A method for marine sensor fault self-checking and calibration

Abstract

The application discloses a kind of marine sensor fault self-checking and calibration method, it is related to marine sensor monitoring technical field, comprising: obtaining the historical health data of target sensor and auxiliary sensor, historical health sea state is divided and historical sea state process is constructed, health response range and health response trajectory are established;According to current monitoring data matching historical health sea state, construct current sea state process and determine sea state state deviation, response span deviation and process response deviation;According to the deviation, identify zero drift, sensitivity drift and dynamic response anomaly;For zero drift and sensitivity drift, determine calibration control quantity and update calibration parameter, for dynamic response anomaly, output maintenance information;Calibration result is rechecked using subsequent matching data.The application can reduce the interference of real sea state change on fault judgment, improve the fault identification and calibration reliability in long-term in-situ operation of marine sensor.

Description

Technical Field

[0001] This invention relates to the field of marine sensor monitoring technology, specifically to a method for self-diagnosis and calibration of marine sensor faults. Background Technology

[0002] In marine environmental monitoring, water quality sensors such as pH, dissolved oxygen, conductivity, and turbidity are usually deployed on buoys, shore-based monitoring stations, or underwater platforms for a long time to continuously acquire data on changes in marine water quality. Due to long-term immersion in complex seawater environments, the sensors are easily affected by factors such as biological attachment, aging of sensitive elements, temperature and salinity changes, and device performance degradation, which causes the measurement performance to change over time. Currently, marine water quality sensors are mainly maintained through methods such as regular salvage, manual cleaning, standard solution calibration, or laboratory comparison. Some online monitoring systems also use fixed thresholds, historical measurement data, data from adjacent sensors, or soft measurement results to identify anomalies and compensate based on measurement deviations. The above-mentioned methods can detect some measurement anomalies, but their judgment usually relies on time series changes or single-moment measurement differences. When real sea conditions change, such as tides and water mass exchanges, the duration and rate of change of different sea state transitions are not fixed. Directly comparing based on time or measured values ​​can easily confuse real environmental changes with sensor anomalies. At the same time, zero drift, sensitivity drift, and slow response have different response characteristics, and using a uniform deviation compensation method may lead to misjudgment of fault types and incorrect calibration. Therefore, existing technologies cannot simultaneously achieve fault type identification, reliable calibration, and calibration result verification under long-term in-situ operation of sensors. Based on this, it is necessary to propose a self-testing and calibration method for marine sensors. Summary of the Invention

[0003] To address the aforementioned shortcomings in the prior art, this invention provides a method for self-diagnosis and calibration of marine sensor faults.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for self-diagnosing and calibrating faults in marine sensors, comprising: S1. Acquire historical health data of the target sensor and auxiliary sensor, divide historical health sea conditions according to the auxiliary sensor data, construct historical sea condition process based on the changes between different historical health sea conditions, and establish the health response range of the target sensor and the health response trajectory that changes with the historical sea condition process. S2. Acquire current monitoring data, determine the current sea state based on auxiliary sensors and match historical healthy sea states, compare the current response of the target sensor with the corresponding healthy response range to obtain the sea state deviation; when the current sea state changes, construct the current sea state process, compare it with the healthy response trajectory according to the same sea state process position to obtain the response span deviation and process response deviation. S3. Based on the sea state deviation, and in conjunction with the response span deviation and process response deviation at the time of the current sea state transition, determine the fault type, which includes zero drift, sensitivity drift and dynamic response anomaly. S4. When the fault type is either zero-point drift or sensitivity drift, determine the calibration control quantity based on the corresponding health response and the current response, and update the calibration parameters of the target sensor; when the fault type is abnormal dynamic response, maintain the current calibration parameters and output maintenance information. S5. After the calibration parameters are updated, at least one of the following should be selected from the historical healthy sea state and sea state conversion process for retesting: if the retest passes, the updated calibration parameters should be retained; if the retest fails, the parameters before calibration should be restored.

[0005] Preferably, in step S1, classifying historical sea state health based on auxiliary sensor data includes: Extract continuous monitoring windows from historical data where both the target sensor and auxiliary sensors are in a healthy state; The average values ​​and variation ranges of each auxiliary sensor within the continuous monitoring window are used to form sea state characteristics, and the continuous monitoring windows are classified according to the similarity between the sea state characteristics. The marine environmental state corresponding to the same type of continuous monitoring window is determined as the same historical healthy sea state, and the corresponding health response range is determined based on the target sensor data under the historical healthy sea state.

[0006] Preferably, in step S1, constructing a historical sea state process based on changes between different historical healthy sea states includes: Determine the first historical sea state and the second historical sea state, and obtain the status values ​​of each auxiliary sensor under the first historical sea state and the second historical sea state; For the historical monitoring process of transitioning from the first historical sea state to the second historical sea state, the historical individual process value is determined according to the degree of change of the monitoring value of each auxiliary sensor relative to the two state values ​​during the transition process. Historical single-item process values ​​obtained when the direction of change of the corresponding auxiliary sensor during the historical monitoring process is inconsistent with the direction of change determined by its two state values ​​are excluded. The remaining historical single-item process values ​​are then fused to obtain the historical sea state process that characterizes the degree of sea state change during the historical monitoring process.

[0007] Preferably, in step S1, establishing the health response trajectory that changes with the historical sea state process includes: Determine the first and second historical healthy sea states from historical healthy sea states, obtain the historical health monitoring process of multiple transitions from the first historical healthy sea state to the second historical healthy sea state, and align the process positions according to the corresponding historical sea state process. The target sensor health responses corresponding to each sea state process position are statistically analyzed, the process health response value of each sea state process position is determined, and the health response trajectory is formed according to the chronological relationship of the sea state process positions.

