residual rbar. The window... The residuals are recursively accumulated to form the residual trend quantity Qt, where the recursion rule is that Qt is equal to λ multiplied by the Qt of the previous window plus 1 minus λ multiplied by the rbar of the current window. The forgetting factor λ is between 0.85 and 0.98, and in this embodiment, it is 0.92. The residual trend quantity Qt is mapped to the process credibility evidence Ep. The mapping is implemented using a monotonically decreasing function. For example, Ep is equal to exp(-Qt) divided by σp, where σp is the scale parameter. In this embodiment, σp is between 0.03 and 0.08 and is 0.05, so that the process credibility evidence monotonically decreases as the residual trend quantity increases. The process credibility evidence Ep is used as the input for subsequent credibility score fusion and is used to trigger the convergence of the control authority set in advance.

[0062] In step S3, the local control terminal performs remote proof or integrity verification to output the platform trusted evidence Ec. For example, the remote management terminal sends a one-time challenge random number nonce to the local control terminal. The local control terminal reads the startup metric m, which includes at least the firmware version identifier, key program digest hash, configuration whitelist hash, and secure startup status identifier. The nonce is concatenated with m and signed to obtain the proof token, which is then sent back to the remote management terminal. The remote management terminal verifies the signature and certificate chain and compares whether m meets the preset trusted baseline. When the trusted baseline is met, the platform trusted evidence Ec is 1.00. When the trusted baseline is not met, the platform trusted evidence Ec is 0.10 to 0.30. In this embodiment, it is 0.20, which is used to force the convergence of the control permission set and prohibit high-risk waveform calls. The remote proof refresh period Tr is 1s to 5s. In this embodiment, it is 2s.

[0063] In step S4, the local control terminal calculates the network controllability index based on network transmission status data and outputs the network credible evidence En. For example, within each sampling window, the 95th quantile d95 of the delay and the root mean square jitter jrms are calculated, where jrms is the root mean square of the jitter sequence j. A network risk quantity Gn is constructed, which is equal to wd multiplied by d95 divided by Dref plus wj multiplied by jrms divided by Jref, where wd and wj are weights and wd plus wj equals 1. In this embodiment, wd is 0.7, wj is 0.3, Dref is 80ms, and Jref is 15ms. The network risk quantity Gn is mapped to the network credible evidence En. For example, En is equal to exp(-Gn), so that En decreases as the delay and jitter increase. When En is below 0.35, it is considered that the network controllability has deteriorated to the threshold condition.

[0064] In step S5, the process credible evidence Ep, platform credible evidence Ec, and network credible evidence En are fused to calculate a credibility score C, and the credibility score is mapped to a set of control permissions. For example, the credibility score C is linearly fused and saturated, where C equals wp multiplied by Ep plus wc multiplied by Ec plus wn multiplied by En, and wp plus wc plus wn equals 1. In this embodiment, wp is 0.35, wc is 0.45, and wn is 0.20, and C is truncated to the range of 0 to 1. C is mapped to the permission state word State, and a first credibility threshold C1 and a second credibility threshold C2 are set, with C1 less than C2. In this embodiment, C1 is 0.45 and C2 is 0.75. When C is not higher than C1, State is strongly convergent; when... When C is higher than C1 but not higher than C2, the State is moderately convergent; when C is higher than C2, the State is weakly convergent. The control permission set is determined by the State and includes at least the set of allowed control modes, the allowed subset of waveform libraries, the energy input boundary Pmax, and the power change rate boundary Kmax. Among them, strong convergence corresponds to only allowing the safe waveform library and Pmax is 0.60 to 0.75 of the rated power, and 0.70 in this embodiment. Kmax is the rated power change per second not exceeding 0.20, and 0.15 in this embodiment. Moderate convergence corresponds to allowing restricted step size updates and Pmax is 0.80 and Kmax is 0.25. Weak convergence corresponds to allowing the expanded waveform library and Pmax is 1.00 and Kmax is 0.40.

[0065] In step S6, control waveform parameters are generated under the constraints of the control permission set to form a laser-driven control waveform, and a physical watermark sequence is embedded in the laser-driven control waveform. For example, the waveform library includes a secure waveform library and an extended waveform library. The pulse width τ0 of the secure waveform library is 200ns to 600ns, and the repetition frequency F0 is 5kHz to 20kHz. The extended waveform library allows the addition of preamble pulses, trailing annealing pulses, and slope shaping. The physical watermark sequence is generated by the remote management terminal based on a seed generated by a proof random number bound to the platform's trusted evidence, and a reference watermark sequence Wref is generated and distributed. In this embodiment, a length L of 64 is generated within each sampling window. The binary sequence has a symbol period Tc, which is approximately 3.9ms (Tw divided by L). The local control terminal maps the watermark sequence to a pulse width perturbation coding amount, such that when the symbol is 1, the pulse width is τ0 multiplied by 1 plus δ, and when the symbol is 0, the pulse width is τ0 multiplied by 1 minus δ. The perturbation coefficient δ is between 0.003 and 0.008, and in this embodiment, it is 0.005. The unit window energy deviation introduced by this perturbation coding amount is limited to a preset quality influence threshold. In this embodiment, the energy deviation threshold is set to 0.5% and the mapping table between energy deviation and melt depth fluctuation is established during the factory calibration stage to confirm that the melt depth standard deviation increment does not exceed 0.02mm within the 0.5% range.

[0066] In step S7, the local control terminal executes the laser drive control waveform and drives the laser output in hard real-time mode, and continuously monitors the credibility score C and the network controllability index En. When C is detected to be lower than C1 or En is lower than 0.35, it enters the autonomous degradation mode. In the autonomous degradation mode, the safe waveform library is forcibly selected and Pmax and Kmax are converged to the strong convergence boundary and the remote side's access to extended waveforms is frozen. The exit of the autonomous degradation mode adopts a minimum hold window mechanism to suppress jitter switching. In this embodiment, the minimum hold window number Nhold is set to 3, that is, only when C is continuously higher than C2 and En is continuously not lower than 0.60 and the platform's credible evidence Ec is 1.00 for 3 consecutive sampling windows, it is allowed to migrate from strong convergence to medium convergence and then to weak convergence.

