Spherical mirror multi-working-condition surface shape stability measurement method and system

Abstract

The invention provides a spherical mirror multi-working-condition surface shape stability measuring method and system, and relates to the technical field of optical testing, and the method comprises the steps: constructing a double-layer state machine, monitoring a supporting contact state and an air path state in real time, generating a corresponding state identifier based on load sensing data and light path monitoring data, and determining the stability of a spherical mirror surface shape. Determining topological annealing sequence parameters by using the state identifier, executing topological annealing, and finally executing interference measurement in a stable state and generating stability evaluation index data. According to the method, the problem of measurement uncertainty caused by support topology micro-slip and air path disturbance coupling in the prior art is solved, and the accuracy and reliability of profile stability evaluation of the large-aperture spherical mirror under multiple working conditions are improved.

Description

Technical Field

[0001] This application relates to the field of optical testing technology, and more specifically, to a method and system for measuring the surface stability of spherical mirrors under multiple operating conditions. Background Technology

[0002] In practical applications of large-aperture spherical mirrors, the surface stability assessment of the mirror typically needs to be conducted under different orientations, temperature and humidity conditions, or after long-term static storage to ensure that the spherical mirror can stably meet the application requirements of high-precision optical systems. Current mainstream methods usually involve predicting surface changes under typical operating conditions through finite element simulation, supplemented by interferometry measurements of several finite discrete operating conditions to determine the PV or RMS value of the spherical mirror. However, this static, decentralized assessment method is only suitable for stable single or a few specific operating conditions. For spherical mirror systems with attitude fine-tuning, locking mechanisms, or long-term operation, it does not consider the contact state uncertainties caused by frictional viscosity, micro-slippage, and differences in locking sequence within the support structure. This uncertainty often makes the assessment results difficult to reproduce or interpret, thus seriously affecting the stability and reliability of the optical system in actual operation.

[0003] Furthermore, current evaluation methods often neglect the impact of air disturbances in the long optical path during interferometry on the measurement results. Especially under large-aperture mirror measurement conditions, the path reconstruction problem of air disturbances caused by temperature gradients, air convection, or attitude changes introduces non-mirror-related air disturbance components into the measurement wavefront, further exacerbating measurement uncertainty and non-repeatability. Existing technologies have not adequately distinguished, analyzed, or eliminated the impact of such air disturbances on measurement accuracy, nor have they considered the potential cross-coupling effect between the support structure state and the air disturbance state from a mechanistic perspective. Therefore, overcoming the measurement uncertainties caused by the support topology path dependence effect and air path disturbance coupling when evaluating the surface stability of spherical mirrors under multiple operating conditions, in order to more accurately determine the stable measurement state and obtain accurate and reliable multi-condition surface stability evaluation indicators, has become a key technical problem urgently needing to be solved in this field. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides a method and system for measuring the surface stability of spherical mirrors under multiple operating conditions.

[0005] Firstly, this application provides a method for measuring the surface stability of spherical mirrors under multiple operating conditions, including:

[0006] A two-layer state machine is constructed, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state.

[0007] Collect load sensing data related to the support contact state and optical path monitoring data related to the air path state;

[0008] Based on the load sensing data, a first state identifier representing the support contact state is generated; based on the optical path monitoring data, a second state identifier representing the air path state is generated; based on the first state identifier and the second state identifier, the topology annealing sequence parameters are determined, and the topology annealing sequence is executed.

[0009] During the execution of the topology annealing sequence, the first state identifier and the second state identifier are updated in real time, and an initial state confirmation signal is generated when both simultaneously meet their respective preset stability conditions.

[0010] Based on the initial state confirmation signal, a formal interferometric measurement is triggered to obtain the surface shape data of the spherical mirror under the current working condition. Based on the surface shape data and the corresponding first state identifier and second state identifier, evaluation index data for characterizing the surface shape stability of the spherical mirror under multiple working conditions is generated.

[0011] Optionally, constructing the two-level state machine includes:

[0012] Based on the output data of multiple load sensors of the multi-point support mechanism, the contact micro-state characteristics of each support point are determined, and clustering operation is performed on the contact micro-state characteristics to generate the state identifier of each support point and form the state mapping table of the first state machine.

[0013] Based on the output data of the optical path peripheral monitoring channel and environmental monitoring sensor of the interferometric measurement component, the disturbance characteristics of the air path are determined, and the disturbance characteristics are classified and calculated to generate the state mapping table of the second state machine.

[0014] Establish a cross-state index between the first state machine and the second state machine to identify the coupling mode descriptor between the support contact state and the air path state.

[0015] Optionally, the contact micro-state characteristics are a feature vector composed of the load fluctuation of the load sensor and the displacement response of the support point within a continuous time window.

[0016] The disturbance characteristic is the modal coefficient fluctuation rate after the air path disturbance amount within a continuous time window is projected onto a preset low-order spatial modal basis.

[0017] The state mapping tables of the first state machine and the second state machine respectively record the state switching order and generate path history codes. The cross-state index is established by associating the path history codes of the first state machine and the second state machine.

[0018] Optionally, generating a first state identifier characterizing the support contact state based on the load sensing data includes:

[0019] Within a preset time window, the load average value, load fluctuation amplitude, and load change rate are calculated from the load sensing data of each support point, and the above three quantities are combined as the load feature vector of the support point.

[0020] Each load feature vector is normalized according to the spatial distribution of support points to obtain a relative load feature vector after eliminating the influence of overall temperature drift, and the offset from the historical baseline is calculated in the relative load feature vector.

[0021] The offset is input into a micro-state classification model based on support point partitioning. In the model, the displacement response data or micro-vibration response data of the support point are combined to generate state labels of bearing state, light touch state or departure state.

[0022] The state markers of all support points are combined to form a first state identifier, and the time sequence code of the state switching of each support point is appended to the first state identifier to characterize the path dependency feature of the support contact topology.

[0023] Optionally, generating a second state identifier characterizing the air path state based on the optical path monitoring data includes:

[0024] The air disturbance data collected by the peripheral monitoring channel of the interferometric measurement component is subjected to attitude normalization processing to obtain normalized air disturbance data;

[0025] Based on the normalized air disturbance data, disturbance feature quantities are determined, and the disturbance feature quantities are input into the air path classification model to generate air path state labels.

[0026] The status markers of the air path are combined to form a second status identifier.

[0027] Optionally, the attitude normalization process includes performing an optical path geometric remapping operation on the air disturbance data based on the optical path geometric offset caused by the pitch or azimuth change of the spherical mirror, so as to eliminate the geometric influence of attitude change on air disturbance measurement.

[0028] The determination of the disturbance characteristic quantity includes projecting the normalized air disturbance data onto a preset low-order spatial mode basis, calculating the volatility and phase consistency index of each mode coefficient within a preset time window, and combining the volatility and phase consistency index as the disturbance characteristic quantity.

[0029] The second state identifier further includes a time sequence code for air path state switching to characterize the historical dependence of air path disturbances.

[0030] Optionally, determining the topology annealing sequence parameters based on the first state identifier and the second state identifier, and executing the topology annealing sequence, includes:

[0031] Read the coupling mode entry corresponding to the current time window from the cross state index, and generate a coupling mode description quantity to characterize the coupling relationship between the support contact state and the air path state;

[0032] Based on the coupling mode descriptor, the annealing action primitives are selected and parameterized in the controller to obtain the topology annealing control vector.

[0033] A topology annealing sequence parameter set is generated based on the topology annealing control vector, and an annealing step sequence including unlocking, micro-attitude perturbation, resting and locking is driven.

[0034] After the annealing step sequence is completed, the first state identifier and the second state identifier are updated, and the preset annealing termination condition is determined. If the condition is not met, the topology annealing control vector is regenerated based on the updated coupling mode descriptor and the process is repeated.

[0035] Optionally, the coupling mode description includes coupling correlation strength and coupling preference direction;

[0036] The annealing action primitive includes at least the attitude jitter direction, attitude jitter amplitude, jitter rhythm, rest duration and locking sequence, and identifies the set of coupled hot spot support points through the state mapping table of the first state machine to determine that the locking sequence is non-hot spot support points first and then hot spot support points.

[0037] The sampling triggering of the annealing step sequence is limited by the second state identifier to be activated when the air path meets the preset stability criterion;

[0038] When the coupling correlation strength does not meet the threshold, the attitude jitter amplitude and jitter rhythm are adjusted by a preset attenuation along the coupling priority direction before entering the next annealing step sequence.

[0039] Optionally, triggering the formal interferometric measurement based on the initial state confirmation signal includes:

[0040] After the initial state confirmation signal is generated, the control channels of the attitude actuator and the locking actuator are frozen, and the first state identifier and the second state identifier are continuously read during the sampling period. When a switch of either state identifier is detected, the sampling is stopped and the process is rolled back to the initial state establishment process.

[0041] Using the second state identifier as the sampling gate, a dual sampling strategy is adopted to acquire the first interferometric measurement frame and the second interferometric measurement frame in adjacent time slots. The modal coefficient fluctuation rate of the second state identifier is required to be lower than a preset threshold and its time sequence code is consistent, so as to form a candidate frame pair.

[0042] Perform a back-mapping registration operation based on optical path geometry on the candidate frame pair to unify the two frames of interferometric measurement data into the initial state reference coordinate system;

[0043] Based on the air disturbance mode coefficients recorded by the second state identifier, the interferometric measurement data of the candidate frame pair is processed by air component subtraction to obtain surface-specific wavefront data, and the surface-specific wavefront data is written into the measurement data buffer as input to generate evaluation index data.

[0044] Secondly, this application provides a multi-condition surface stability measurement system for spherical mirrors, including:

[0045] A construction module is used to construct a two-layer state machine, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state.

[0046] The acquisition module is used to acquire load sensing data related to the support contact state and optical path monitoring data related to the air path state.

[0047] The processing module is configured to generate a first state identifier representing the support contact state based on the load sensing data; generate a second state identifier representing the air path state based on the optical path monitoring data; determine the topology annealing sequence parameters based on the first state identifier and the second state identifier, and execute the topology annealing sequence.