[0008] Preferably, in step S2, obtaining the sea state deviation, response span deviation, and process response deviation includes: When the current response of the target sensor exceeds the health response range corresponding to the matched historical healthy sea state, the sea state deviation is determined based on the directed difference between the current response and the exceeded boundary value. When the current response is within the health response range, the sea state deviation is determined to be zero. The response span deviation is determined based on the difference between the change in the target sensor response before and after the current sea state change and the change in the response at both ends of the health response trajectory. The current response trajectory is formed by combining the current responses of the target sensors at each sea state process position during the current sea state transition. The current response trajectory is then translated and scaled to align its two ends with the two ends of the healthy response trajectory. The response differences between the aligned current response trajectory and the healthy response trajectory at the same sea state process position are compared to obtain the process response deviation corresponding to each sea state process position.

[0009] Preferably, step S3 includes: During the current sea state transition, when the process response deviation exceeds the dynamic response allowable range at multiple consecutive sea state process positions, it is determined to be a dynamic response anomaly; During the current sea state transition, if the dynamic response anomaly judgment conditions are not met, when the response span deviation exceeds the allowable span range, it is determined to be sensitivity drift. If the target sensor matches the same historical healthy sea state in multiple consecutive monitoring windows and produces a sea state deviation in the same direction, it is determined to be zero drift when the current sea state has not changed. When the current sea state changes, if the target sensor produces a sea state deviation in the same direction under the historical healthy sea state before and after the current sea state change, and neither the dynamic response anomaly judgment condition nor the sensitivity drift judgment condition is met, it is determined to be zero drift.

[0010] Preferably, in step S4, determining the calibration control quantity based on the corresponding health response and the current response includes: When zero-point drift is determined, a health response reference value is determined based on the upper and lower boundaries of the health response range corresponding to the matched historical healthy sea state. The zero-point calibration control value is determined based on the offset between the health response reference value and the current response, and the zero-point calibration parameters are updated. When sensitivity drift is identified, the sensitivity calibration control quantity is determined based on the proportional relationship between the target sensor response span during the current sea state transition and the response change amplitude at both ends of the healthy response trajectory, and the sensitivity calibration parameters are updated.

[0011] Preferably, in step S4, when the fault type is dynamic response anomaly, the current calibration parameters of the target sensor are maintained, the abnormal state of the current output data of the target sensor is marked, and maintenance information characterizing the dynamic response anomaly of the target sensor is generated.

[0012] Preferably, in step S5, at least one of the following is selected for re-examination from the subsequently matched historical sea state and sea state transition process: Select subsequent matching data that were not involved in the determination of the calibration control quantities in this study; When the calibration parameter updated this time is the zero-point calibration parameter, the sea state deviation is re-determined using the historical healthy sea state of subsequent matching, and the re-determined sea state deviation is used as the zero-point re-examination deviation. When the calibration parameters updated this time are sensitivity calibration parameters, the response span deviation and process response deviation are re-determined using the subsequent matched sea state conversion process, and the re-determined response span deviation and process response deviation are combined into a sensitivity re-examination deviation group. When the updated calibration parameter is the zero-point calibration parameter and the zero-point re-inspection deviation is within the corresponding allowable range, the re-inspection is deemed successful and the updated calibration parameter is maintained. When the updated calibration parameter is the sensitivity calibration parameter and all deviations in the sensitivity retest deviation group are within the corresponding allowable range, the retest is deemed passed and the updated calibration parameter is maintained.

[0013] Except for the above-mentioned cases where the re-inspection passed, if the re-inspection fails, the parameters before calibration should be restored.

[0014] Preferably, after the re-inspection is passed, the procedure further includes: Set the target sensor data and the synchronously collected auxiliary sensor data when the re-inspection is passed as the health data to be confirmed; Select another set of subsequent matching data that did not participate in the previous re-examination and re-examine it. If the re-examination passes, add the health data to be confirmed to the historical health data to update at least one of the health response range and health response trajectory that match the health data to be confirmed.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a sea state process using auxiliary sensors and compares the current response of the target sensor with its historical health response at the same sea state process position. This solves the problem of difficulty in directly aligning responses caused by different durations and rates of change in different tidal and water mass exchange processes, reducing the interference of real sea state changes on fault diagnosis. By classifying the target sensor based on sea state deviation, response span deviation, and process response deviation, it addresses the problem of easy confusion between zero-point drift, sensitivity drift, and dynamic response anomalies, improving the specificity of fault type identification and avoiding incorrect compensation for dynamic response anomalies using zero-point or sensitivity parameters. Furthermore, by implementing in-situ calibration or maintenance based on the fault type and using independent follow-up data to re-check and reconfirm the calibration results, it solves the problem of lack of verification of calibration results and the contamination of historical health benchmarks by abnormal data during long-term operation. This improves the reliability of fault self-checking and calibration of marine sensors under long-term unattended conditions and reduces the frequency of manual salvage and maintenance. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, a method for self-diagnosing and calibrating faults in marine sensors includes: S1. Acquire historical health data of the target sensor and auxiliary sensor, divide historical health sea conditions according to the auxiliary sensor data, construct historical sea condition process based on the changes between different historical health sea conditions, and establish the health response range of the target sensor and the health response trajectory that changes with the historical sea condition process. In this embodiment, the initial historical health data consists of continuous monitoring data obtained after the target sensor and auxiliary sensor have undergone their most recent manual calibration, maintenance, or standard sample comparison and confirmed to be functioning correctly. During this continuous monitoring period, neither the target sensor nor the auxiliary sensor generated any equipment fault codes, communication anomalies, or over-range conditions. Subsequent data confirmed to be normal by this method can be added to the historical health data.

[0019] In this embodiment, a pH sensor is used as the target sensor, and a temperature sensor, a salinity sensor, a pressure sensor, and a dissolved oxygen sensor are used as auxiliary sensors. The target sensor and auxiliary sensors simultaneously collect monitoring data, with a sampling period of 1 minute, and 30 consecutive sampling points constitute a 30-minute monitoring window.