[0067] In step S8, the remote management terminal receives the returned multimodal operation data and detects the physical watermark sequence to verify the execution consistency of the laser drive control waveform, and at the same time forms a traceable audit record;

[0068] For example, the remote management terminal uses the reflected light data R as the watermark carrier signal. It segments the returned R according to the sampling window Tw and calculates the symbol energy feature sequence Vm within each segment according to the symbol period Tc. Vm is the bandpass energy or root mean square value of R within each symbol time slice. Vm is normalized to obtain Vmn, and then correlated with the reference watermark sequence Wref to obtain the watermark confidence ρ. The correlation matching can be achieved using Pearson correlation coefficient or equivalent centralized correlation calculation. A watermark confidence threshold ρth is set; in this embodiment, ρth is 0.65. When ρ is lower than ρth, it is determined that there is insufficient execution consistency and a strong convergence lock is triggered. The lock is only released and the tiered unlocking path is entered when ρ continuously satisfies ρth and meets the Nhold window number. Simultaneously, the remote management terminal writes the timestamp, Ep, Ec, En, C, State, Pmax, Kmax, autonomous degradation trigger flag, ρ, and the judgment result of each window into the audit record, and calculates a hash chain for the audit record to prevent tampering, thereby achieving post-event traceability and responsibility determination.

[0069] As can be seen from the above implementation, process credible evidence, platform credible evidence and network credible evidence are uniformly integrated into a credibility score and further solidified into an authorization status word, thereby incorporating the available waveform library subset, energy boundary and power change rate boundary into the same constraint chain.

[0070] Meanwhile, the detectability of physical watermarks is used as quantitative evidence of execution consistency and directly participates in the access control and step-by-step unlocking path, thus elevating remote management from traditional alarms and logs to verifiable closed-loop constraints.

[0071] like Figure 1-5 As shown, the cross-modal physical consistency residuals are obtained and used to form credible evidence of the process in the following manner:

[0072] After performing time alignment and sampling window constraint processing on the laser output power data and back-reflection data in the process feature vector, a back-reflection prediction value is generated based on a preset power-back-reflection coupling constraint relationship. The deviation between the back-reflection data and the back-reflection prediction value is used as the instantaneous residual. The instantaneous residual is further recursively accumulated to form a residual trend quantity. The residual trend quantity is then mapped to the process credibility evidence, such that the process credibility evidence monotonically decreases as the residual trend quantity increases. The process credibility evidence is used as a leading quantity for the convergence constraint of the control authority set in the fusion calculation of the credibility score, thereby triggering the entry condition of the autonomous degradation mode in advance.

[0073] In this embodiment: the local control terminal performs unified time base alignment on the power data P and the back-reflected light data R within each sampling window Tw and limits the sampling window boundary. Specifically, the two data streams are resampled to the same time grid point, and transition segments of no more than 5ms at the beginning and end of the window are removed to reduce the impact of trigger transients on the residual. Subsequently, a back-reflected light prediction value Rhat is generated using a preset power back-reflected light coupling constraint relationship. In this embodiment, the coupling constraint relationship is implemented using a polynomial coupling function, such that Rhat is equal to the square of a0 plus a1 multiplied by P plus a2 multiplied by P, where a0, a1, and a2 are coupling parameters obtained from factory calibration. During calibration, the power is scanned step by step under rated operating conditions, and back-reflected light is collected synchronously and fitted with least squares. The instantaneous residual rt is calculated within the window, where rt is the absolute value of the difference between the measured back-reflected light value R and the back-reflected light prediction value Rhat. The average value of rt within the window is taken to obtain the window residual rbar. The window residual rbar is recursively accumulated to form the residual. The trend quantity Qt is recursively calculated as follows: Qt equals λ multiplied by the previous window Qt plus 1 minus λ multiplied by the current window rbar. The forgetting factor λ ranges from 0.85 to 0.98, and in this embodiment, it is 0.92 to achieve smooth suppression of short-term spikes. The residual trend quantity Qt is mapped to process credibility evidence Ep. The mapping is implemented using a monotonically decreasing function, such that Ep equals exp(-Qt) divided by σp. σp is a scale parameter determined by the 95th quantile of Qt under normal operating conditions during factory calibration. In this embodiment, σp is 0.05. As Qt increases, Ep monotonically decreases and participates in the convergence of the control authority set as a leading quantity in the credibility score fusion calculation. This allows the autonomous degradation mode entry condition to be triggered preferentially when Ep drops below the threshold Ep1. Ep1 ranges from 0.55 to 0.70, and in this embodiment, it is 0.60. Thus, the allowable waveform set can be converged and the energy input boundary limited in advance through the residual trend quantity in the early stages of power back-back coupling mismatch or back-back anomaly.

[0074] like Figure 1-5 As shown, the power back-reflection coupling function is an updatable coupling function, and the update of the coupling parameters is only allowed to be triggered under the condition that the platform's credible evidence meets the preset credible baseline and the verification result of the physical watermark sequence based on the back-transmitted multimodal operation data of the remote management terminal is consistent with the execution. After being allowed to be triggered, the spectral drift index is extracted from the spectral data within the same sampling window as the residual trend amount, and the spectral drift index is written into the update objective function of the coupling parameters to solve the updated coupling parameters. The back-reflection prediction value is corrected using the updated coupling parameters to obtain the corrected residual trend amount and update the process credible evidence accordingly, so that the sensitivity of the process credible evidence decreases when the spectral drift index increases and the control authority set is further converged.

[0075] When the platform's trusted evidence does not meet the preset trusted baseline or the physical watermark sequence verification result does not indicate consistent execution, the coupling parameter is locked and the process trusted evidence is updated according to the monotonically decreasing rule to force convergence of the control authority set and trigger the autonomous degradation mode.