[0048] The update module is used to update the first state identifier and the second state identifier in real time during the execution of the topological annealing sequence, and generate an initial state confirmation signal when both of them simultaneously meet their respective preset stability conditions.

[0049] The generation module is used to trigger formal interferometric measurement based on the initial state confirmation signal, obtain the surface shape data of the spherical mirror under the current working condition, and generate evaluation index data to characterize the surface shape stability of the spherical mirror under multiple working conditions based on the surface shape data and the corresponding first state identifier and second state identifier.

[0050] Compared with existing technologies, this application introduces a two-layer state machine to characterize the contact state and air path state of the spherical mirror support, and separately determines the support state identifier and air path state identifier. This allows the surface stability assessment process to fully consider the path dependence effect of the support structure state and the dynamic changes of air path disturbances. Before actual measurement, this method actively eliminates state uncertainties caused by micro-slippage of the support structure, changes in locking sequence, or changes in air path through a topological annealing sequence, thus achieving accurate determination of the stable measurement state.

[0051] Furthermore, the method in this application combines real-time updated dual-layer state markers with strict stability judgment conditions, triggering formal interferometry only when the state is truly stable, thereby fundamentally improving the reliability and repeatability of the measurement data. Moreover, the evaluation index data formed by the surface profile data obtained from the stable measurement state and the state markers comprehensively and accurately reflects the combined impact of the cross-coupling between the support contact state and air path disturbances on the stability of the spherical mirror surface profile, significantly improving the objectivity and accuracy of the multi-condition surface profile stability evaluation results, and demonstrating significant advantages over existing technologies. Attached Figure Description

[0052] Figure 1 A flowchart of a method for measuring the surface stability of a spherical mirror under multiple working conditions, provided in an embodiment of this application;

[0053] Figure 2 A flowchart illustrating a method for constructing a two-level state machine, as provided in this application embodiment;

[0054] Figure 3 A flowchart illustrating a method for generating a first state identifier representing a support contact state, provided in an embodiment of this application;

[0055] Figure 4 This is a schematic diagram of a spherical mirror multi-condition surface stability measurement system provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0057] See Figure 1 The flowchart shown is a method for measuring the stability of a spherical mirror under multiple working conditions according to an embodiment of this application, including steps S101 to S105, wherein:

[0058] S101: Construct a two-layer state machine, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state.

[0059] S102: Collect load sensing data related to the support contact state and optical path monitoring data related to the air path state;

[0060] S103: Based on the load sensing data, generate a first state identifier representing the support contact state; based on the optical path monitoring data, generate a second state identifier representing the air path state; based on the first state identifier and the second state identifier, determine the topology annealing sequence parameters and execute the topology annealing sequence;

[0061] S104: During the execution of the topological annealing sequence, the first state identifier and the second state identifier are updated in real time, and an initial state confirmation signal is generated when both simultaneously meet their respective preset stability conditions.

[0062] S105: Based on the initial state confirmation signal, a formal interferometric measurement is triggered to obtain the surface shape data of the spherical mirror under the current working condition. Based on the surface shape data and the corresponding first state identifier and second state identifier, evaluation index data for characterizing the surface shape stability of the spherical mirror under multiple working conditions is generated.

[0063] Regarding the above S101:

[0064] In one embodiment, to effectively identify key states affecting the surface stability of a spherical mirror under various operating conditions, this application first constructs a dual-layer state machine. This dual-layer state machine includes a first state machine and a second state machine, used to characterize the support contact state and air path state of the spherical mirror, respectively. The first state machine takes the outputs of multiple load sensors deployed on the back or sides of the spherical mirror as input, and can classify the force conditions at different support points into finite state categories, such as a bearing state, a light touch state, or a removal state. The second state machine takes monitoring channel data and environmental sensor data deployed around the interferometer's optical path as input, and is used to abstract air disturbance conditions into finite states, such as a relatively stable laminar flow corridor state, a local convection disturbance state, or a mixed zone state.

[0065] Through this two-layer modeling approach, different physical influence sources are uniformly converted into traceable state identifiers, thereby providing a data foundation for subsequent annealing control and stability determination.

[0066] It is important to emphasize that the "state machine" referred to in this application is an abstract logical modeling method used to discretize continuously changing physical quantities into a finite set of states and define the transition conditions between these states. In this way, complex physical processes can be represented and tracked using a finite set of states and state transitions, thus facilitating their use in subsequent data processing and decision-making stages.

[0067] For example, in the support structure of a spherical mirror, multiple load sensors continuously output values ​​that change continuously over time. By introducing a state machine, the force situation at each support point can be discretized into finite states such as "bearing," "touching," and "leaving," and the sequence of state transitions can be recorded. Similarly, for air disturbance paths, although the refraction changes in the airflow field are continuous, the state machine can abstract them into finite states such as "laminar corridor," "convective plume," and "mixing zone." In this way, the complex support topology and air disturbances are mapped to discrete states that are easy to calculate and compare. Subsequently, when performing topology annealing and stability determination, it is only necessary to check whether the state indicators meet preset conditions, without having to solve for all continuous physical quantities in real time.

[0068] In practical applications, such as when inspecting a vertically mounted spherical mirror with a diameter of approximately 3 meters, 48 ​​back support points and several lateral support points simultaneously provide real-time load data. Through a first state machine model, it is possible to quickly identify whether some support points transition from a load-bearing state to a light-touch state, a transition that, if left uncontrolled, can easily cause localized drift in the overall surface shape. Simultaneously, under the same inspection environment, with an interferometer optical path approximately 30 meters long, the distribution of air disturbances changes with temperature gradients and pitch angles. The second state machine can provide a judgment result within the sampling window regarding whether the current optical path is in a stable laminar flow state.

[0069] For example, during the process of the ambient temperature rising from 20°C to 26°C, if the fluctuation rate of the air disturbance mode exceeds the preset threshold, the second state machine will output "convective disturbance state".

[0070] The evaluation method utilizes the output of the aforementioned two-layer state machine to provide a dual characterization of the support topology and air disturbance state at an early stage, serving as a prerequisite for subsequent stability assessment procedures. Those skilled in the art can select appropriate sensor arrangements and state sets for modeling based on different apertures, support layouts, or environmental conditions; the specific state categories are not limited to the examples described above.

[0071] Regarding S102 above:

[0072] In one embodiment, to provide input data for the two-layer state machine, the method of this application includes acquiring load sensing data related to the support contact state and optical path monitoring data related to the air path state. Load sensing data refers to the output signals of force sensors deployed at multiple support points or lateral support positions on the back of the spherical mirror; these signals characterize the magnitude and variation of the force borne by each support point. Optical path monitoring data refers to data characterizing air disturbance conditions acquired by peripheral channels or auxiliary sensors in the interferometer's detection optical path, such as spot tilt, phase disturbance, light intensity fluctuation, and temperature and humidity parameters. Both types of data are accessed through a unified data acquisition system, and timestamps are added during acquisition to ensure alignment of data from different sources on the timeline.

[0073] In specific implementations, for example, for a vertical spherical mirror with a diameter of about 3 meters, strain gauge force sensors are installed at the 48-point support on the back and some lateral supports. The range of the sensors is 0 to 2000 N, the sampling frequency is set to 100 Hz, and the signals are transmitted to the data acquisition card via the CAN bus.

[0074] In the same detection scenario, the optical path monitoring section deploys four tilt detectors and two temperature and humidity sensors next to the interferometer's optical path, with a sampling frequency of 20 Hz, and transmits the data to the main control computer via an Ethernet interface. The acquired raw data undergoes zero-point calibration and moving average filtering before entering the state machine to remove high-frequency noise and reference drift. Within a single sampling cycle, the system can simultaneously obtain force curves for dozens of support points and multiple air disturbance curves, organizing them into a unified multi-dimensional matrix input to a two-layer state machine for generating subsequent state identifiers.

[0075] Those skilled in the art can freely choose different types of force sensors, air monitoring devices, sampling frequencies and data interface methods depending on the aperture size of the spherical mirror, the number of supports and the detection environment. The above example is only one feasible implementation method and does not constitute a limitation on the method of this application.

[0076] Regarding the above S103:

[0077] In one embodiment, the generation of the first state identifier is based on the continuous acquisition and synchronization of the aforementioned load sensing data. Specifically, the load sequence of each support point can be statistically analyzed within a fixed time window, such as 0.5 s to 2 s, to obtain a set of characteristic quantities used to characterize the steady state and disturbance degree of the support point force, such as the time window average value of the load, the fluctuation amplitude within the time window, and the rate of change between adjacent samples. After completing zero-point removal, drift removal, and simple smoothing, the system calls a preset set of state criteria to discretize the single-point force from the continuous numerical domain into a finite set of states, such as bearing, light touch, or removal, and forms a support point-state mapping table.

[0078] The state criterion can be implemented using threshold band discrimination, cluster-based interval partitioning, or other equivalent supervised / semi-supervised discrimination processes, and is not limited to a specific algorithm. Subsequently, the system merges the discrete states of all support points under a unified coordinate system to obtain a frame for the "first state identifier" at the current moment.

[0079] For example, when the spherical mirror is tilted at a small angle, the average load of a group of support points near the lower edge decreases and the fluctuation range increases within a short time window, and the judgment result tends to be a light touch state; this spatial distribution is included in the first state identifier to indicate that subsequent sequences should pay attention to the load recovery of this area.

[0080] In another embodiment, the generation of the second state identifier is based on the spatiotemporal stability determination of air disturbances using optical path monitoring data. Observations related to air refraction disturbances, such as light spot tilt, phase disturbance, light intensity fluctuations, and temperature and humidity changes, are obtained from the peripheral monitoring channels of the optical path. Within the same time window, the fluctuation amplitude, stability duration, and correlation or consistency indicators among multiple points are calculated and compared with preset criteria. The air path is then discretized into a finite set of states, such as laminar flow corridors, convective disturbances, or mixing zones, and the output is the second state identifier.