[0020] In this embodiment, the monitoring window is composed of sampling points that are continuous in time. The continuous monitoring window will be referred to as the monitoring window below.

[0021] In step S1, historical sea condition health is classified based on auxiliary sensor data, including: Extract continuous monitoring windows from historical data where both the target sensor and auxiliary sensors are in a healthy state; The average values ​​and variation ranges of each auxiliary sensor within the continuous monitoring window are used to form sea state characteristics, and the continuous monitoring windows are classified according to the similarity between the sea state characteristics. The marine environmental state corresponding to the same type of continuous monitoring window is determined as the same historical healthy sea state, and the corresponding health response range is determined based on the target sensor data under the historical healthy sea state.

[0022] Specifically, the average values ​​of temperature, salinity, pressure, and dissolved oxygen are calculated for each 30-minute monitoring window. At the same time, the difference between the maximum and minimum values ​​of each auxiliary sensor within the window is calculated. The average values ​​and the range of change of these values ​​together constitute the sea state characteristics of the corresponding monitoring window.

[0023] Since temperature, salinity, pressure, and dissolved oxygen have different dimensions, each type of characteristic is normalized before comparing sea state characteristics. In this embodiment, the normalization interval is determined based on the 5th percentile and 95th percentile of the corresponding characteristic in the initial historical health data, and the corresponding characteristic is converted to the range of 0 to 1.

[0024] Specifically, when the corresponding eigenvalue is not higher than the 5th percentile, the normalization result is set to 0; when the corresponding eigenvalue is not lower than the 95th percentile, the normalization result is set to 1; when the corresponding eigenvalue is between the 5th and 95th percentiles, linear normalization is performed according to its position relative to the two percentiles. When the 5th and 95th percentiles of the corresponding eigenvalue are the same, the eigenvalue is not used in the calculation of the sea state characteristic distance.

[0025] After normalization, the differences between the corresponding sea state characteristics of the two monitoring windows are calculated item by item. The average of the squared differences is then taken and the square root is taken to obtain the sea state characteristic distance between the two monitoring windows.

[0026] In this embodiment, monitoring windows with a sea state characteristic distance of no more than 0.15 are grouped into the same group. Specifically, during classification, the earliest unclassified monitoring window can be selected as an initial category. The category center is determined based on the average sea state characteristic of the existing monitoring windows in the initial category, and other monitoring windows with a sea state characteristic distance of no more than 0.15 from the category center are added to the category. After each addition, the category center is recalculated until no new monitoring windows meet the addition criteria. Then, the same process is performed on the remaining unclassified monitoring windows.

[0027] Monitoring windows within the same category represent marine environmental states with similar combinations of temperature, salinity, pressure, and dissolved oxygen, and these marine environmental states are identified as the same historical healthy sea state.

[0028] For each historical healthy sea state, the health response of the target pH sensor within the corresponding monitoring window is statistically analyzed. To reduce the impact of occasional fluctuations on the health response range, this implementation method takes the 5th percentile value of the target sensor's health response as the lower boundary of the health response range and the 95th percentile value as the upper boundary of the health response range.

[0029] For example, in one specific embodiment, the 5th percentile value of the pH health response corresponding to a certain type of historical healthy sea state is 8.09 and the 95th percentile value is 8.15, then the corresponding health response range is 8.09 to 8.15.

[0030] In step S1, the historical sea state process is constructed based on the changes between different historical healthy sea states, including: Determine the first historical sea state and the second historical sea state, and obtain the status values ​​of each auxiliary sensor under the first historical sea state and the second historical sea state; For the historical monitoring process of transitioning from the first historical sea state to the second historical sea state, the historical individual process value is determined according to the degree of change of the monitoring value of each auxiliary sensor relative to the two state values ​​during the transition process. Historical single-item process values ​​obtained when the direction of change of the corresponding auxiliary sensor during the historical monitoring process is inconsistent with the direction of change determined by its two state values ​​are excluded. The remaining historical single-item process values ​​are then fused to obtain the historical sea state process that characterizes the degree of sea state change during the historical monitoring process.

[0031] Specifically, for the same historical healthy sea state, the median value of the average value of multiple monitoring windows of each auxiliary sensor is taken as the state value of the corresponding auxiliary sensor under the historical healthy sea state, so as to reduce the impact of abnormal fluctuations in individual windows.

[0032] For the i-th auxiliary sensor, the short-term fluctuation standard deviation is determined based on the sampled data in each monitoring window of the historical healthy sea state, and the median value of the short-term fluctuation standard deviation corresponding to multiple historical healthy monitoring windows is taken as the historical healthy sea state short-term fluctuation standard deviation of the auxiliary sensor.

[0033] When the difference between the state values ​​of the auxiliary sensor between the first historical healthy sea state and the second historical healthy sea state is less than three times the short-term fluctuation standard deviation, it is considered that the change of the auxiliary sensor is insufficient to reliably reflect the progress of the current sea state transition. It is retained for sea state matching, but its historical individual progress value is not calculated.

[0034] When the difference in state values ​​is not less than three times the standard deviation of the short-term fluctuations, let the state value of the i-th auxiliary sensor under the first historical healthy sea state A be... The state value under the second historical healthy sea state B is The monitoring value at the k-th sampling time during the historical monitoring process is Then the historical single process value is: In the formula, Let be the historical single-item process value of the i-th auxiliary sensor at the k-th sampling time; the first historical healthy sea state A corresponds to sea state process 0, and the second historical healthy sea state B corresponds to sea state process 1; when the calculated historical single-item process value is less than 0, it is taken as 0, and when it is greater than 1, it is taken as 1.

[0035] For example, when the monitoring device is fixed at the same water depth, the pressure change between the first historical healthy sea state and the second historical healthy sea state may be very small. In this case, the pressure can be used to determine whether the monitoring device is still at the same deployment depth, but it is not used as the main parameter for judging the progress of water mass conversion.