[0076] In this embodiment, the local control terminal defines the coupling parameter θ as an updatable parameter group based on the coupling function. θ includes a0, a1, and a2. At the end of each sampling window Tw, an update permission determination is performed before deciding whether to enter the parameter update process. The update permission determination is established when the following simultaneous conditions are met: First, the platform's trusted evidence Ec satisfies the preset trusted baseline. Specifically, the remote management terminal completes the signature verification of the proof token within the most recent proof refresh cycle Tr, and the metric m is consistent with the whitelist, making Ec equal to 1.00. Second, the physical watermark sequence verification result... The result is characterized as consistent execution, specifically, the watermark confidence ρ obtained by the remote management terminal within the most recent Nhold sampling windows is not lower than the threshold ρth. In this embodiment, ρth is 0.65 and Nhold is 3. When the update permission determination is not valid, the local control terminal locks θ and updates the process credible evidence Ep according to the monotonically decreasing rule. For example, Ep is set to the smaller of minEp and the previous window Ep, where minEp is between 0.20 and 0.40 and is 0.30 in this embodiment, thereby forcibly driving the convergence of the control permission set and triggering the autonomous degradation mode entry condition.

[0077] When the update permission determination is successful, the local control terminal extracts the spectral drift index Δλ from the spectral data S within the same sampling window Tw to participate in the coupling parameter update. Δλ is defined as the offset of the spectral peak wavelength λpeak relative to the reference wavelength λref within the window. Δλ equals λpeak minus λref, where λref is the average peak wavelength obtained under normal operating conditions during the factory calibration phase and is stored. The local control terminal constructs the coupling parameter update objective function J, which is equal to the weighted residual sum of squares within the window. The weighted residual sum of squares is the sum of wk multiplied by Rk minus Rhatk over each time grid point k, where Rhatk is calculated from the current θ and Pk, and the weight wk is generated by the spectral drift index. For example, wk equals 1 plus γ multiplied by the absolute value Δλ divided by Δλref, where γ ranges from 0.5 to 2.0 in this embodiment. The value is set to 1.0, and Δλref is set to 0.2nm to 1.0nm, with 0.5nm in this embodiment. This amplifies the contribution of the residual to the objective function when the spectral drift increases, thereby improving the sensitivity to mismatch conditions. The local control terminal minimizes the objective function J to obtain the updated θ. This embodiment uses recursive least squares or an equivalent online least squares solution. To ensure the stability of the update, an upper limit is applied to the update step size of the parameter change. For example, the L2 norm of the parameter increment in each window is limited to no more than η, where η is set to 0.01 to 0.10, with 0.05 in this embodiment. The back-reflection prediction value is recalculated using the updated θ to obtain the corrected residual trend Qt and the credible evidence Ep is updated accordingly. When Δλ increases, the sensitivity of Ep to decrease due to the amplification of wk increases, thereby further converging the control authority set.

[0078] In this embodiment, the typical parameter setting range is as follows: sampling window Tw is 250ms, proof refresh period Tr is 2s, watermark threshold ρth is 0.65, minimum hold window number Nhold is 3, Δλref is 0.5nm, γ is 1.0, and parameter step size upper limit η is 0.05. Furthermore, when Ec does not meet the trusted baseline or ρ is lower than ρth, θ is immediately locked and Ep is updated according to the monotonically decreasing rule. This ensures that parameter updates are only allowed when platform trust and execution consistency are simultaneously met, thereby avoiding the introduction of erroneous parameter updates when there is a risk of tampering or execution deviation in the control link and improving the security, controllability, and process stability of remote control.

[0079] like Figure 1-5As shown, the mapping from the confidence score to the control permission set is implemented using a step-by-step convergence segmentation rule. The segmentation rule sets a first confidence threshold and a second confidence threshold, with the first confidence threshold being less than the second confidence threshold. The permission status word is determined according to the comparison result between the confidence score and the first confidence threshold and the second confidence threshold. The permission status word is used to uniquely determine the set of allowed control modes and the corresponding set of available control waveforms, and simultaneously determine the energy input boundary and the power change rate boundary.

[0080] When the confidence score is not higher than the first confidence threshold, the permission status word is placed in a strong convergence state so that the available set of control waveforms converges to a safe waveform set and the energy input boundary and power change rate boundary converge to a preset minimum boundary and the autonomous degradation mode is forcibly maintained.

[0081] When the confidence score is higher than the first confidence threshold and not higher than the second confidence threshold, the permission status word is placed in the mid-convergence state to allow only limited step size updates to the control waveform parameters under the premise that the autonomous degradation mode is lifted, and to keep the power change rate boundary from exceeding the upper limit corresponding to the network controllability index.

[0082] When the confidence score is higher than the second confidence threshold, the permission status word is placed into a weak convergence state to unlock the call to the extended waveform set in the control waveform library and update the energy input boundary and power change rate boundary according to the monotonically relaxed rule, so that the control permission set is unlocked step by step as the confidence score increases and converges step by step as the confidence score decreases.

[0083] In this embodiment, to achieve a step-by-step convergence mapping from the confidence score to the control permission set, the local control terminal calculates the confidence score C and generates the permission state word State at the end of each sampling window Tw. The confidence score C is obtained using the fusion method and is clipped to the range of 0 to 1. A first confidence threshold C1 and a second confidence threshold C2 are set, with C1 being less than C2. In this embodiment, C1 is set to 0.45 to 0.55 and 0.45, and C2 is set to 0.70 to 0.85 and 0.75. The permission state word State is determined according to the following segmentation rules: when C is not higher than C1, State is set to a strong convergence state; when C is higher than C1 but not higher than C2, State is set to a medium convergence state; when C is higher than C2, State is set to a weak convergence state. The permission state word State is used to uniquely determine the allowed control mode set and the corresponding available control waveform set, and simultaneously determine the energy input boundary Pmax and the power change rate boundary Kmax. The control modes include at least the local autonomous mode and the remote collaborative mode, and the available control waveform set includes at least the safe waveform set and the extended waveform set.

[0084] In the strong convergence state, the local control terminal only allows the use of the safety waveform set and forces the local autonomous mode to be maintained. The energy input boundary Pmax is converged to the range of 0.60 to 0.75 of the rated power and is set to 0.70. At the same time, the power change rate boundary Kmax is converged to the range of the rated power change per second not exceeding 0.10 to 0.20 and is set to 0.15. In this embodiment, the safety waveform set is formed by limiting the pulse width τ and the repetition frequency F, where τ is 200ns to 600ns and F is 5kHz to 20kHz. The rise slope is limited to not exceeding the preset slope upper limit to reduce the risk of thermal shock.