[0081] Furthermore, threshold hysteresis and minimum duration constraints can be introduced to avoid frequent switching caused by transient disturbances. For example, during the process of the plant temperature slowly rising from 20 ℃ to 24 ℃, if the light spot is observed to drift slowly in the same direction across multiple channels with increased correlation, the system can determine it as a laminar flow corridor state. However, when the air conditioning starts or stops and introduces local eddies, the fluctuation amplitude of each channel increases and the consistency decreases in a short period of time, which the system can determine as a convective disturbance or mixing zone, and update the second state label accordingly.

[0082] In another embodiment, based on the first state identifier and the second state identifier, the topology annealing sequence parameters are determined and the topology annealing sequence is executed, which is a preprocessing process aimed at state convergence.

[0083] For ease of understanding, this application interprets the topological annealing sequence parameters as a parameterized description of a set of standardized actions, including but not limited to unlocking, micro-attitude perturbation, resting and locking, which can be executed cyclically in a predetermined order.

[0084] In specific implementation, the current first state identifier and second state identifier are first read, and the direction of implementation of micro-attitude perturbation is selected according to their spatial distribution and stability indication. For example, a small perturbation, perturbation amplitude and rhythm are applied to a degree of freedom of pitch or azimuth, such as small angle, small step, intermittent, as well as the resting time and the triggering order of locking operation. Then, the unlocking-micro-attitude perturbation-resting-locking step sequence is executed according to the parameters.

[0085] After each sub-step, the system immediately refreshes two types of status indicators: if the first status indicator shows that the previously lightly touched area has returned to bearing capacity, and the second status indicator shows that the air path is in a stable state, then the system proceeds to the next sub-step; if either state still shows instability, the system maintains or fine-tunes the current parameters and continues to promote state convergence by alternating small-amplitude perturbations and rest on the same degree of freedom. Through this "determine-execute-refresh" closed loop, the topological annealing sequence gently and traceably pushes the support contact state and air path state into the stable region, thereby creating repeatable initial conditions for subsequent initial state confirmation and formal interferometry.

[0086] For example, when the first state indicator indicates a local touch at the lower edge and the second state indicator indicates convective disturbance, the system can choose to make a smaller, shorter-rhythm perturbation in the pitch degree of freedom and extend the rest period to reduce the probability of triggering the air disturbance switching again; when both types of indicators show stability, the topology annealing in this embodiment is completed in one stage.

[0087] It is understandable that the above parameter selection and action sequence can be configured according to different equipment, venues and control resolutions, and are not limited to a single implementation form.

[0088] Regarding S104 above:

[0089] In one embodiment, during the execution of the topological annealing sequence, i.e., the cyclic process of unlocking—micro-attitude perturbation—resting—locking, the first state identifier and the second state identifier are updated in real time, and an initial state confirmation signal is generated when both simultaneously meet their respective preset stability conditions.

[0090] For ease of understanding, the meaning and implementation of real-time updates, stability conditions, and initial state confirmation signals are explained below.

[0091] For real-time updates:

[0092] In each sub-stage of the annealing sequence, such as after completing a micro-attitude perturbation or resting, the system reads a new round of load sensing data and optical path monitoring data under the synchronous timestamp from the data acquisition system. After zero-point correction, drift removal and smoothing, the data is input into the first state machine and the second state machine respectively to obtain the first state identifier and the second state identifier at that moment.

[0093] To mitigate the impact of brief bursts on decision-making, the system can maintain a sliding time window, such as 1–5 seconds, to perform consistency statistics on the status flags of recent frames. This includes counts of unchanged states, counts of state transitions, and variances or rates of change of characteristic quantities. Therefore, real-time updates encompass both refreshing the status flags of the latest frame and cumulatively evaluating the stability of states over a recent period.

[0094] In this embodiment, the update of the first state identifier is based on the spatial distribution of the discrete states (bearing / touching / leaving) of each support point; the update of the second state identifier is based on the discrete states of the air path (laminar corridor / convective disturbance / mixing zone) and their accompanying disturbance intensity indications. The aforementioned time window, sampling rhythm, and statistical caliber can all be configured according to the device resolution and environmental disturbance characteristics, and are not limited to a single one.

[0095] For stable conditions:

[0096] To avoid misclassifying transient pseudo-stable states as usable initial states, this embodiment employs a dual criterion with hysteresis and minimum duration constraints:

[0097] First, the first stability condition corresponding to the first state identifier is: within the sliding time window, the number of all support points marked as lightly touched or removed does not exceed a preset threshold, and their spatial positions do not spread; for the set of support points identified as historical hotspots, their state must continuously maintain the trend of bearing or converging towards bearing; at the same time, the load characteristics of each support point (such as fluctuation amplitude and rate of change) are generally lower than the preset upper limit, and there is no reverse transition between two adjacent sub-stages.

[0098] Secondly, there is a second stability condition corresponding to the second state identifier: within the sliding time window, the air path state remains within a stable set, such as a laminar flow corridor, or its disturbance intensity indicator remains below a threshold; the consistency and correlation indicators between multi-point monitoring channels are not lower than the lower limit; if a state switch has just occurred, a minimum duration must be met before it can be considered stable. This embodiment can set a hysteresis interval for the second stability condition to reduce repeated switching near the critical value.

[0099] The threshold, minimum duration, hysteresis width, and correlation lower limit mentioned above are configurable parameters that can be tuned according to different factory scales, optical path lengths, and sensor noise characteristics.

[0100] When the first and second stability conditions conflict—for example, the support side shows stability while the air side is still perturbed—the system does not output an initial state confirmation signal. Instead, it returns to the annealing sequence loop: maintaining small-amplitude, low-energy perturbations and rest periods alternately, slightly extending the rest period or reducing the perturbation rhythm to make the second state converge. Conversely, if the air is stable but the support is not, it prioritizes finer-step attitude perturbations and stricter locking sequence control for the hotspot areas of the first state. For cases where simultaneous stability is not achieved for an extended period, a timeout branch can be triggered: recording the reasons for non-convergence and key statistics, stopping the current round of annealing, and using the information for parameter retuning or manual verification.

[0101] The conflict handling strategy in this embodiment is based on the principle of not amplifying the source of disturbance, and avoids causing new state transitions due to excessive intervention.

[0102] Furthermore, when both the first and second stability conditions are met within the same decision window, the system generates an initial state confirmation signal. This signal may include, but is not limited to: a timestamp, a snapshot of the current attitude reading and the locked state, a summary of the first and second state identifiers (e.g., a hash or compressed description), key statistics used to determine stability (e.g., invariant count, upper limit of rate of change, lower limit of correlation), and a parameter version identifier. The initial state confirmation signal indicates that the current operating condition has entered a repeatable and traceable stable measurement starting point; subsequent processes can use this signal as a trigger to initiate formal interferometric measurements, but this embodiment does not limit the measurement details.

[0103] For example, taking a spherical mirror with an aperture of about 3 m and installed vertically as an example, in one annealing cycle, the first state indicator converges from sporadic light touch at the lower edge to full-area bearing within 12 seconds, while the second state indicator transitions from the mixing zone to the laminar flow corridor within 8 seconds, and the stable intervals of the two overlap by 4 seconds; based on this, the system generates an initial state confirmation signal at time 20 seconds.

[0104] If a brief bounce of the air path occurs in the subsequent perturbation phase (the second state indicator leaves the stable set), the system cancels the incomplete decision, returns to rest, and extends the duration until the bistable condition is met again. The aforementioned duration, threshold, and rhythm parameters may vary depending on the device and environment and do not constitute a limitation of this application.

[0105] In this way, the execution of the annealing sequence, the updating of the status flag, and the dual criteria for stability form a tight closed loop, which can provide a repeatable starting reference for subsequent surface measurement without relying on forced numerical compensation.

[0106] Regarding the above S105:

[0107] After generating the initial state confirmation signal, the system enters the formal interferometric measurement phase. To maintain consistency between the measurement and the initial reference, the controller freezes the attitude and locking execution permissions, keeping only the data acquisition and measurement trigger channels active.

[0108] Understandably, the measurement can be triggered by a command from the host computer or automatically by an initial state confirmation signal. The acquired interferometric data is processed under an effective aperture mask to form surface data. The processing includes removing piston and tilt terms; whether the power term is retained is determined by the configuration.

[0109] To facilitate traceability and comparison, the surface data was unified to the initial state reference coordinate system and recorded together with the timestamp of this measurement, attitude readings, locking state snapshot, environmental parameters, and the corresponding first and second state identifiers.

[0110] To control data quality, a short-term stabilization window after initial state confirmation can be used as the sampling interval. If a switch between the first and second state identifiers is detected within this interval, the current frame is discarded and the process returns to the initial state establishment procedure. Qualified frames within the same stabilization window undergo registration and synthesis to mitigate the impact of random disturbances. Registration employs a back-mapping method or equivalent registration method based on optical path geometry to unify all frames to the same reference frame. Synthesis can use simple averaging, weighted averaging, or robust estimation, with weights determined by the stability indication provided by the second state identifier. Through gating, registration, and synthesis, surface data consistent with the initial state reference and satisfying the stability criteria are obtained.

[0111] Based on the surface shape data and its corresponding first and second state identifiers, the system generates evaluation index data to characterize the surface shape stability of spherical mirrors under multiple operating conditions. The evaluation index data includes three levels: single-condition index, cross-condition stability index, and path coupling related index.

[0112] The single-condition indices include PV and RMS values, band-limited RMS divided by spatial frequency bands, the set of low-order mode coefficients obtained from modal decomposition, and slope statistics and the location of the maximum residual. The cross-condition stability indices are obtained by aggregating and comparing the single-condition indices of multiple conditions, including the global stability index of the maximum RMS offset and its confidence value within the condition set, the modal stability spectrum composed of the amplitude and phase variation and consistency of low-order modes across different conditions, the sensitivity matrix of the influence intensity of attitude and temperature on RMS or key low-order modes, and the stability map showing the usable area and the degraded area on the attitude-temperature plane.

[0113] The path coupling related indicators are calculated using the first and second state identifiers recorded synchronously with the measurement. These include the path dependence indicator of the surface shape difference when the same set point is taken through different return paths, the correlation between the state change of the support hotspot area and the low-order aberration change, the correlation between air path stability and mid-to-low frequency surface shape change, and the proportion of complementary stability windows in which the measurement is completed within the time percentage that satisfies the dual-state stability.