[0036] Furthermore, the actual direction of change of each auxiliary sensor involved in the calculation of sea state process is determined.

[0037] For example, when the temperature is 18.2℃ under the first historical healthy sea state A and 15.6℃ under the second historical healthy sea state B, its normal transition direction should be downward. If the monitoring values ​​of the auxiliary sensor at three consecutive sampling times all change in the opposite direction to the change direction determined by the two state values, and the magnitude of the reverse change between the first and last monitoring values ​​at the three sampling times exceeds three times the standard deviation of the short-term fluctuation of the corresponding historical healthy sea state, then the historical single-item process values ​​corresponding to the three sampling times are excluded.

[0038] The remaining valid historical individual process values ​​are fused using the median to obtain the historical sea state process value at the corresponding sampling time. These historical sea state process values ​​are then arranged in chronological order of sampling time to form a historical sea state process characterizing the degree of sea state change during the historical monitoring process. Using the median for fusion avoids significant shifts in the historical sea state process caused by short-term anomalies in a single auxiliary sensor.

[0039] In this embodiment, at least two auxiliary sensors are required to provide valid individual process values. When the number of valid auxiliary sensors is less than two, the corresponding conversion process is not used as the healthy sea state conversion template.

[0040] For the historical sea state process, the median of three consecutive sampling points can be used to smooth each historical sea state process value to reduce small process regressions caused by sampling noise. When the smoothed historical sea state process value decreases sequentially over three consecutive sampling times, and the historical sea state process regression between the first and third sampling times exceeds 0.05, the corresponding historical monitoring process is considered not a valid transition from the first historical healthy sea state to the second historical healthy sea state, and it is not used to establish a health response trajectory.

[0041] In step S1, the establishment of the health response trajectory that changes with the historical sea state process includes: Determine the first and second historical healthy sea states from historical healthy sea states, obtain the historical health monitoring process of multiple transitions from the first historical healthy sea state to the second historical healthy sea state, and align the process positions according to the corresponding historical sea state process. The target sensor health responses corresponding to each sea state process position are statistically analyzed, the process health response value of each sea state process position is determined, and the health response trajectory is formed according to the chronological relationship of the sea state process positions.

[0042] Specifically, this implementation method does not directly align multiple historical monitoring processes according to the actual time elapsed during the sea state transition process, but aligns them according to the aforementioned sea state process.

[0043] For example, a transition from sea state A to sea state B might take 20 minutes, while another transition of the same type might take 45 minutes. If the target sensor response is directly compared based on the elapsed time, the change in the seawater exchange rate itself could easily be mistaken for a change in the sensor response rate.

[0044] Therefore, multiple sea state processes 0 to 1 are set at fixed intervals. In this embodiment, five sea state process positions are used: 0, 0.25, 0.50, 0.75, and 1.00. Among them, sea state processes 0 and 1.00 correspond to the first historical healthy sea state and the second historical healthy sea state, respectively, and sea state processes 0.25, 0.50, and 0.75 are used to characterize the degree of sea state change at different stages between the two historical healthy sea states.

[0045] For each historical sea state transition, linear interpolation is performed based on the target sensor monitoring time and the corresponding sea state process to obtain the health response of the target sensor at the above 5 sea state process positions.

[0046] After obtaining multiple health monitoring processes of the same first historical healthy sea state transitioning to the second historical healthy sea state, the health responses of multiple target sensors corresponding to the same sea state process position are statistically analyzed, and the median value is used as the process health response value of the sea state process position.

[0047] The health response values ​​of each process are arranged in order from 0 to 1 according to the position of the sea state process, and the health response trajectory corresponding to the target sensor is obtained.

[0048] The process health response value is a single-point health response benchmark of the health response trajectory at the corresponding sea state process position. The health response trajectory is composed of multiple process health response values ​​corresponding to different sea state process positions in chronological order of the sea state process positions.

[0049] For example, in a set of exemplary historical health data used to illustrate the processing procedure, the state parameters of the first historical healthy sea state A are: temperature approximately 18.2°C, salinity approximately 31.4 PSU, pressure approximately 0.02 MPa gauge pressure, and dissolved oxygen approximately 6.45 mg / L; the state parameters of the second historical healthy sea state B are: temperature approximately 15.6°C, salinity approximately 33.2 PSU, pressure approximately 0.02 MPa gauge pressure, and dissolved oxygen approximately 7.10 mg / L.

[0050] Since the pressure change between the two historical healthy sea states is very small, the current sea state process is mainly obtained by fusing the effective individual process values ​​corresponding to temperature, salinity and dissolved oxygen.

[0051] Based on statistics from multiple healthy sea state transitions, the process health response values ​​of the target pH sensor at sea state transition positions 0, 0.25, 0.50, 0.75, and 1.00 are approximately 8.12, 8.06, 7.99, 7.92, and 7.86, respectively.

[0052] The numerical values ​​described herein are for illustrative purposes only and do not imply a fixed relationship between temperature, salinity, dissolved oxygen, and pH in a specific sea area. For the actual monitored sea area, a corresponding health response trajectory should be established based on historical monitoring data obtained during the target sensor's confirmed health period.

[0053] S2. Acquire current monitoring data, determine the current sea state based on auxiliary sensors and match historical healthy sea states, compare the current response of the target sensor with the corresponding healthy response range to obtain the sea state deviation; when the current sea state changes, construct the current sea state process, compare it with the healthy response trajectory according to the same sea state process position to obtain the response span deviation and process response deviation. Specifically, the sea state characteristics are calculated for the current monitoring data using the same monitoring window length as the historical health data. The current monitoring window is then matched with each historical healthy sea state using the same normalization and sea state characteristic distance calculation method as the historical healthy sea state classification process. When multiple historical healthy sea states meet the sea state characteristic distance requirements, the historical healthy sea state with the smallest sea state characteristic distance is selected as the current matching sea state. When no historical healthy sea state meets the requirements, the current monitoring window is not used for fault calibration.