[0085] In the convergence state, the local control terminal is allowed to enter the remote cooperative mode after deactivating the local autonomous mode, but only limited step size updates are allowed for the control waveform parameters. The limited step size update means that the relative change of the pulse width τ within each sampling window does not exceed 1% and the relative change of the repetition frequency F does not exceed 1%. At the same time, Pmax is set to the range of 0.75 to 0.85 of the rated power and is taken as 0.80, and Kmax is limited to not exceeding the upper limit Knet corresponding to the network controllability index, where Knet is monotonically determined by the network credible evidence En. In this embodiment, Knet is equal to 0.10 plus 0.30 multiplied by En, so that Knet decreases when En decreases, thereby further limiting the power change rate.

[0086] In the weak convergence state, the local control terminal allows the invocation of the extended waveform set and allows the extended invocation of the waveform library in the remote collaborative mode. At the same time, Pmax and Kmax are updated according to the monotonically relaxed rule. The monotonically relaxed rule means that when the state transitions from medium convergence to weak convergence, Pmax is increased to 0.90 to 1.00 of the rated power and is set to 1.00, and Kmax is increased to the rated power change of no more than 0.30 to 0.50 per second and is set to 0.40. In this embodiment, the extended waveform set allows the addition of a preamble pulse and a trailing annealing pulse on the basis of the safety waveform set, and allows the rise slope to be relaxed under the condition that it does not exceed the preset upper limit.

[0087] Through the above segmented threshold and parameter mapping, this implementation method enables the control permission set to be unlocked step by step as the confidence score increases and converged step by step as the confidence score decreases, and makes the energy input boundary and power change rate boundary have a clear reproducible numerical range under different permission states.

[0088] like Figure 1-5As shown, the update of the permission status word follows a state transition rule with hysteresis constraints. The state transition rule includes maintaining a state holding register corresponding to the permission status word on the local control terminal and setting a minimum holding window number for each state transition. The transition of the permission status word from a weak convergence state to a medium convergence state and from a medium convergence state to a strong convergence state is configured as an immediate effective transition. The immediate effective transition is triggered by the platform's trusted evidence not meeting the preset trusted baseline, the physical watermark sequence verification result not representing execution consistency, or the trust score not being higher than the first trust threshold. After triggering, the state holding register is locked.

[0089] Furthermore, the migration of the permission status word from a strong convergence state to a medium convergence state is only allowed to be triggered when the platform's trusted evidence continuously meets the preset trusted baseline and the physical watermark sequence verification results are all characterized as consistent within a continuous first minimum hold window. The migration of the permission status word from a medium convergence state to a weak convergence state is only allowed to be triggered when, after the aforementioned triggering is completed, the network controllability index does not deteriorate within a continuous second minimum hold window and the trust score is continuously higher than the second trust threshold. Thus, permission unlocking must go through a step-by-step release path from strong convergence to medium convergence and then to weak convergence and is subject to the prior gating constraints of platform trust and watermark consistency.

[0090] In this embodiment, to ensure that the permission state word update rule with hysteresis constraint can be implemented, the local control terminal obtains the candidate permission state word Statecand based on the segmentation rule at the end of each sampling window Tw, and maintains the current permission state word Statecur and its corresponding state holding register H and minimum holding window number Nmin on the local control terminal, where H is used to record the consecutive window count since the most recent state transition was completed; in this embodiment, Tw is taken as 250ms, Nmin is taken as 2 to 8 and is taken as 3, and H is updated by incrementing 1 in each window until H is cleared to zero after a state transition occurs.

[0091] When the candidate permission status word Statecand is detected to point to a stronger convergence direction, an immediate migration is executed. The stronger convergence direction includes migration from a weak convergence state to a medium convergence state and migration from a medium convergence state to a strong convergence state. The immediate migration is triggered when any of the following conditions are met: the platform's trusted evidence Ec does not meet the preset trusted baseline, or the physical watermark sequence verification result does not represent execution consistency, or the trust score C is not higher than the first trust threshold C1. When the immediate migration is established, the local control terminal immediately updates Statecur to Statecand and locks the state to maintain the register H. The locking means that during the locking period, migration to a weaker convergence direction is not allowed and H is fixed to 0 until the release condition is met, thereby ensuring that permission convergence takes effect immediately when a risk occurs and is not offset by short-term fluctuations.

[0092] When the candidate permission status word Statecand is detected to point to a weaker convergence direction, migration is only allowed after the window maintenance gate is satisfied. The weaker convergence direction includes migration from a strong convergence state to a medium convergence state and migration from a medium convergence state to a weak convergence state. The allowed triggering condition for migration from a strong convergence state to a medium convergence state is that the platform's trusted evidence Ec continuously satisfies a preset trusted baseline, and the physical watermark sequence verification result is consistently characterized within N consecutive sampling windows, while the trust score C is higher than the first trust threshold C1 within N consecutive sampling windows. The allowed triggering condition for migration from a medium convergence state to a weak convergence state is that the aforementioned allowed triggering condition for migration from a strong convergence state to a medium convergence state is met. Further, the network trust evidence En corresponding to the network controllability index is not degraded within N consecutive sampling windows, and the trust score C is continuously higher than the second trust threshold C2 within N consecutive sampling windows. When the allowable triggering condition for any weaker convergence direction is met, the local control terminal unlocks the state holding register H and updates Statecur to Statecand, while clearing H and restarting the count. This ensures that unlocking permissions must go through a step-by-step release path from strong convergence to medium convergence and then to weak convergence, and is subject to the prior gating constraint of platform trust and watermark consistency. Hysteresis is formed by the minimum holding window number Nmin to avoid frequent switching caused by short-term network recovery or noise fluctuations.

[0093] like Figure 1-5 As shown, the physical watermark sequence is determined by the remote management terminal based on the proof random number generation seed bound to the platform's trusted evidence and synchronously forms a reference watermark sequence. The reference watermark sequence is mapped at the local control terminal to a perturbation coding amount of the laser drive control waveform to perform window-allocated micro-modulation on the pulse width or pulse repetition frequency of the pulse sequence, so that the energy deviation introduced by the perturbation coding amount is always limited to the preset quality influence threshold.