[0114] The specific definitions, orders, and thresholds of the above indicators can be adjusted according to the equipment and environment, and are not the only limitations.

[0115] The evaluation index data is output in the form of a structured data package. The data package includes index items for the current and previous measurements, operating condition labels, corresponding first and second state identifiers summaries, statistical confidence levels and quality control markers, and can provide visual references to stability maps and sensitivity matrices. Taking a spherical mirror with an aperture of approximately three meters and employing long-range autocollimation measurement as an example, the RMS obtained by the system under the 0° / 20 ℃ condition is on the order of λ / 160, and the RMS obtained under the +1° / 26 ℃ condition is on the order of λ / 120. The global stability index across operating conditions shows a difference of more than 30%.

[0116] In this paper, λ refers to the nominal operating wavelength used in the interferometry, such as 632.8 nm He–Ne light, used to normalize the wavefront or surface shape error. For example, 0.02λ represents 2% of the operating wavelength.

[0117] When retested using different zero-return sequences, the path-dependent indicators showed distinguishable differences. Furthermore, during periods when the air path was within the laminar flow corridor, the mid-to-low frequency RMS fluctuations were significantly lower than during periods of convective disturbance. These results illustrate the formation mechanism of the indicators and typical observable phenomena; actual values ​​and thresholds can be configured based on field conditions.

[0118] Through the above process, formal interferometry and initial state confirmation, state identification tracing, and index generation form a closed-loop output. The obtained evaluation index data not only reflects the surface profile level under a single working condition, but also comprehensively characterizes the impact of support contact state and air path state on surface profile stability across multiple working conditions. This provides comparable, traceable, and decision-making data for the assembly, operation, and maintenance of spherical mirrors.

[0119] Optional, see Figure 2 The flowchart of a method for constructing a two-level state machine provided in this application embodiment includes steps S201 to S203, wherein:

[0120] S201: Based on the output data of multiple load sensors of the multi-point support mechanism, determine the contact micro-state characteristics of each support point, and perform clustering operation on the contact micro-state characteristics to generate a state identifier for each support point and form a state mapping table of the first state machine.

[0121] S202: Based on the output data of the optical path peripheral monitoring channel of the interferometric measurement component and the environmental monitoring sensor, determine the disturbance characteristics of the air path, perform classification calculations on the disturbance characteristics, and generate the state mapping table of the second state machine;

[0122] S203: Establish a cross-state index between the first state machine and the second state machine to identify the coupling mode descriptor between the support contact state and the air path state.

[0123] In one embodiment, the construction of the two-layer state machine first processes the load sensor outputs of the multi-point support mechanism. The system preprocesses the load sequences of each support point synchronized in time, including zero-point removal, drift removal, and smoothing, and then calculates statistical and response features that can characterize the steady state and the degree of disturbance within a sliding time window.

[0124] In order to discretize continuous numerical values ​​into finite states, the system performs clustering operations or equivalent discrimination processes in the aforementioned feature space and marks the current force state of each support point as a category such as bearing, lightly touching, or leaving.

[0125] Subsequently, using timestamps as indexes, the support point numbers and their discrete states are combined and registered to form a state mapping table for the first state machine. This mapping table is continuously updated over time, reflecting both the current spatial distribution and preserving the state evolution trajectory of recent periods, which is used to support subsequent stability assessments and process control.

[0126] For air path conditions, the system extracts characteristic quantities that can characterize the degree and stability of refractive field disturbances based on the synchronous data of the optical path peripheral monitoring channel of the interferometric measurement component and the environmental monitoring sensor. These include multi-point tilt and amplitude indication of phase or light intensity shift, cross-channel consistency and correlation indication, temperature and humidity change rate, etc.

[0127] It is important to note that, to achieve discretization, a classification operation is performed within this feature space, labeling the current air path as a laminar corridor, convective disturbance, or mixing zone, and then writing the category label and feature summary into the state mapping table of the second state machine using a timestamp as an index. This mapping table is also maintained in a time-series manner to reflect the stability and switching of the air path during the measurement phase.

[0128] After establishing two state mapping tables, the system aligns them based on a unified time axis to form a cross-state index for identifying the relationship between the support contact state and the air path state. This index measures the covariance strength between local state changes on the support side and changes in air path disturbances by statistically analyzing the co-occurrence and sequential relationships of the two types of states within the same time window. Combined with the sequence information and spatial distribution information of state switching, a set of descriptive quantities for quantifying the coupling relationship between the two is obtained, serving as coupling mode descriptive quantities for subsequent steps.

[0129] For example, when the spherical mirror enters the sampling phase under the conditions of pitch angle θ = +1° and ambient temperature T = 26℃, the state mapping table of the first state machine shows that several support points at the lower edge change from bearing load to light contact. Simultaneously, the state mapping table of the second state machine shows that the air path changes from a laminar flow corridor to convective disturbance. The cross-state index provides a high level of covariance indication and a clear temporal sequence. This result is recorded as the coupling mode descriptor for the current time window, used to drive subsequent parameter tuning and annealing processes. To adapt to different apertures, support arrangements, and environmental conditions, the above feature selection, clustering and classification methods, time window length, and correlation aperture can all be tuned according to engineering needs and are not considered as unique limitations.

[0130] This alternative implementation addresses the measurement uncertainty caused by the difficulty in simultaneously identifying the path dependence of the support contact state and the remapping of air path disturbances in existing assessments. By establishing a first state machine and a second state machine and forming two types of state mapping tables through clustering and classification operations, and then constructing a cross-state index with a unified time axis to obtain the coupling mode description, it achieves quantitative characterization and real-time tracking of the support-air dual-domain relationship.

[0131] This transforms continuous, noisy observations into determinable discrete states, significantly improving the robustness and reproducibility of identification under unstable conditions. Coupled mode descriptors provide a calculable basis for selecting topological annealing parameters, reducing blind adjustments and the release of "pseudo-stable" conditions. Recording data in temporal order and spatial distribution creates traceable data assets, facilitating cross-condition comparisons and long-term trend assessments, thereby improving the overall accuracy and operability of multi-condition surface stability assessments.

[0132] Optionally, the contact micro-state characteristics are a feature vector composed of the load fluctuation of the load sensor and the displacement response of the support point within a continuous time window.

[0133] The disturbance characteristic is the modal coefficient fluctuation rate after the air path disturbance amount within a continuous time window is projected onto a preset low-order spatial modal basis.

[0134] The state mapping tables of the first state machine and the second state machine respectively record the state switching order and generate path history codes. The cross-state index is established by associating the path history codes of the first state machine and the second state machine.

[0135] In order to accurately characterize the minute state changes of the support contact and the low-order disturbances of the air path under multiple operating conditions, and to incorporate their historical order into the same index in order to reveal path dependence and coupling relationships.

[0136] In this embodiment, a contact micro-state feature vector is constructed for each support point within a continuous time window. The load fluctuation can be taken as the standard deviation or range within this window, or as the residual energy after median filtering. The displacement response can be read by a micro-displacement sensor at the support point, or estimated by the actuator encoder and stiffness model. The two together constitute the feature vector of the support point within this window. The window length can be set from 0.5 s to 2 s according to the sampling frequency and environmental disturbance characteristics, and the window sliding step size can be an integer multiple of the sampling period.

[0137] To eliminate the effects of overall temperature drift or slow shift, the feature vectors can undergo mean reduction and normalization before entering clustering or discrimination, and neighborhood constraints can be introduced to prevent anomalies of adjacent support points from being amplified. The system processes the feature vectors of all support points in parallel, outputting discrete states such as bearing light touch and departure, and updates the state mapping table of the first state machine accordingly.

[0138] For air path disturbances, the system summarizes the observations such as tilt, phase disturbance, and light intensity fluctuation of the peripheral monitoring channel of the optical path within the same time window. It first performs attitude and geometry back mapping to make the data fall into a unified reference frame, and then projects it onto the preset low-order spatial mode basis.

[0139] The modal basis can be a low-order Zernike term, an orthogonal polynomial, or a sine-cosine basis, with the order and basis type determined by the field conditions. The system statistically analyzes the volatility of each modal coefficient within a window, using the ratio of the coefficient standard deviation to the mean or the root mean square of the differences between adjacent frames as the volatility measure. Based on the combination of the volatility of each coefficient and threshold rules, the window is classified as a laminar flow corridor, convective disturbance, or mixing zone, and the state mapping table of the second state machine is updated accordingly.

[0140] To avoid threshold boundary jitter, the category determination can be set with a minimum duration and hysteresis band. The specific values ​​are determined by the device resolution and the airflow conditions in the plant.

[0141] The two state mapping tables record not only the current category but also the order of state transitions. The system generates a path history code for each time window. The first path history code is obtained by compressing the category change sequence of each support point in the window queue; this can be achieved using a hash concatenation of the support point index and state sequence number, or by using an ordered list based on spatial partitioning. The second path history code is obtained by compressing the change sequence of the air path category over time, along with the sign and amplitude range identifier of the dominant low-order mode. The cross-state index aligns the two types of path history codes via a unified time axis, generating a co-occurrence table and a sequence relationship table, further providing the covariance strength and time priority direction in the coupled mode descriptors.

[0142] For example, taking a 3 m vertical spherical mirror as an example, with a 1 s window and a 0.1 s step size, in the stage where θ = +1° and T = 26℃, the load fluctuation at several support points at the lower edge increases while the displacement response increases, indicating a transition from load-bearing to light contact. Simultaneously, the low-order term volatility of the air mode increases and transforms from a laminar flow corridor to convective disturbance. The first and second path history codes of this time window show significant co-occurrence in the cross-index, indicating that the support side precedes the flow. Based on this, the system marks the covariance intensity of this window as higher and records the time-priority direction. This processing provides calculable input for subsequent parameter tuning and annealing processes, while preserving complete sequence information, facilitating playback and comparison across operating conditions.

[0143] Optional, see Figure 3 The flowchart below illustrates a method for generating a first state identifier representing a support contact state, as provided in an embodiment of this application. The method includes:

[0144] S301: Calculate the average load, load fluctuation amplitude and load change rate of the load sensing data for each support point within a preset time window, and combine the above three quantities as the load feature vector of the support point.