[0054] When the current sea state transitions from the first historical healthy sea state to the second historical healthy sea state, the state values ​​of each auxiliary sensor under both the first and second historical healthy sea states are used as the two endpoints of the current sea state process. The current individual process value of each auxiliary sensor is determined using the same calculation method as the historical individual process values. Invalid current individual process values ​​are excluded using the same change direction filtering rule. The remaining valid current individual process values ​​are then fused using the median to obtain the current sea state process value at the corresponding sampling time. The current sea state process values ​​are arranged in chronological order of sampling time to form the current sea state process. Therefore, both the historical sea state process and the current sea state process share the same sea state process benchmark, with the first historical healthy sea state corresponding to process 0 and the second historical healthy sea state corresponding to process 1.

[0055] When monitoring windows that match the first and second historical healthy sea states appear successively, and the smoothed current sea state progress value between the two monitoring windows increases from no more than 0.10 to no less than 0.90, and there is no invalid transition where the aforementioned sea state progress continuously regresses by more than 0.05, the monitoring process between the two monitoring windows is defined as a current sea state transition process from the first historical healthy sea state to the second historical healthy sea state. The preceding monitoring window is designated as the transition start monitoring window, and the following monitoring window is designated as the transition end monitoring window.

[0056] In step S2, the sea state deviation, response span deviation, and process response deviation are obtained, including: When the current response of the target sensor exceeds the health response range corresponding to the matched historical healthy sea state, the sea state deviation is determined based on the directed difference between the current response and the exceeded boundary value. When the current response is within the health response range, the sea state deviation is determined to be zero. The response span deviation is determined based on the difference between the change in the target sensor response before and after the current sea state change and the change in the response at both ends of the health response trajectory. The current response trajectory is formed by combining the current responses of the target sensors at each sea state process position during the current sea state transition. The current response trajectory is then translated and scaled to align its two ends with the two ends of the healthy response trajectory. The response differences between the aligned current response trajectory and the healthy response trajectory at the same sea state process position are compared to obtain the process response deviation corresponding to each sea state process position.

[0057] Specifically, for sea state deviation, when the current response of the target sensor is higher than the upper boundary of the corresponding healthy response range, the sea state deviation is obtained by subtracting the upper boundary from the current response; when the current response of the target sensor is lower than the lower boundary of the healthy response range, the sea state deviation is obtained by subtracting the lower boundary from the current response; when the current response of the target sensor is within the healthy response range, the sea state deviation is determined to be 0.

[0058] This preserves the direction of the sea state deviation, where a positive value indicates that the target sensor's current response is above the healthy response range, and a negative value indicates that the current response is below the healthy response range, providing a basis for subsequent identification of continuous zero-point drift in the same direction.

[0059] For the current sea state transition process, the median value of the target sensor response within the monitoring window at the start of the transition is taken as the current response at sea state process 0, and the median value of the target sensor response within the monitoring window at the end of the transition is taken as the current response at sea state process 1.00. For sea state process positions 0.25, 0.50, and 0.75, linear interpolation is performed based on the sampling time of the target sensor and the corresponding current sea state process to obtain the current response at the corresponding sea state process position. The current response trajectory is composed of the current responses at the above five sea state process positions. The change amplitude of the response at both ends of the current transition and the change amplitude of the response at both ends of the healthy response trajectory are both taken as the absolute value of the difference between the corresponding response values ​​at both ends.

[0060] The magnitude of the change in the responses at both ends of the current conversion is denoted as... The magnitude of the response changes at both ends of the health response trajectory is recorded as The response span deviation is: In the formula, In response to span deviation; This represents the magnitude of the change in the target sensor's response before and after the current sea state change; The magnitude of response changes at both ends of the health response trajectory.

[0061] The minimum effective response span is determined based on the short-term fluctuations of the target sensor under historical healthy sea conditions. Specifically, the standard deviations of the short-term fluctuations of the target sensor within the monitoring window corresponding to each historical healthy sea condition are statistically analyzed, the median value of the obtained short-term fluctuation standard deviations is taken, and three times the median value is taken as the minimum effective response span.

[0062] when When the response span is lower than the minimum effective response span determined based on the historical health noise of the target sensor, the corresponding sea state conversion template is not used to calculate the response span deviation.

[0063] A negative response span deviation indicates that the current target sensor's response amplitude to the same sea state change is smaller than that of the historical healthy state, while a positive response amplitude indicates that the current response amplitude is larger than that of the historical healthy state.

[0064] Furthermore, to avoid the overall zero-point offset and sensitivity changes directly affecting the dynamic response judgment, endpoint alignment is performed on the current response trajectory.

[0065] Let the position of the sea state process be P, then the aligned current response is: In the formula, The target sensor's current response after endpoint alignment at position P in the sea state process; The current response of the target sensor at position P in the sea state process; and These are the endpoint responses of the target sensors before and after the current sea state change; and These are the response values ​​at both ends of the health response trajectory.

[0066] The above processing ensures that the start and end points of the current response trajectory are consistent with those of the healthy response trajectory. Specifically, the start point position adjustment eliminates the impact of overall zero-point offset, and the response span ratio adjustment eliminates the impact of changes in overall response amplitude.

[0067] After completing endpoint alignment, calculate the process response deviation based on the progress position under the same sea state: In the formula, The process response deviation corresponding to the sea state process position P; The current response after endpoint alignment; The health response value is the process health response value corresponding to the process position P in the same sea state for the health response trajectory.

[0068] Therefore, the sea state deviation reflects whether the overall position of the current response has shifted; the response span deviation reflects whether the response amplitude of the target sensor to the same sea state change has changed; and the process response deviation, after eliminating the influence of the overall position shift and the response span, reflects whether the dynamic response process of the target sensor is still consistent with the historical health state.