[0094] After receiving the returned reflective data, the remote management terminal performs synchronous alignment on the reflective data under the same time reference as the sampling window, and extracts watermark features after alignment. Then, it obtains the watermark confidence by performing correlation matching with the reference watermark sequence, and uses the watermark confidence as a quantitative representation of the verification result of the physical watermark sequence. When the watermark confidence is lower than a preset threshold, the process credible evidence is updated according to the monotonically decreasing rule, and the permission status word is forcibly migrated to a strong convergence state and the state is locked to maintain the register. Only when the watermark confidence continuously meets the preset threshold and meets the minimum holding window number is the lock allowed to be released and enter the step-by-step unlocking path, so that the detectability of the physical watermark sequence becomes a necessary condition for controlling permission unlocking and forms a closed-loop constraint relationship with the confidence score.

[0095] In this embodiment, in order to ensure that the generation, embedding, detection of the physical watermark sequence and the closed-loop constraint of the watermark confidence gating permission status word can be realized, this embodiment uses the reflected light signal as the watermark carrier and pulse width perturbation as the watermark embedding method.

[0096] Within each remote proof refresh cycle Tr, the remote management terminal obtains the proof random number nonce corresponding to the platform's trusted evidence Ec, and concatenates the nonce with the device identifier ID, then hashes it to obtain the generation seed. The seed is used to generate the reference watermark sequence Wref. The reference watermark sequence Wref is a binary sequence of length L. In this embodiment, L is set to 64 and corresponds one-to-one with the sampling window Tw. The symbol period Tc is equal to Tw divided by L. In this embodiment, Tw is set to 250ms, so Tc is approximately 3.9ms. The remote management terminal sends Wref and the window sequence number together to the local control terminal and caches them on the remote side for subsequent verification.

[0097] After receiving the reference watermark sequence Wref, the local control terminal maps it to a perturbation coding amount Δτ for the laser drive control waveform. This ensures that the pulse width within each symbol period is bidirectionally perturbed near the reference pulse width τ0. In this implementation, the perturbation amplitude is determined by Δτ being equal to δ multiplied by τ0, with δ ranging from 0.003 to 0.008 and set to 0.005. When the symbol is 1, the pulse width within that symbol period is τ0 multiplied by 1 plus δ; when the symbol is 0, the pulse width within that symbol period is τ0 multiplied by 1 minus δ. The pulse repetition frequency F is kept constant or maintained at a constant value. The Kmax boundary is adjusted slowly. To ensure that the impact on process quality is controllable, the local control terminal calculates the relative energy deviation ΔE introduced by the watermark in each sampling window. ΔE is the absolute value of the difference between the waveform energy after perturbation and the reference waveform energy without perturbation in the window divided by the reference waveform energy. ΔE is limited to not exceeding the preset quality impact threshold Emax. In this embodiment, Emax is taken as 0.2% to 0.8% and 0.5%. When it exceeds the threshold, δ is automatically converged to the maximum allowable value that meets Emax and the convergence event is recorded.

[0098] After receiving the returned reflective data R, the remote management terminal segments the reflective data according to the time base consistent with the sampling window Tw, and divides each window into L symbol segments according to the symbol period Tc. For each symbol segment, the watermark feature value Vm is calculated. In this embodiment, the watermark feature value Vm is taken as the root mean square value or bandpass energy value of R in the symbol segment. The obtained L-dimensional feature sequence V is then normalized by removing the mean and variance to obtain Vn. The remote management terminal performs correlation matching between Vn and the reference watermark sequence Wref to obtain the watermark confidence ρ. In this embodiment, a centered correlation coefficient is used to achieve ρ, and the watermark confidence threshold ρth is set to 0.55 to 0.75 and is set to 0.65. When ρ is not lower than ρth, the physical watermark sequence verification result is determined to be consistent with the execution. When ρ is lower than ρth, the physical watermark sequence verification result is determined not to be consistent with the execution.

[0099] When ρ is lower than ρth, the remote management terminal sends a strong convergence constraint flag to the local control terminal and triggers the monotonically decreasing rule update process trusted evidence Ep on the local control terminal. In this embodiment, Ep is updated to min the previous window Ep and 0.30, and the permission status word is forcibly migrated to the strong convergence state and the step-by-step unlocking path is locked. When ρ continuously satisfies ρth and continuously satisfies the minimum holding window number Nmin, the local control terminal allows the lock to be released and enters the step-by-step unlocking path from strong convergence to medium convergence and then to weak convergence, where Nmin is 2 to 8 and is 3 in this embodiment. Through the above embedding, detection and gating mechanism, the detectability of the physical watermark sequence becomes a necessary condition for controlling the unlocking of permissions, and the watermark confidence ρ and the confidence score C together form a reproducible closed-loop constraint relationship, thereby ensuring that the execution consistency in the remote management scenario is verifiable and the release of permissions is controllable.

[0100] A multi-mode driven intelligent control and remote management system for lasers, such as Figure 1-5 As shown, it includes a laser, a local control terminal, a remote management terminal, and a communication link for connecting the local control terminal and the remote management terminal;

[0101] The local control terminal is connected to the laser and configured to collect at least two types of operating data in different modes during laser operation to form a process feature vector. The operating data in different modes includes at least two of the following: laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Simultaneously, the network transmission status data of the communication link is acquired. The network transmission status data includes at least delay parameters and jitter parameters.

[0102] The local control terminal is further configured to construct cross-modal physical consistency residuals based on the process feature vector to output process credible evidence, and to perform remote proof or integrity verification on the local control terminal to output platform credible evidence, while calculating network controllability indicators based on the network transmission status data to output network credible evidence.

[0103] The local control terminal is further configured to fuse the process credible evidence, the platform credible evidence, and the network credible evidence to calculate a credibility score and generate a control permission set accordingly. The control permission set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary.