[0145] S302: Perform neighborhood normalization on each load feature vector according to the spatial distribution of support points to obtain the relative load feature vector after eliminating the influence of overall temperature drift, and calculate the offset from the historical baseline in the relative load feature vector.

[0146] S303: Input the offset into a micro-state classification model based on support point partitioning, and combine the displacement response data or micro-vibration response data of the support point in the model to generate a state label of bearing state, light touch state or departure state.

[0147] S304: Combine the state markers of all support points to form a first state identifier, and attach the time sequence code of the state switching of each support point to the first state identifier to characterize the path dependency features of the support contact topology.

[0148] After collecting load sensing data, if the judgment is made based solely on the absolute value of a single load or a simple threshold, it is easily affected by overall temperature drift, slow offset, and mechanical common-mode disturbance, leading to misjudgment of load and light touch. At the same time, without introducing neighborhood constraints and historical references, occasional fluctuations of edge support points will be amplified, and the classification results will fluctuate and bounce, making it difficult to stably support subsequent processes.

[0149] Therefore, this implementation performs differential and normalization processing in both time and space dimensions, and combines mechanical characteristics with displacement or micro-vibration response for determination, thereby improving the robustness and repeatability of support contact state identification.

[0150] Regarding the above S301:

[0151] When judging based solely on the instantaneous load value at a single point, misjudgments can easily be introduced due to environmental vibrations and sampling noise, and it is impossible to distinguish between steady-state load and short-term disturbances.

[0152] Therefore, within a preset time window, three quantities are extracted from the load sensing data of each support point and arranged in a fixed order to form a load feature vector, which is used for subsequent state discrimination and mapping. The time window can be 1.0 s, the sliding step size can be 0.1 s, and the sampling frequency of the load signal can be 100 Hz.

[0153] For load samples within a window, the average load is calculated using an arithmetic mean to reflect the bearing baseline level of the support point in the current window. The load fluctuation amplitude is characterized by the difference between the upper and lower bounds within the window. The upper and lower bounds are obtained by performing median smoothing and mild outlier removal on the window samples to suppress the influence of occasional spikes on amplitude estimation. The load change rate is obtained by averaging the absolute values ​​of the differences between adjacent samples within the window and normalizing them over time, which is used to characterize the dynamic activity of the support point in the current window.

[0154] The three quantities mentioned above are denoted as the average load, load fluctuation amplitude, and load change rate of the support point, respectively. They are concatenated in this order to form a one-dimensional load feature vector, which is then stored in association with the spatial index and timestamp of the support point. Taking a 3 m vertical spherical mirror as an example, under the conditions of θ = +1° and T = 26℃, sampling a support point at the lower edge within a 1.0 s window yields an average load of approximately 940 N, a load fluctuation amplitude of approximately 32 N, and a load change rate of approximately 5.8 N per second. Therefore, the load feature vector of this support point within this window is [940, 32, 5.8]. The time window, step size, sampling frequency, and the measurement caliber of fluctuation and rate can be adjusted according to the equipment resolution and the characteristics of on-site disturbances, and are not considered unique limitations.

[0155] Regarding the above S302:

[0156] To reduce the impact of overall temperature drift and slow offset on the judgment, the system performs neighborhood normalization based on the spatial distribution of the support points after generating the load feature vector, and then compares it with the historical baseline to obtain the offset. The neighborhood can be determined by dividing the support points into several rings or sectors according to the mechanical layout, or by determining a fixed-size neighborhood set based on several nearest neighbor numbers for each support point. For each support point, the system first calculates three local statistics within its neighborhood, corresponding to the local center value and local scale value of the load average, load fluctuation amplitude, and load change rate, respectively. The local center value is preferably a truncated mean or median to improve the anti-outlier capability, and the local scale value is preferably an interquartile range or robust standard deviation. Then, the three-dimensional load feature vector of the support point is subtracted from the local center value by each component and divided by the local scale value to obtain the relative load feature vector, which mainly reflects the difference relative to the nearby environment rather than the global common mode change. To avoid sharp discontinuities introduced by neighborhood random noise, the system can perform a slight smoothing according to the spatial diagram structure after normalization, with the smoothing weight set according to the proximity distance and structural connection strength.

[0157] To ensure comparability across different sampling rounds, the system applies a global consistency constraint to the relative vectors, making the relative average component of all support points close to zero within the same time window. Specifically, this is achieved by performing zero-mean correction on the relative components according to rings or sectors, or by removing residual linear trends across the entire field. For points near or near missing sensors, the system uses neighborhood interpolation to complete the missing components and adds quality labels. Low-quality samples are not used in subsequent threshold determination and are only used for interpolation display.

[0158] Historical baselines are solidified from one or more confirmed stable reference phases. The confirmation of a reference phase presupposes a dual-state stability condition. A relative load feature vector is generated using a time window length and preprocessing caliber consistent with the current processing. The baseline center value and baseline scale value are then calculated according to the support point index, and the version number and timestamp are recorded. To address slow equipment aging or long-term drift, the system supports a rolling update strategy, updating the baseline only with a smaller weight when it passes the stability determination again, while retaining the old version for backtracking. The baseline and the current relative vector are aligned under the same coordinates and mask to ensure consistent comparison.

[0159] After obtaining the relative load eigenvector, the system calculates the offset from the historical baseline. The offset includes the differences of three components and a dimensionless offset normalized to the baseline scale. A weighted composite intensity can also be provided for ranking. The component differences reflect the deviation of the current point from the stable reference in terms of load level, disturbance amplitude, and dynamic rate. The dimensionless offset facilitates comparisons across points and time periods. To suppress occasional abrupt changes, the offset can be averaged over time with a short window and a capped pruning. Samples exceeding the cap are marked as anomalies and trigger review or removal.

[0160] For example, under the conditions of θ=+1° and T=26℃, the load feature vector obtained at a certain support point on the lower edge of a 3 m vertical spherical mirror within a 1.0 s window is 940, 32, and 5.8. Its neighborhood local center values ​​are 950, 20, and 3.0, and its local scale values ​​are 20, 8, and 1.0, respectively. After normalization, the relative load feature vector at this point is approximately -0.5, +1.5, and +2.8. The relative vector centers of the historical baseline at the same index are -0.25, +1.0, and +1.5. Therefore, the component differences between the current point and the baseline are approximately -0.25, +0.5, and +1.3. The dimensionless offset obtained after baseline scale normalization falls at a moderate level, and the corresponding quality label is valid. This offset will be used as one of the inputs to the subsequent microstate classification model and will participate in the generation of the time sequence code to characterize the evolution path of the contact topology. The time window, neighborhood division, robust statistical caliber, smoothing and pruning thresholds can be adjusted according to the device resolution and the characteristics of on-site disturbances, and do not constitute a unique limitation.

[0161] Regarding S303 and S304 above:

[0162] In one embodiment, the system feeds the aforementioned offset from the historical baseline into a micro-state classification model constructed by support point partitioning. Partitioning can be based on support geometry, using rings or sectors, allowing the model to learn or set more suitable thresholds and criteria within a local range. The model's input includes at least the three-dimensional offset of the support point in the current time window, while also incorporating displacement response data or micro-vibration response data as auxiliary quantities. The displacement response can be directly provided by the displacement sensor at the support point, or estimated by the actuator encoder in conjunction with the support stiffness relationship. The micro-vibration response can be provided by calculating the root mean square or energy index within a defined frequency band using an accelerometer located close to the support point. The model's output is a state label of one of three categories: bearing, light contact, or departure, along with a confidence value or distance metric for quality control. To reduce jitter, hysteresis and minimum duration constraints are set internally within the model.

[0163] Taking rule implementation as an example, a three-dimensional offset band threshold can be set for each partition and jointly decided with the auxiliary threshold. Different threshold bands are used when entering or exiting a certain state.

[0164] Taking statistical learning as an example, stable load and light touch samples can be collected during the production calibration phase to fit cluster centers and covariance. In the online phase, Mahalanobis distance and auxiliary quantity consistency are used for joint discrimination. Both implementations can deploy quality labels. When the input signal is missing or the noise exceeds the limit, the output maintains the previous stable state and is marked as needing retesting.

[0165] At the same timestamp, the system aggregates the state markers of all support points and combines them by spatial index to form the first state identifier for the current moment. This identifier reflects the discrete spatial distribution of the support contact states across the entire field and serves as the direct input for subsequent cross-indexing and annealing processes. To characterize path dependence, the system also generates a time sequence code to record the order of state transitions for each support point within the sliding window.

[0166] The time sequence code can be stored in the form of an event sequence, including timestamps, support point indexes, pre-switch status, and post-switch status. Alternatively, similar switches of adjacent support points can be merged and compressed at the spatial partition level to generate partition-level sequence entries.

[0167] To balance volume and traceability, the system can deduplicate and threshold-prune the original event sequences, and generate sequence summaries and hashes for rapid comparison. Time sequence codes and first-state identifiers are archived synchronously on the timeline, allowing for simultaneous retrieval and reconstruction of spatial distributions and recent evolutionary paths at the same time.

[0168] For example, taking a 3 m vertical spherical mirror as an example, when θ = +1° and T = 26℃, the lower edge support point of a certain ring shows an increase in load offset and displacement response within a 1.0 s window, with a synchronous increase in micro-vibration energy. Based on this, the partitioning model classifies this point as switching from load-bearing to light-touch, and two adjacent points subsequently experience similar switching at 0.2 s and 0.4 s respectively. The system generates a first state identifier at this timestamp and writes three light-touch tags, while simultaneously recording three time-ordered switching events in the time sequence code. The sequence summary shows that the switching propagates along the edge direction. The subsequent annealing process can then select the direction and rhythm of the micro-amplitude attitude disturbance accordingly, and adopt a stricter locking sequence and a longer settling time in hotspot areas. The above partitioning method, the method of obtaining auxiliary quantities and the calculation caliber, the specific implementation of the classification model, and the specific format of the sequence code can all be adjusted or replaced according to the equipment resolution and field conditions, and should not be construed as limiting the claims of this application.