[0069] When the change in the response of the target sensors at both ends of the current conversion is lower than the minimum effective response span determined based on historical health data, the endpoint scale adjustment and process response deviation calculation are not performed to avoid the denominator being too small and amplifying the calculation error; in this case, only the response span deviation is used for subsequent sensitivity drift judgment.

[0070] S3. Based on the sea state deviation, and in conjunction with the response span deviation and process response deviation at the time of the current sea state transition, determine the fault type, which includes zero drift, sensitivity drift and dynamic response anomaly. Step S3 includes: During the current sea state transition, when the process response deviation exceeds the dynamic response allowable range at multiple consecutive sea state process positions, it is determined to be a dynamic response anomaly; During the current sea state transition, if the dynamic response anomaly judgment conditions are not met, when the response span deviation exceeds the allowable span range, it is determined to be sensitivity drift. If the target sensor matches the same historical healthy sea state in multiple consecutive monitoring windows and produces a sea state deviation in the same direction, it is determined to be zero drift when the current sea state has not changed. When the current sea state changes, if the target sensor produces a sea state deviation in the same direction under the historical healthy sea state before and after the current sea state change, and neither the dynamic response anomaly judgment condition nor the sensitivity drift judgment condition is met, it is determined to be zero drift.

[0071] Specifically, the allowable range of dynamic response is not set arbitrarily, but is obtained based on the historical health sea state conversion data of the target sensor itself.

[0072] For multiple historical sea state health transitions, a leave-one-out verification approach can be used to establish a dynamic response allowable range. That is, each historical sea state health transition is selected as the process to be evaluated, and temporary health response trajectories are established using the remaining historical sea state health transitions. Then, the process to be evaluated is processed according to the endpoint alignment method in S2 to obtain the possible process response deviations at each sea state process position when the target sensor is in a healthy state.

[0073] After summarizing all historical healthy sea state transition processes, the absolute value of the process response deviation is statistically analyzed for each sea state process position. In this implementation method, the 95th percentile value is taken as the dynamic response allowable boundary for the corresponding sea state process position.

[0074] Since the current response trajectory coincides with the two ends of the healthy response trajectory after the endpoint alignment, the process response deviation at the positions of sea state process 0 and 1.00 is used to complete the endpoint constraint and does not participate in the dynamic response anomaly judgment. In this embodiment, the process response deviation at the three intermediate positions of sea state process 0.25, 0.50 and 0.75 is used to judge the dynamic response state of the target sensor.

[0075] For example, in the exemplary data, the allowable boundaries for dynamic response at sea state processes of 0.25, 0.50, and 0.75 are approximately 0.025 pH, 0.030 pH, and 0.028 pH, respectively.

[0076] When the absolute value of the process response deviation of the target sensor at the three intermediate positions of sea state progress of 0.25, 0.50 and 0.75 is greater than the corresponding dynamic response allowable boundary, a dynamic response anomaly is determined to have occurred.

[0077] Using multiple consecutive process positions for judgment, rather than judging based on a single position, can reduce the probability of misjudgment caused by wave impact, instantaneous communication interference, and single-point sampling anomalies.

[0078] The allowable span range is also obtained based on historical sea state transition processes. For each historical sea state transition process, the deviation of its response span from the response spans at both ends of the healthy response trajectory is calculated, and the absolute value of the span deviation for multiple historical sea state transition processes is statistically analyzed.

[0079] The 95th percentile of the absolute value of the span deviation of multiple historical healthy sea state transition processes is taken as the span allowable boundary, and the positive and negative values ​​of the span allowable boundary constitute the span allowable range.

[0080] In one specific embodiment, the allowable range of the span obtained from historical health data statistics is ±0.05. When the absolute value of the current response span deviation is greater than 0.05, and the process response deviation does not meet the dynamic response anomaly condition, it is determined to be sensitivity drift.

[0081] For zero-point drift when the current sea state has not changed, this implementation requires that the target sensor matches the same historical healthy sea state for three consecutive 30-minute monitoring windows, and that the sea state deviation directions corresponding to the three monitoring windows are the same, in order to determine it as zero-point drift.

[0082] For zero-point drift during sea state transitions, the start and end monitoring windows of the transition are determined separately. Zero-point drift is defined as follows: when the current response of the target sensor within both the start and end monitoring windows is below the lower boundary of its corresponding historical healthy sea state response range, or above the upper boundary of its corresponding historical healthy sea state response range, and the response span deviation remains within the allowable range and the process response deviation does not meet the dynamic response anomaly conditions.

[0083] Based on the above judgment order, dynamic response anomalies are judged first, followed by sensitivity drift, and finally zero-point drift.

[0084] Among them, abnormal dynamic response usually corresponds to the target sensor’s slow response or change in response pattern; sensitivity drift corresponds to the target sensor still being able to follow changes in sea state, but the response amplitude caused by the same sea state change changes; zero drift corresponds to the overall output position of the target sensor shifting in the same direction, while the response amplitude and dynamic change process are still basically normal.

[0085] When multiple degradation factors exist simultaneously in the target sensor, the fault that is prioritized in the current judgment sequence should be addressed first. After calibration or maintenance is completed and subsequent re-inspection is passed, S2 and S3 should be executed again to determine whether other types of faults still exist.

[0086] S4. When the fault type is either zero-point drift or sensitivity drift, determine the calibration control quantity based on the corresponding health response and the current response, and update the calibration parameters of the target sensor; when the fault type is abnormal dynamic response, maintain the current calibration parameters and output maintenance information. In step S4, determining the calibration control quantity based on the corresponding health response and the current response includes: When zero-point drift is determined, a health response reference value is determined based on the upper and lower boundaries of the health response range corresponding to the matched historical healthy sea state. The zero-point calibration control value is determined based on the offset between the health response reference value and the current response, and the zero-point calibration parameters are updated. When sensitivity drift is identified, the sensitivity calibration control quantity is determined based on the proportional relationship between the target sensor response span during the current sea state transition and the response change amplitude at both ends of the healthy response trajectory, and the sensitivity calibration parameters are updated.