[0104] The local control terminal is further configured to generate a laser-driven control waveform under the constraints of the control permission set and embed a physical watermark sequence in the laser-driven control waveform, and execute the laser-driven control waveform in hard real-time mode to drive the laser output. At the same time, when the credibility score is lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters an autonomous degradation mode and restricts the control waveform to a safe waveform in the waveform library subset and converges the energy input boundary or the power change rate boundary.

[0105] The remote management terminal is configured to receive multimodal operation data returned by the local control terminal and detect the physical watermark sequence based on the returned data to verify the execution consistency of the laser-driven control waveform and form a traceable remote management audit record. The local control terminal is configured to solidify the mapping result of the confidence score into an authorization status word and use the authorization status word to uniquely define the waveform library subset and the energy input boundary or the power change rate boundary. The authorization status word is determined by a first confidence threshold and a second confidence threshold, and the first confidence threshold is less than the second confidence threshold. By setting a minimum holding window number, a step-by-step unlocking path with hysteresis constraints is formed so that the authorization status word takes effect immediately when it migrates to a stronger convergence direction, and is only allowed to be triggered when it migrates to a weaker convergence direction under the condition that the platform's credible evidence continuously meets the preset credible baseline, the physical watermark sequence verification result continuously represents execution consistency within the continuous minimum holding window number, and the network controllability index does not deteriorate within the corresponding window.

[0106] The remote management terminal is further configured to determine the reference watermark sequence based on the proof random number generated seed bound to the platform's trusted evidence, and to determine the physical watermark sequence in consistency with the local control terminal. The watermark confidence is obtained based on the correlation matching between the returned reflected light data and the reference watermark sequence, and the watermark confidence is used to quantify and characterize the verification result of the physical watermark sequence. When the watermark confidence is lower than a preset threshold, the local control terminal forcibly migrates the permission status word to a strong convergence state and locks the step-by-step unlocking path until the watermark confidence continuously meets the minimum holding window number.

[0107] In this embodiment, the laser intelligent control and remote management system based on multi-mode driving includes a laser, a local control terminal, a remote management terminal, and a communication link. The local control terminal is connected to the laser's power sampling channel, retroreflection sampling channel, spectral sampling channel, temperature sampling channel, and current or voltage ripple sampling channel, and synchronously acquires data using a unified time base. For example, the power sampling frequency is 10kHz, the retroreflection sampling frequency is 50kHz, the spectral frame rate is 200Hz, the temperature sampling frequency is 100Hz, the ripple sampling frequency is 20kHz, and the sampling window Tw is 250ms. The local control terminal resamples the data of each mode to a unified time grid within Tw to form a process feature vector and synchronously acquires network transmission status data. Network transmission status data includes at least latency and jitter. Latency is obtained by carrying the sending and receiving timestamps in the message, and jitter is obtained by statistically analyzing the latency difference between adjacent messages. The local control unit constructs a cross-modal physical consistency residual within each sampling window and outputs process credible evidence Ep. For example, a power back-reflection coupling function Rhat is established, which is equal to a0 + a1 × P + a2 × P squared to obtain the back-reflection prediction value, and the residual trend quantity Qt is calculated. Then, Qt is mapped monotonically decreasingly to generate Ep. Simultaneously, the local control unit performs remote proof or integrity verification to output platform credible evidence Ec. For example, the remote management unit issues a challenge random number nonce, and the local control unit returns a proof token containing a metric value m and a signature. The terminal verifies the signature and compares m with the trusted baseline. If verification passes, Ec is set to 1.00; if verification fails, Ec is set to 0.20. The local control terminal calculates the network controllability index based on network latency and jitter and outputs the network trusted evidence En. For example, the network risk quantity is constructed using the 95th percentile of latency and the root mean square of jitter, and En is obtained using a monotonic function. The local control terminal fuses Ep, Ec, and En to obtain a trust score C and maps C to a control permission set. The control permission set must at least include allowed control modes, a subset of waveform libraries, energy input boundary Pmax, and power change rate boundary Kmax. For example, C is trimmed to 0 to 1 and then segmented into permission state words State using thresholds C1 and C2, where C1 is 0.45, C2 is 0.75, and C is not... When C is above C1, the State is in strong convergence and only the safe waveform library is allowed, and Pmax converges to 0.70 of the rated power and Kmax converges to 0.15. When C is above C1 but not above C2, the State is in medium convergence and only limited step size updates are allowed, and Pmax is set to 0.80 and Kmax does not exceed the upper limit determined monotonically by En. When C is above C2, the State is in weak convergence and the expanded waveform library is allowed, and Pmax is relaxed to 1.00 and Kmax is relaxed to 0.40. The local control terminal generates the laser drive control waveform accordingly and drives the laser output in hard real-time mode. When C is below C1 or En is below the preset threshold, it enters the autonomous degradation mode and forcibly selects the safe waveform library while converging Pmax and Kmax.

[0108] Based on the above, the local control terminal solidifies the mapping result of the credibility score into an authorization status word and maintains the status retention register H and the minimum retention window number Nmin to realize a step-by-step unlocking path with hysteresis constraints. For example, Nmin is 3 and H is incremented by 1 for each window. When a risk occurs, the migration to a stronger convergence direction takes effect immediately and locks the unlocking path. The risk includes at least the platform's credible evidence Ec not meeting the credibility baseline, the physical watermark verification result not representing execution consistency, or C not being higher than C1. The migration to a weaker convergence direction is only allowed to be triggered when Ec is continuously 1.00 for N consecutive windows, the physical watermark verification result continuously represents execution consistency, the network credible evidence En does not continuously deteriorate, and C is continuously higher than the corresponding threshold. Thus, unlocking must go through a step-by-step release from strong convergence to medium convergence and then to weak convergence. In addition to completing remote proof verification, the remote management terminal also generates a reference watermark sequence Wref based on the proof random number generated by the seed bound to the platform's credible evidence and determines the physical watermark sequence in consistency with the local control terminal. For example, each Tw generates a binary sequence with a length L of 64 and a symbol period Tc equal to Tw divided by L is approximately 3.9ms. The local control terminal maps Wref to a perturbation coding quantity and performs micro-modulation on the pulse width or pulse repetition frequency according to the symbol period, limiting the relative energy deviation introduced by this to no more than a preset quality influence threshold Emax. For example, Emax is set to 0.5% and the perturbation coefficient δ is set to 0.005. After receiving the reflected light data returned by the local control terminal, the remote management terminal segments it according to Tw and Tc and extracts the watermark feature sequence. It performs correlation matching with Wref to obtain the watermark confidence ρ, and compares ρ with the threshold ρth to form a physical watermark confidence score. The watermark sequence verification result, in example, ρth is 0.65. When ρ is lower than ρth, the remote management terminal outputs strong convergence constraint information to the local control terminal, forcibly migrating the permission status word to strong convergence and locking the step-by-step unlocking path. Only after ρ continuously satisfies ρth and satisfies Nmin is the lock released and the step-by-step unlocking path allowed. At the same time, the remote management terminal records the timestamp, Ep, Ec, En, C, State, autonomous downgrade trigger event, and ρ and the judgment result of each window to form an audit record, so as to achieve traceable and verifiable execution consistency under remote management.