[0169] Optionally, generating a second state identifier characterizing the air path state based on the optical path monitoring data includes:

[0170] The air disturbance data collected by the peripheral monitoring channel of the interferometric measurement component is subjected to attitude normalization processing to obtain normalized air disturbance data;

[0171] Based on the normalized air disturbance data, disturbance feature quantities are determined, and the disturbance feature quantities are input into the air path classification model to generate air path state labels.

[0172] The status markers of the air path are combined to form a second status identifier.

[0173] In one embodiment, the system takes air disturbance data obtained from the peripheral monitoring channel of the interferometric measurement component as input, first performs attitude normalization processing, then determines the disturbance feature quantity based on it and inputs it into the air path classification model, and finally summarizes the state markers of each channel at the same timestamp to form a second state identifier.

[0174] The purpose of attitude normalization is to eliminate the geometric remapping caused by pitch and azimuth fine-tuning, so that the monitored quantities under different attitudes fall into the same reference coordinate system. Specifically, the relative displacement and incident angle change of the current beam in space are calculated based on the attitude encoder readings and the optical path geometric model. The tilt, phase disturbance, and light intensity fluctuation of each monitoring channel are back-mapped and resampled according to the reference path. At the same time, the time delay is estimated by cross-channel cross-correlation and a small phase alignment is performed. If necessary, a linear scaling correction is performed according to the reference amplitude. This yields the attitude-normalized air disturbance data sequence.

[0175] After attitude normalization is completed, the system calculates characteristic quantities that characterize the strength and stability of disturbances for each channel within a preset time window. The characteristic quantities include at least two components: root mean square of tilt, root mean square of phase disturbance, coefficient of variation of light intensity fluctuation, consistency and correlation index across channels, and temperature and humidity change rate.

[0176] To enhance sensitivity to low-order fringe structures, the system projects the spatial distribution of phase perturbations onto a low-order modal basis, recording several dominant modal coefficients and their volatility. These features, after mean removal and scale normalization, are input into an air path classification model. The model can employ a threshold rule with hysteresis and minimum duration constraints, or an offline-calibrated statistical discriminator, outputting time-level air path state labels. The state set can include laminar corridors, convective disturbances, and mixing zones, along with a stability indicator for quality control.

[0177] At the same timestamp, the system summarizes the status markers and stability indicators of all monitoring channels in a consistent manner and generates a second status identifier for that moment.

[0178] The aggregation strategy prioritizes the majority consensus principle and uses stability weighting. Conflicting channels are removed or downweighted when necessary. The strategy also records the state retention duration and the most recent switch time for subsequent stability assessment.

[0179] For example, in a real-world scenario, for a 3-meter vertical spherical mirror operating at a small positive angle θ and a relatively high temperature range, within a short-term window after attitude normalization, the root mean square of the tilt increases from a low level to a significantly higher level; the root mean square of the phase perturbation jumps from near-baseline to significantly higher than baseline; cross-channel correlation gradually decreases from high consistency to moderate consistency; simultaneously, the rate of change of temperature and humidity shows an upward trend; and the volatility of the dominant term in the low-order modes remains consistently higher than a preset threshold. Based on these joint characteristics, the classification model determines that the air path is in a convective perturbation state and provides a low stability indication; the second state identifier records this state and accumulates its duration.

[0180] For example, the air path classification model adopts a two-level interpretable structure. The first level is feature standardization and weighted scoring. Features such as root mean square of tilt, root mean square of phase perturbation, coefficient of variation of light intensity fluctuation, cross-channel consistency and correlation, temperature change rate and humidity change rate, and low-order mode dominance volatility are first normalized with median and absolute deviation, and then linearly combined with the weight vector obtained from offline calibration to form a score S. The weights can be obtained from labeled data through constrained least squares or log-probability regression, and fine-tuned on-site with a small number of operating conditions. The second level is threshold determination with hysteresis and minimum duration. An upper threshold U and a lower threshold L are set. When S is greater than U, it is judged as convective disturbance; when S is less than L, it is judged as laminar corridor; and when it is between the two, it is judged as mixed zone. The hysteresis interval width and minimum duration are taken from a range that matches the on-site sampling frequency and disturbance time scale to reduce boundary jitter. The model incorporates cross-channel consistency weighting and channel health masks. When some channels are lost or marked as low quality, the weights are automatically reduced and robust aggregation is employed. The final state is determined by a weighted majority consensus principle. Window length and sliding step size, initial values ​​of the weight vector, initial suggested values ​​of U and L, hysteresis width, minimum duration, and the order of low-order modes are distributed as configurable parameters during engineering deployment. Calibration is performed on-site through a small number of comparative measurements, and the calibration process is recorded in the parameter version for auditing and backtracking.

[0181] The normalization method, feature set, classification implementation, and summary criteria mentioned above can all be adjusted according to the equipment and environment, and do not constitute a limitation on this application.

[0182] Optionally, the attitude normalization process includes performing an optical path geometric remapping operation on the air disturbance data based on the optical path geometric offset caused by the pitch or azimuth change of the spherical mirror, so as to eliminate the geometric influence of attitude change on air disturbance measurement.

[0183] The determination of the disturbance characteristic quantity includes projecting the normalized air disturbance data onto a preset low-order spatial mode basis, calculating the volatility and phase consistency index of each mode coefficient within a preset time window, and combining the volatility and phase consistency index as the disturbance characteristic quantity.

[0184] The second state identifier further includes a time sequence code for air path state switching to characterize the historical dependence of air path disturbances.

[0185] When determining the air path state solely based on raw optical path monitoring data, subtle changes in pitch and azimuth introduce geometric remapping errors, causing the same disturbance to be treated as different phenomena under different attitudes. Furthermore, it is difficult to spatially separate low-order fringe structures from random disturbances. This implementation employs a two-step process of attitude normalization and low-order modal projection, and introduces a time sequence code into the second state identifier to simultaneously address the issues of geometric consistency, feature interpretability, and history-dependent representation.

[0186] Attitude normalization is based on the geometric calibration parameters of the spherical mirror and the interferometric optical path, as well as the pitch and azimuth readings given by the attitude encoder. It calculates the lateral displacement and incident angle change of the interferometric beam on the monitoring plane at the current moment. The tilt, phase disturbance, and intensity fluctuation data collected from each monitoring channel are geometrically back-mapped and resampled according to the reference path to align them to a common reference coordinate system. To reduce phase shift caused by synchronization errors, cross-channel cross-correlation can be used to estimate small time delays and perform subsampling-level alignment after back-mapping. To reduce the impact of sensor gain inconsistencies, a linear scaling correction can be performed on the amplitude according to the reference scale. The normalized air disturbance data has cross-attitude comparability and can be used as a unified input for subsequent feature calculations.

[0187] When determining the perturbation characteristics, the system calculates two types of complementary indices on the normalized data within a preset time window. The first type is intensity and instability indices, including the root mean square of tilt and phase perturbation, the coefficient of variation of light intensity fluctuations, and multi-channel consistency and correlation.

[0188] The second category consists of spatial morphology and temporal phase indices. Several dominant mode coefficients are obtained by projecting the perturbation field onto a preset low-order spatial mode basis, and the volatility and phase consistency of each coefficient are calculated within a window. The mode basis is preferably a low-order orthogonal basis with good characterization of long-distance low-frequency fringes; the order and basis type can be configured according to site conditions. Volatility is used to quantify the temporal fluctuations of each mode intensity, and phase consistency is used to characterize the degree of temporal stability of the mode phase. After mean removal and scale normalization, the two types of indices are combined according to weights to form perturbation feature quantities, which serve as input to the air path classification model. The classification model outputs state markers such as laminar corridors, convective disturbances, or mixing zones, as well as stability indicators.

[0189] Under the same timestamp, the system aggregates the status markers of all channels to generate a second status identifier, and appends a time sequence code for air path status switching to characterize historical dependencies. The time sequence code records the state entry, maintenance, and exit events on the sliding timeline, including the event time, preceding and following states, and maintenance duration, and filters critical jitter through hysteresis and minimum duration constraints. To facilitate storage and rapid comparison, the sequence code can be digested while ensuring traceability, and the original event queue is retained to support verification.

[0190] In this way, the second state identifier is spatially independent of attitude, has interpretable low-order modal components in terms of features, and retains switching sequence information in time. It can establish a cross-index with the first state identifier on a unified time axis to identify coupling modes and drive subsequent annealing and measurement processes.

[0191] Optionally, determining the topology annealing sequence parameters based on the first state identifier and the second state identifier, and executing the topology annealing sequence, includes:

[0192] Read the coupling mode entry corresponding to the current time window from the cross state index, and generate a coupling mode description quantity to characterize the coupling relationship between the support contact state and the air path state;

[0193] Based on the coupling mode descriptor, the annealing action primitives are selected and parameterized in the controller to obtain the topology annealing control vector.

[0194] A topology annealing sequence parameter set is generated based on the topology annealing control vector, and an annealing step sequence including unlocking, micro-attitude perturbation, resting and locking is driven.

[0195] After the annealing step sequence is completed, the first state identifier and the second state identifier are updated, and the preset annealing termination condition is determined. If the condition is not met, the topology annealing control vector is regenerated based on the updated coupling mode descriptor and the process is repeated.

[0196] When annealing is performed with fixed parameters, a mismatch can easily occur where the support side has converged while the air side remains disturbed, or vice versa, leading to prolonged uncertainty or repeated oscillations in the initial state. Therefore, in this embodiment, the system does not directly use fixed parameters. Instead, it uses the coupling information generated jointly by the first and second state identifiers as a basis. The system adaptively determines the annealing action primitives and their parameters based on the coupling information, and performs a closed-loop parameter update based on the latest state identifier after each round of annealing.

[0197] In practice, the coupling mode entry corresponding to the current time window is first read from the cross-state index, and coupling mode descriptors, which characterize the coupling relationship between the support contact state and the air path state, are calculated. These coupling mode descriptors include covariance strength, time priority direction, hotspot distribution, and stability summary. Covariance strength measures the joint significance of local state changes on the support side and state changes on the air side within the same window. Time priority direction characterizes the sequential relationship between the two types of changes. Hotspot distribution is given by the first state identifier, indicating the set of regions concentrated in the light touch or departure phases, along with spatial order. The stability summary is given by the second state identifier, indicating the duration of the current air state and a consistency index. These quantities are calculated and archived on the same timeline, making the coupling relationships within the same window directly accessible input.