[0087] Specifically, when it is determined to be a zero-point drift, the midpoint between the upper and lower boundaries of the corresponding historical sea state health response range is taken as the health response benchmark value.

[0088] The zero-point calibration control value is obtained by subtracting the current response of the target sensor from the health response baseline value, and the existing zero-point parameters of the target sensor are corrected based on the zero-point calibration control value.

[0089] If the zero-point drift is determined during a complete sea state transition, the zero-point calibration control quantity is calculated based on the start and end monitoring windows of the transition, respectively.

[0090] When the two zero-point calibration control quantities are in the same direction and the difference between them is not greater than the consistency allowable value set according to historical health data, the average value of the two zero-point calibration control quantities is used to perform zero-point calibration.

[0091] For example, in this embodiment, the allowable value for consistency of the zero-point calibration control value can be set to 0.02 pH. When the difference between the zero-point calibration control values ​​obtained from the two monitoring windows exceeds 0.02 pH, automatic calibration is not performed directly, but the system waits for the next set of matching data for reconfirmation.

[0092] When sensitivity drift is identified, the sensitivity calibration control value is determined based on the response change amplitudes at both ends of the healthy response trajectory and the response change amplitudes at both ends of the current sea state transition: in, This is the sensitivity calibration control value; The magnitude of response changes at both ends of the health response trajectory; This represents the magnitude of the change in the target sensor's response before and after the change in sea state.

[0093] For target sensors with zero-point parameter and sensitivity parameter configuration interfaces, the corresponding calibration control quantities can be converted into zero-point parameter correction quantities or sensitivity parameter correction quantities in the device's existing calibration model, and then written through the sensor parameter interface.

[0094] For target sensors that do not support direct modification of internal calibration parameters, the zero-point calibration control value and sensitivity calibration control value can be stored in the data processing controller bound to the target sensor, and the corresponding correction parameters can be called during the original response conversion and compensation process of the target sensor.

[0095] This implementation does not limit the pH sensor to using a specific form of linear output expression, thus it is applicable to target sensors using electrode slope, ISFET response parameters or other existing sensor calibration models.

[0096] Taking a monitoring device that supports RS485 communication and parameter configuration as an example, the corresponding calibration parameters can be updated through the parameter register reserved by the device; for target sensors that do not support parameter distribution, the buoy-end data processing controller saves the calibration parameters and corrects them accordingly for subsequent monitoring results.

[0097] Therefore, the in-situ calibration described in this method does not require removing the target sensor from the marine environment after each drift is detected. Instead, the calibration control quantity is determined based on the historical health response of the sensor itself, which has been confirmed during long-term operation, and the current real sea conditions reflected by other auxiliary sensors.

[0098] In step S4, when the fault type is dynamic response anomaly, the current calibration parameters of the target sensor are maintained, the abnormal state of the current output data of the target sensor is marked, and maintenance information characterizing the dynamic response anomaly of the target sensor is generated.

[0099] The reason why zero-point or sensitivity calibration is not performed directly for abnormal dynamic response is that when the target sensor has a significant slow response due to factors such as probe surface adhesion, sensitive film contamination, or device aging, even if the endpoints before and after the conversion are corrected to normal values ​​through static parameters, the intermediate change process still cannot be restored to normal.

[0100] Therefore, in the event of an abnormal dynamic response, the current calibration parameters are maintained, and the target sensor identifier, the time of the abnormality, and the dynamic response abnormality marker are recorded in the maintenance information.

[0101] Furthermore, the system can record the number of process locations that continuously exceed the allowable range of dynamic response. If the target sensor experiences abnormal dynamic response during two consecutive sea state transitions, the maintenance priority can be increased, and a probe cleaning or on-site inspection prompt can be output.

[0102] S5. After the calibration parameters are updated, at least one of the following should be selected from the historical healthy sea state and sea state conversion process for retesting: if the retest passes, the updated calibration parameters should be retained; if the retest fails, the parameters before calibration should be restored.

[0103] In step S5, at least one type of historical healthy sea state and sea state transition process is selected for re-examination, including: Select subsequent matching data that were not involved in the determination of the calibration control quantities in this study; When the calibration parameter updated this time is the zero-point calibration parameter, the sea state deviation is re-determined using the historical healthy sea state of subsequent matching, and the re-determined sea state deviation is used as the zero-point re-examination deviation. When the calibration parameters updated this time are sensitivity calibration parameters, the response span deviation and process response deviation are re-determined using the subsequent matched sea state conversion process, and the re-determined response span deviation and process response deviation are combined into a sensitivity re-examination deviation group. When the updated calibration parameter is the zero-point calibration parameter and the zero-point re-inspection deviation is within the corresponding allowable range, the re-inspection is deemed successful and the updated calibration parameter is maintained. When the updated calibration parameter is the sensitivity calibration parameter and all deviations in the sensitivity retest deviation group are within the corresponding allowable range, the retest is deemed passed and the updated calibration parameter is maintained.

[0104] Except for the above-mentioned cases where the re-inspection passed, if the re-inspection fails, the parameters before calibration should be restored.

[0105] Specifically, the data used to determine the calibration control values ​​is not intended to prove the correctness of the calibration results.

[0106] For zero-point calibration, after the zero-point parameters are updated, subsequent monitoring data is acquired, and the target sensor is waited to re-enter the historical sea state corresponding to this zero-point calibration.

[0107] When subsequent independent monitoring windows re-match the historical sea state health, the sea state deviation is recalculated based on the updated target sensor response. When the target sensor response re-enters the corresponding healthy response range, the zero-point re-inspection deviation is 0, confirming that the first re-inspection has passed.

[0108] For sensitivity calibration, the process waits for the subsequent conversion from the first historical sea state to the second historical sea state corresponding to this calibration to occur again, and the subsequent conversion process is not involved in the calculation of the current sensitivity calibration control value.