[0109] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0110] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for intelligent control and remote management of lasers based on multi-mode driving, characterized in that, Includes the following steps: S1: During the operation of the laser, at least two types of operation data of different modes are collected and process feature vectors are formed. The operation data of different modes include at least two types of laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Network transmission status data is collected simultaneously. The network transmission status data includes at least time delay parameters and jitter parameters. S2: Construct cross-modal physical consistency residuals based on the process feature vectors. The cross-modal physical consistency residuals are used to characterize the degree of deviation of different modal operation data under preset physical constraints and output process credibility evidence. S3: Perform remote verification or integrity check on the local control terminal of the laser to output platform trust evidence, which is used to characterize the degree to which the hardware and software status of the local control terminal meets the preset trust baseline; S4: Calculate the network controllability index based on the network transmission status data and output network trust evidence. The network trust evidence is used to characterize the degree of influence of the remote control link on the stability of closed-loop control. S5: The process credibility evidence, the platform credibility evidence, and the network credibility evidence are fused together to obtain a credibility score, and the credibility score is mapped to a control authority set. The control authority set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary. S6: Generate control waveform parameters and form a laser-driven control waveform under the constraints of the control authority set, and embed a physical watermark sequence in the laser-driven control waveform. The physical watermark sequence is achieved by perturbation of the pulse sequence in the time domain or frequency domain and satisfies the constraint of not exceeding the preset quality influence threshold.

2. The laser intelligent control and remote management method based on multi-mode driving according to claim 1, characterized in that, It also includes the following steps: S7: The local control terminal executes the laser drive control waveform in hard real-time mode to drive the laser output, and continuously monitors the confidence score and the network controllability index. When the confidence score is detected to be lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters the autonomous degradation mode and restricts the control waveform to the safe waveform in the waveform library subset, while converging the energy input boundary or the power change rate boundary. S8: The remote management terminal receives the returned multimodal operation data and detects the physical watermark sequence based on the returned data to verify the execution consistency of the laser drive control waveform. At the same time, the credibility score, the control permission set mapping result, the autonomous degradation trigger event, and the physical watermark sequence verification result are recorded to form a traceable remote management audit record.

3. The laser intelligent control and remote management method based on multi-mode driving according to claim 2, characterized in that, The cross-modal physical consistency residuals are obtained and used to form credible evidence of the process in the following manner: After performing time alignment and sampling window constraint processing on the laser output power data and back-reflection data in the process feature vector, a back-reflection prediction value is generated based on a preset power-back-reflection coupling constraint relationship. The deviation between the back-reflection data and the back-reflection prediction value is used as the instantaneous residual. The instantaneous residual is further recursively accumulated to form a residual trend quantity. The residual trend quantity is then mapped to the process credibility evidence, such that the process credibility evidence monotonically decreases as the residual trend quantity increases. The process credibility evidence is used as a leading quantity for the convergence constraint of the control authority set in the fusion calculation of the credibility score, thereby triggering the entry condition of the autonomous degradation mode in advance.

4. The intelligent control and remote management method for lasers based on multi-mode driving according to claim 2, characterized in that, The power back-reflection coupling function is an updatable coupling function, and the update of the coupling parameters is only allowed to be triggered under the condition that the platform's credible evidence meets the preset credible baseline and the verification result of the physical watermark sequence based on the back-transmitted multimodal operation data of the remote management terminal is consistent with the execution. After being allowed to be triggered, the spectral drift index is extracted from the spectral data within the same sampling window as the residual trend amount, and the spectral drift index is written into the update objective function of the coupling parameters to solve the updated coupling parameters. The updated coupling parameters are used to correct the back-reflection prediction value to obtain the corrected residual trend amount and update the process credible evidence accordingly, so that the sensitivity of the process credible evidence decreases when the spectral drift index increases and the control authority set is further converged. When the platform's trusted evidence does not meet the preset trusted baseline or the physical watermark sequence verification result does not indicate consistent execution, the coupling parameters are locked and the process trusted evidence is updated according to the monotonically decreasing rule to force convergence of the control authority set and trigger the autonomous degradation mode.

5. The laser intelligent control and remote management method based on multi-mode driving according to claim 4, characterized in that, The mapping from the confidence score to the control permission set is implemented using a stepwise convergence segmentation rule. The segmentation rule sets a first confidence threshold and a second confidence threshold, with the first confidence threshold being less than the second confidence threshold. The permission status word is determined according to the comparison result between the confidence score and the first confidence threshold and the second confidence threshold. The permission status word is used to uniquely determine the set of allowed control modes and the corresponding set of available control waveforms, and simultaneously determine the energy input boundary and the power change rate boundary. When the confidence score is not higher than the first confidence threshold, the permission status word is placed in a strong convergence state so that the available set of control waveforms converges to a safe waveform set and the energy input boundary and power change rate boundary converge to a preset minimum boundary and the autonomous degradation mode is forcibly maintained. When the confidence score is higher than the first confidence threshold and not higher than the second confidence threshold, the permission status word is placed in the mid-convergence state to allow only limited step size updates to the control waveform parameters under the premise that the autonomous degradation mode is lifted, and to keep the power change rate boundary from exceeding the upper limit corresponding to the network controllability index. When the confidence score is higher than the second confidence threshold, the permission status word is placed into a weak convergence state to unlock the call to the extended waveform set in the control waveform library and update the energy input boundary and power change rate boundary according to the monotonically relaxed rule, so that the control permission set is unlocked step by step as the confidence score increases and converges step by step as the confidence score decreases.