[0198] The controller selects and parameterizes annealed action primitives from the action primitive library based on the coupling mode descriptor, resulting in a topology annealed control vector. The action primitive library must contain at least the attitude jitter direction, attitude jitter amplitude, jitter rhythm, rest duration, and locking sequence. Selection and parameterization follow two rules.

[0199] For example, the first rule is to prioritize small-step jitter along the coupling priority direction to reduce the risk of remapping, use smaller amplitude and longer rest when the covariance intensity is high, and enable hot spot priority locking sequence.

[0200] For example, the second rule is to reduce the shaking rhythm and extend the static period when the air-side stability is low, and to adopt a locking strategy of first non-hot spots and then hot spots when the support side hot spots are concentrated, and to extend the static period in the hot spot area.

[0201] For example, in a scenario with a 3-meter vertical spherical mirror, θ = +1°, and a relatively high temperature, if the cross-state index shows a hotspot at the lower edge and the air is disturbed, the controller can select a negative pitch perturbation with an initial amplitude of 5″ to 10″, a rhythm of 0.3 to 0.5 seconds per cycle, a rest period of 3 to 5 seconds, and a locking sequence of first the upper edge and then both sides, followed by the lower edge. The above ranges can be adjusted according to the device resolution and the ambient disturbance.

[0202] The system generates a specific set of topology annealing sequence parameters based on the topology annealing control vector, driving the sequence of steps for unlocking, micro-attitude perturbation, resting and locking. During execution, the first state identifier and the second state identifier are refreshed in real time at the end of each sub-stage.

[0203] If the first state indicator shows that the hotspot area is carried by the return of light touch and no new diffusion occurs, and the second state indicator shows that the air path remains in a stable set and meets the minimum duration, then the round step is considered valid and the next sub-stage is entered; if either side is still unstable, the amplitude and rhythm are attenuated by a preset factor without changing the direction, and the resting time is extended before repeating the sub-stage until the maximum number of attempts for the current round is reached or the stability condition is met.

[0204] For example, within one cycle, if the air stability is still low and the covariance intensity has not dropped below the threshold, the amplitude decreases from 10″ to 7″ and then to 5″, the rhythm increases from 0.3 s to 0.5 s, the rest period increases from 3 s to 5 s, and the hot spot locking sequence strategy is maintained; conversely, if the air is stable but the support still shows local light touches, the rhythm remains unchanged, moderately increased to 12″, and the rest period is shortened to 2 s to promote force chain redistribution, but the locking sequence still follows non-hot spots first and then hot spots.

[0205] After the annealing step sequence is completed, the system determines the preset annealing termination condition based on the updated first and second state identifiers.

[0206] For example, termination conditions may include the covariance intensity being below a threshold, both states remaining stable within the same decision window without new hotspots, and the parameter convergence of this round meeting the requirements. If the termination conditions are met, an initial state confirmation signal is output and formal measurement begins; if not, the topological annealing control vector is regenerated using the latest coupling mode descriptor as input and the next round sequence begins.

[0207] Through the above-mentioned process of adaptive parameter selection based on coupling mode, fine-tuning based on sub-stage feedback, and closed-loop determination based on termination conditions, annealing promotes the coordinated convergence of the support and air sides with a smaller disturbance energy level, reducing the risk of excessive intervention and pseudo-stable release caused by fixed parameters, while being able to stably generate reproducible initial states in typical field scenarios.

[0208] Optionally, the coupling mode description includes coupling correlation strength and coupling preference direction;

[0209] The annealing action primitive includes at least the attitude jitter direction, attitude jitter amplitude, jitter rhythm, rest duration and locking sequence, and identifies the set of coupled hot spot support points through the state mapping table of the first state machine to determine that the locking sequence is non-hot spot support points first and then hot spot support points.

[0210] The sampling triggering of the annealing step sequence is limited by the second state identifier to be activated when the air path meets the preset stability criterion;

[0211] When the coupling correlation strength does not meet the threshold, the attitude jitter amplitude and jitter rhythm are adjusted by a preset attenuation along the coupling priority direction before entering the next annealing step sequence.

[0212] In one implementation, the coupling mode descriptor consists of coupling correlation strength and coupling preference direction. Coupling correlation strength measures the joint significance of support-side state changes and air-side state changes within the same time window. It can be obtained by weighting the co-occurrence rate, cross-window consistency, and low-order statistical correlation of the two state mapping tables. The value is between zero and one, with a larger value indicating a stronger correlation.

[0213] Among them, the coupling priority direction is used to characterize the sequential relationship between the two types of changes. It can be obtained from the time lag corresponding to the cross-correlation peak of the first state identifier and the second state identifier. Then, combined with the geometric mapping of the device, the lag sign and magnitude are mapped to the positive or negative direction of the attitude control axis.

[0214] For example, if the hysteresis is supported first and the amplitude falls within a short time range, the priority direction is set to a small step movement in the negative pitch direction. Both of these quantities are calculated and stored within each decision window, serving as direct inputs for parameter selection.

[0215] In practical implementation, the action primitive library includes at least the attitude jitter direction, attitude jitter amplitude, jitter rhythm, rest duration, and locking sequence. The attitude jitter direction is either consistent with or opposite to the coupling priority direction; the default is consistent to reduce remapping risk. The attitude jitter amplitude and jitter rhythm are jointly determined by the coupling correlation strength and air-side stability. A smaller amplitude and slower rhythm are used when the correlation is strong, while a slightly larger amplitude is allowed when the correlation is weak to accelerate force chain redistribution. The rest duration is set to be longer or shorter depending on the air-side stability; a longer duration is used when the stability is low to wait for the air field to converge. The locking sequence identifies the set of coupling hotspot support points based on the state mapping table of the first state machine. The hotspot set is defined as the area where touch or leave events occur in clusters in space. The sequence strategy is to prioritize non-hotspots over hotspots to reduce the probability of secondary switching.

[0216] Taking a 3 m vertical spherical mirror as an example, when the lower edge is identified as a hotspot and the air stability is low, the sequence usually selects small-step jitter in the negative pitch direction, with the amplitude falling at the low end of the several-second angle level, the rhythm slowing down to the sub-second level, the static time being several seconds, the upper edge and both sides being locked first, and the lower edge being locked later.

[0217] The sampling triggering of the annealing step sequence is limited by a second state identifier, which indicates that the air path meets a preset stability criterion. The stability criterion requires that the air path state be in a stable set and maintain a minimum duration, while cross-channel consistency and correlation must not fall below the lower limit. The sequence performs one or more samplings within the trigger window. If the second state identifier changes during this period, the current sample is discarded and the process reverts to the annealing cycle. This gating strategy ensures that the data used for decision-making and quality control is consistent with the stable periods of the air field, reducing misjudgments caused by convective plumes from the source.

[0218] When the coupling strength does not meet the threshold, the attitude jitter amplitude and jitter rhythm are adjusted according to a preset decay along the coupling priority direction before proceeding to the next annealing step sequence. The decay strategy requires the amplitude to decrease monotonically until it reaches the allowable lower limit, the rhythm to slow down monotonically or remain unchanged, and the resting time to be extended as needed.

[0219] Through the above-mentioned parameter selection based on coupling quantity, sampling with second state identifier as gate, and iterative mechanism constrained by decay law, the annealing process can drive the two states to converge together with a small disturbance energy level under typical field conditions, and provide repeatable preconditions for initial state confirmation.

[0220] Optionally, after the initial state confirmation signal is generated, the controller closes the write channels of the attitude actuator and the locking actuator, leaving only the read-only monitoring and measurement trigger channels active. Simultaneously, it continuously reads the first and second state identifiers along a unified timeline. If a switch in either state identifier is detected during sampling, the current sampling process is immediately terminated, the cached measurement frames for this round are discarded, and the system reverts to the initial state establishment process, recording the trigger event and timestamp for subsequent parameter tuning and verification. This freeze and rollback mechanism ensures consistency between the formal measurement and the initial steady state, avoiding the generation of incomparable data during state drift.

[0221] Formal measurements employ a dual sampling strategy gated by the second state identifier. The system triggers sampling only within the time window where the second state identifier indicates that the air path satisfies a preset stability criterion, and acquires the first and second interferometric measurement frames within adjacent time slots. The span and sampling rhythm of adjacent time slots are configured based on the interferometer readout speed and the air field time scale, forming tightly coupled frame pairs.

[0222] It is important to note that, in order to form a candidate frame pair, the second state identifiers of the two frames must have modal coefficient volatility below a threshold and identical time sequence codes. If either condition is not met, the system discards the pair and waits for the next stable time slot. This gating and dual sampling design is used to acquire data from two frames under similar conditions during periods of slow rather than abrupt changes in the air field, providing stable input for subsequent registration and subtraction.

[0223] For candidate frame pairs that pass the gated check, the system performs back-mapping registration based on optical path geometry. Registration targets the initial state reference coordinate system and, based on attitude encoder readings and optical path calibration parameters, back-maps the interferometric phase fields of the two frames to the reference geometry, eliminating spatial misalignment caused by minute pitch or azimuth errors. Subsequently, cross-channel cross-correlation is used to estimate minute time delays and perform subsampling-level phase alignment. If necessary, linear scaling correction is performed according to the reference amplitude, thereby aligning the two frames spatially and amplitude-wise. The registration residuals and alignment quality are recorded as quality control quantities and written into metadata for subsequent screening and traceability.

[0224] After registration, the system performs air component subtraction on the candidate frame pair based on the air disturbance mode coefficients recorded in the second state identifier. Specifically, the phase fields of the two frames are projected onto a preset low-order spatial mode basis, the dominant air disturbance mode coefficients and stability indicators of the corresponding time slots in the second state identifier are read, and the air mode components of the two frames are jointly estimated according to robust weights and subtracted from their respective phase fields to obtain the air-free surface-specific wavefront data. To suppress over-subtraction or under-subtraction, hysteresis and upper limit pruning can be set for the weights and thresholds, and the registration residual and stability indicator are used to jointly determine whether to accept the result. If the quality control quantity does not meet the standard, the candidate frame pair is discarded and waits for the next stable time slot.