6. The intelligent control and remote management method for lasers based on multi-mode driving according to claim 5, characterized in that, The update of the permission status word follows a state transition rule with hysteresis constraints. The state transition rule includes maintaining a state holding register corresponding to the permission status word on the local control terminal and setting a minimum holding window number for each state transition. The transition of the permission status word from a weak convergence state to a medium convergence state and from a medium convergence state to a strong convergence state is configured as an immediate effective transition. The immediate effective transition is triggered by the platform's trusted evidence not meeting the preset trusted baseline, the physical watermark sequence verification result not representing execution consistency, or the trust score not being higher than the first trust threshold. After triggering, the state holding register is locked. Furthermore, the migration of the permission status word from a strongly converged state to a moderately converged state is only allowed to be triggered when the platform's trusted evidence continuously meets the preset trusted baseline and the physical watermark sequence verification results are consistently performed within a continuous first minimum hold window. The migration of the permission status word from a moderately converged state to a weakly converged state is only allowed to be triggered after the aforementioned triggering is completed, and further when the network controllability index does not deteriorate within a continuous second minimum hold window and the trust score is continuously higher than the second trust threshold. Thus, permission unlocking must go through a step-by-step release path from strongly converged to moderately converged and then to weakly converged, and is subject to the prior gating constraints of platform trust and watermark consistency.

7. The intelligent control and remote management method for lasers based on multi-mode driving according to claim 6, characterized in that, The physical watermark sequence is determined by the remote management terminal based on the proof random number generation seed bound to the platform's credible evidence and synchronously forms a reference watermark sequence. The reference watermark sequence is mapped at the local control terminal to a perturbation coding amount of the laser drive control waveform to perform window-allocated micro-modulation on the pulse width or pulse repetition frequency of the pulse sequence, so that the energy deviation introduced by the perturbation coding amount is always limited to the preset quality influence threshold. After receiving the returned reflective data, the remote management terminal performs synchronous alignment on the reflective data under a time reference consistent with the sampling window, and extracts watermark features after alignment. Then, it obtains the watermark confidence by performing correlation matching with the reference watermark sequence, and uses the watermark confidence as a quantitative representation of the verification result of the physical watermark sequence. When the watermark confidence is lower than a preset threshold, the process credible evidence is updated according to the monotonically decreasing rule, and the permission status word is forcibly migrated to a strong convergence state and the state is locked to maintain the register. Only when the watermark confidence continuously meets the preset threshold and meets the minimum holding window number is the lock allowed to be released and enter the step-by-step unlocking path, so that the detectability of the physical watermark sequence becomes a necessary condition for controlling permission unlocking and forms a closed-loop constraint relationship with the confidence score.

8. A laser intelligent control and remote management system based on multi-mode driving, the laser intelligent control and remote management method based on multi-mode driving according to any one of claims 1-7, characterized in that, It includes a laser, a local control terminal, a remote management terminal, and a communication link for connecting the local control terminal and the remote management terminal; The local control terminal is connected to the laser and configured to collect at least two types of operating data in different modes during laser operation to form a process feature vector. The operating data in different modes includes at least two of the following: laser output power data, back reflection data, spectral data, temperature data, and current or voltage ripple data. Simultaneously, the network transmission status data of the communication link is acquired. The network transmission status data includes at least delay parameters and jitter parameters. The local control terminal is further configured to construct cross-modal physical consistency residuals based on the process feature vector to output process credible evidence, and to perform remote proof or integrity verification on the local control terminal to output platform credible evidence, while calculating network controllability indicators based on the network transmission status data to output network credible evidence. The local control terminal is further configured to fuse the process credible evidence, the platform credible evidence, and the network credible evidence to obtain a credibility score and generate a control permission set accordingly. The control permission set includes at least a set of allowed control modes, a set of allowed instructions, a set of allowed waveforms or a subset of waveform libraries, and an energy input boundary or a power change rate boundary. The local control terminal is further configured to generate a laser-driven control waveform under the constraints of the control permission set, embed a physical watermark sequence in the laser-driven control waveform, and execute the laser-driven control waveform in hard real-time mode to drive the laser output. At the same time, when the credibility score is lower than the threshold or the network controllability index deteriorates to the threshold condition, it enters an autonomous degradation mode, restricts the control waveform to a safe waveform in the waveform library subset, and converges the energy input boundary or the power change rate boundary.

9. The intelligent control and remote management system for lasers based on multi-mode driving according to claim 8, characterized in that, The local control terminal is configured to solidify the mapping result of the confidence score into an authorization status word and use the authorization status word to uniquely define the waveform library subset and the energy input boundary or the power change rate boundary. The authorization status word is determined by a first confidence threshold and a second confidence threshold, with the first confidence threshold being less than the second confidence threshold. A step-by-step unlocking path with hysteresis constraints is formed by setting a minimum hold window number, so that the authorization status word takes effect immediately when it migrates to a stronger convergence direction, while it is only allowed to be triggered when it migrates to a weaker convergence direction, provided that the platform's credible evidence continuously meets the preset credible baseline, the physical watermark sequence verification result is continuously characterized as consistent within the continuous minimum hold window number, and the network controllability index does not deteriorate within the corresponding window. The remote management terminal is further configured to determine a reference watermark sequence based on a proof random number seed bound to the platform's trusted evidence, and to determine the physical watermark sequence in consistency with the local control terminal. Furthermore, it obtains the watermark confidence level based on the correlation matching between the returned reflected light data and the reference watermark sequence, and uses the watermark confidence level to quantify and characterize the verification result of the physical watermark sequence. This ensures that when the watermark confidence level is lower than a preset threshold, the local control terminal forcibly migrates the permission status word to a strong convergence state and locks the step-by-step unlocking path until the watermark confidence level continuously meets the minimum hold window number.

Images