[0225] The surface-specific wavefront data, along with the timestamp of this measurement, attitude readings, lockout snapshot, environmental parameters, and corresponding first and second state identifier summaries, are written into the measurement data buffer and used as direct input for generating evaluation index data. To ensure comparability, the system preferably accumulates a number of qualified surface-specific wavefront data within a single stability window, performs spatial registration consistency verification and robust synthesis, and the synthesis method can be simple averaging, stability-weighted averaging, or robust estimation with outlier suppression as the objective. The weights and elimination criteria used are consistent with the stability indication of the second state identifier.

[0226] In this way, the synthesized results serve as representative wavefront data specific to the surface pattern under this working condition, while also outputting quality markers and confidence indices, providing a reliable basis for subsequent cross-working-condition stability calculations and report generation.

[0227] The aforementioned freezing and rollback strategies, gating and dual sampling calibers, back-mapping registration methods, air component subtraction weight settings, and quality control thresholds can be adjusted according to the equipment resolution and on-site disturbance characteristics, and do not constitute a limitation of this application.

[0228] Based on the same inventive concept, this application also provides a spherical mirror multi-condition surface stability measurement system corresponding to the spherical mirror multi-condition surface stability measurement method. Since the principle of the system in this application is similar to the spherical mirror multi-condition surface stability measurement method described above in this application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0229] Reference Figure 4 The diagram shown is a schematic of a spherical mirror multi-condition surface stability measurement system provided in an embodiment of this application. The system includes:

[0230] Module 10 is used to construct a two-layer state machine, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state.

[0231] The acquisition module 20 is used to acquire load sensing data related to the support contact state and optical path monitoring data related to the air path state.

[0232] Processing module 30 is configured to generate a first state identifier representing the support contact state based on the load sensing data; generate a second state identifier representing the air path state based on the optical path monitoring data; determine the topology annealing sequence parameters based on the first state identifier and the second state identifier, and execute the topology annealing sequence.

[0233] The update module 40 is used to update the first state identifier and the second state identifier in real time during the execution of the topology annealing sequence, and generate an initial state confirmation signal when both of them simultaneously meet their respective preset stability conditions.

[0234] The generation module 50 is used to trigger formal interferometric measurement based on the initial state confirmation signal, obtain the surface shape data of the spherical mirror under the current working condition, and generate evaluation index data to characterize the surface shape stability of the spherical mirror under multiple working conditions based on the surface shape data and the corresponding first state identifier and second state identifier.

[0235] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for measuring the surface stability of a spherical mirror under multiple working conditions, characterized in that, include: A two-layer state machine is constructed, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state. Collect load sensing data related to the support contact state and optical path monitoring data related to the air path state; Based on the load sensing data, a first state identifier representing the support contact state is generated. Based on the optical path monitoring data, a second state identifier representing the air path state is generated; Based on the first state identifier and the second state identifier, the topology annealing sequence parameters are determined, and the topology annealing sequence is executed; During the execution of the topology annealing sequence, the first state identifier and the second state identifier are updated in real time, and an initial state confirmation signal is generated when both simultaneously meet their respective preset stability conditions. Based on the initial state confirmation signal, a formal interferometric measurement is triggered to obtain the surface shape data of the spherical mirror under the current working condition. Based on the surface shape data and the corresponding first state identifier and second state identifier, evaluation index data for characterizing the surface shape stability of the spherical mirror under multiple working conditions is generated.

2. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 1, characterized in that, The construction of the two-level state machine includes: Based on the output data of multiple load sensors of the multi-point support mechanism, the contact micro-state characteristics of each support point are determined, and clustering operation is performed on the contact micro-state characteristics to generate the state identifier of each support point and form the state mapping table of the first state machine. Based on the output data of the optical path peripheral monitoring channel and environmental monitoring sensor of the interferometric measurement component, the disturbance characteristics of the air path are determined, and the disturbance characteristics are classified and calculated to generate the state mapping table of the second state machine. Establish a cross-state index between the first state machine and the second state machine to identify the coupling mode descriptor between the support contact state and the air path state.

3. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 2, characterized in that, The contact micro-state characteristics are a feature vector composed of the load fluctuation of the load sensor and the displacement response of the support point within a continuous time window. The disturbance characteristic is the modal coefficient fluctuation rate after the air path disturbance amount within a continuous time window is projected onto a preset low-order spatial modal basis. The state mapping tables of the first state machine and the second state machine respectively record the state switching order and generate path history codes. The cross-state index is established by associating the path history codes of the first state machine and the second state machine.

4. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 1, characterized in that, The generation of a first state identifier characterizing the support contact state based on the load sensing data includes: Within a preset time window, the load average value, load fluctuation amplitude, and load change rate are calculated from the load sensing data of each support point, and the above three quantities are combined as the load feature vector of the support point. Each load feature vector is normalized according to the spatial distribution of support points to obtain a relative load feature vector after eliminating the influence of overall temperature drift, and the offset from the historical baseline is calculated in the relative load feature vector. The offset is input into a micro-state classification model based on support point partitioning. In the model, the displacement response data or micro-vibration response data of the support point are combined to generate state labels of bearing state, light touch state or departure state. The state markers of all support points are combined to form a first state identifier, and the time sequence code of the state switching of each support point is appended to the first state identifier to characterize the path dependency feature of the support contact topology.

5. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 1, characterized in that, The generation of a second state identifier characterizing the air path state based on the optical path monitoring data includes: The air disturbance data collected by the peripheral monitoring channel of the interferometric measurement component is subjected to attitude normalization processing to obtain normalized air disturbance data; Based on the normalized air disturbance data, disturbance feature quantities are determined, and the disturbance feature quantities are input into the air path classification model to generate air path state labels. The status markers of the air path are combined to form a second status identifier.

6. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 5, characterized in that, The attitude normalization process includes performing an optical path geometric remapping operation on the air disturbance data based on the optical path geometric offset caused by the pitch or azimuth change of the spherical mirror, so as to eliminate the geometric influence of attitude change on air disturbance measurement. The determination of the disturbance characteristic quantity includes projecting the normalized air disturbance data onto a preset low-order spatial mode basis, calculating the volatility and phase consistency index of each mode coefficient within a preset time window, and combining the volatility and phase consistency index as the disturbance characteristic quantity. The second state identifier further includes a time sequence code for air path state switching to characterize the historical dependence of air path disturbances.

7. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 2, characterized in that, The step of determining the topology annealing sequence parameters based on the first state identifier and the second state identifier, and then executing the topology annealing sequence, includes: Read the coupling mode entry corresponding to the current time window from the cross state index, and generate a coupling mode description quantity to characterize the coupling relationship between the support contact state and the air path state; Based on the coupling mode descriptor, the annealing action primitives are selected and parameterized in the controller to obtain the topology annealing control vector. A topology annealing sequence parameter set is generated based on the topology annealing control vector, and an annealing step sequence including unlocking, micro-attitude perturbation, resting and locking is driven. After the annealing step sequence is completed, the first state identifier and the second state identifier are updated, and the preset annealing termination condition is determined. If the condition is not met, the topology annealing control vector is regenerated based on the updated coupling mode descriptor and the process is repeated.

8. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 7, characterized in that, The coupling mode description includes coupling correlation strength and coupling preference direction; The annealing action primitive includes at least the attitude jitter direction, attitude jitter amplitude, jitter rhythm, rest duration and locking sequence, and identifies the set of coupled hot spot support points through the state mapping table of the first state machine to determine that the locking sequence is non-hot spot support points first and then hot spot support points. The sampling triggering of the annealing step sequence is limited by the second state identifier to be activated when the air path meets the preset stability criterion; When the coupling correlation strength does not meet the threshold, the attitude jitter amplitude and jitter rhythm are adjusted by a preset attenuation along the coupling priority direction before entering the next annealing step sequence.

9. The method for measuring the surface stability of a spherical mirror under multiple working conditions according to claim 1, characterized in that, The formal interferometric measurement triggered based on the initial state confirmation signal includes: After the initial state confirmation signal is generated, the control channels of the attitude actuator and the locking actuator are frozen, and the first state identifier and the second state identifier are continuously read during the sampling period. When a switch of either state identifier is detected, the sampling is stopped and the process is rolled back to the initial state establishment process. Using the second state identifier as the sampling gate, a dual sampling strategy is adopted to acquire the first interferometric measurement frame and the second interferometric measurement frame in adjacent time slots. The modal coefficient fluctuation rate of the second state identifier is required to be lower than a preset threshold and its time sequence code is consistent, so as to form a candidate frame pair. Perform a back-mapping registration operation based on optical path geometry on the candidate frame pair to unify the two frames of interferometric measurement data into the initial state reference coordinate system; Based on the air disturbance mode coefficients recorded by the second state identifier, the interferometric measurement data of the candidate frame pair is processed by air component subtraction to obtain surface-specific wavefront data, and the surface-specific wavefront data is written into the measurement data buffer as input to generate evaluation index data.

10. A multi-condition surface stability measurement system for spherical mirrors, characterized in that, include: A construction module is used to construct a two-layer state machine, which includes a first state machine for characterizing the contact state of the spherical mirror support and a second state machine for characterizing the air path state. The acquisition module is used to acquire load sensing data related to the support contact state and optical path monitoring data related to the air path state. The processing module is used to generate a first state identifier representing the support contact state based on the load sensing data. Based on the optical path monitoring data, a second state identifier representing the air path state is generated; Based on the first state identifier and the second state identifier, the topology annealing sequence parameters are determined, and the topology annealing sequence is executed; The update module is used to update the first state identifier and the second state identifier in real time during the execution of the topological annealing sequence, and generate an initial state confirmation signal when both of them simultaneously meet their respective preset stability conditions. The generation module is used to trigger formal interferometric measurement based on the initial state confirmation signal, obtain the surface shape data of the spherical mirror under the current working condition, and generate evaluation index data to characterize the surface shape stability of the spherical mirror under multiple working conditions based on the surface shape data and the corresponding first state identifier and second state identifier.

Images