[0109] The response span deviation and process response deviation are recalculated for the independent sea state transition process. When the response span deviation is within the allowable span range determined based on historical health data, and the absolute values ​​of the process response deviations at the sea state transition positions of 0.25, 0.50, and 0.75 are not greater than the corresponding dynamic response allowable boundaries, the sensitivity calibration is considered to have passed the first retest.

[0110] If the retest fails, cancel the updated calibration parameters, restore the parameters saved before calibration, and record the automatic calibration failure status to prevent erroneous parameters from being used in subsequent outputs.

[0111] After the re-inspection is passed, it also includes: Set the target sensor data and the synchronously collected auxiliary sensor data when the re-inspection is passed as the health data to be confirmed; Select another set of subsequent matching data that did not participate in the previous re-examination and re-examine it. If the re-examination passes, add the health data to be confirmed to the historical health data to update at least one of the health response range and health response trajectory that match the health data to be confirmed.

[0112] Specifically, the data that passes the first re-inspection is not immediately entered into the historical health database, but is set as health data to be confirmed. At the same time, the auxiliary sensor data collected synchronously with it, the corresponding historical health sea state label, and the current calibration parameters are saved.

[0113] After continuing the operation, wait for another set of independent subsequent matching data that were neither involved in the calibration control calculation nor in the first retest.

[0114] For monitoring window data that matches historical healthy sea states, re-evaluate whether the calibrated target sensor response is within the corresponding healthy response range; for sea state transition data, re-evaluate whether the response span deviation and process response deviation are within the corresponding allowable range.

[0115] Only after the second re-examination is passed will the pending health data generated from the first re-examination be added to the historical health data.

[0116] When the health data to be confirmed corresponds to historical sea state health, it can be used to recalculate the corresponding health response range; when the health data to be confirmed contains a complete and confirmed normal sea state transition process, it can be further used to update the process health response value and health response trajectory of the corresponding sea state process location.

[0117] This enables historical health data to be updated gradually, while avoiding the situation where monitoring data after an erroneous automatic calibration immediately becomes the new health benchmark, causing the health response range or health response trajectory to continuously deviate with the erroneous result.

[0118] Example: The pH sensor on a nearshore marine monitoring buoy is used as the target sensor, and temperature, salinity, pressure and dissolved oxygen sensors are used as auxiliary sensors for explanation.

[0119] Based on historical health data from the target sensor and auxiliary sensors, two historical healthy sea states, A and B, were established. Sea state A corresponds to a temperature of approximately 18.2℃, salinity of approximately 31.4 PSU, and dissolved oxygen of approximately 6.45 mg / L, while sea state B corresponds to a temperature of approximately 15.6℃, salinity of approximately 33.2 PSU, and dissolved oxygen of approximately 7.10 mg / L. A health response trajectory was established based on multiple historical health monitoring processes transitioning from sea state A to sea state B. The health response values ​​of the target pH sensor at sea state transitions of 0, 0.25, 0.50, 0.75, and 1.00 were approximately 8.12, 8.06, 7.99, 7.92, and 7.86, respectively; the corresponding health response ranges for sea state A and sea state B were 8.09–8.15 and 7.83–7.89, respectively.

[0120] After long-term operation, the auxiliary sensor once again detected a sea state change process that matched the historical transition from sea state A to sea state B. The current responses of the target pH sensor at the aforementioned sea state transition points are approximately 7.95, 7.89, 7.82, 7.75, and 7.69, respectively.

[0121] The pH response change before and after the current sea state transition is 0.26 pH, consistent with the response change at both ends of the healthy response trajectory. After translating and scaling the current response trajectory, the process response deviations at sea state progression positions of 0.25, 0.50, and 0.75 are all 0, within the corresponding allowable range for dynamic response. Therefore, the conditions for sensitivity drift and dynamic response anomaly determination are not met.

[0122] Meanwhile, under sea state A, the current response of the target pH sensor is 7.95, which is lower than the lower boundary of the healthy response range of 8.09. Under sea state B, the current response is 7.69, which is lower than the lower boundary of the healthy response range of 7.83. The sea state deviations corresponding to the two sea states are both negative, thus determining that the target pH sensor has experienced zero-point drift.

[0123] The median values ​​of the health response ranges corresponding to sea states A and B are taken as the health response baseline values, respectively. The zero-point calibration control values ​​corresponding to the two sea states are approximately +0.17 pH. Therefore, the zero-point calibration parameters of the target sensor are updated according to the zero-point calibration control values.

[0124] After the parameter update is completed, the calibration results are not directly verified using the current sea state conversion data. Instead, subsequent independent monitoring data are acquired. When the data is matched to sea state A again, the target pH sensor response is 8.11, which is within the healthy response range of 8.09 to 8.15, confirming that the first retest has passed. Subsequently, another set of independent data is matched to sea state B, and the target pH sensor response is 7.85, which is within the healthy response range of 7.83 to 7.89, confirming that the retest has passed again. The pending health data generated from the first retest is then added to the historical health data.

[0125] Therefore, this method can use auxiliary sensors to determine the degree of change in actual sea state and use the target sensor's own historical health response to distinguish between environmental changes and sensor drift, and complete zero-point drift identification, in-situ calibration and subsequent independent re-inspection without disassembling the target sensor.

[0126] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for self-diagnosing and calibrating faults in marine sensors, characterized in that, Includes the following steps: S1. Acquire historical health data of the target sensor and auxiliary sensor, divide historical health sea conditions according to the auxiliary sensor data, construct historical sea condition process based on the changes between different historical health sea conditions, and establish the health response range of the target sensor and the health response trajectory that changes with the historical sea condition process. S2. Acquire current monitoring data, determine the current sea state based on auxiliary sensors and match historical healthy sea states, compare the current response of the target sensor with the corresponding healthy response range to obtain the sea state deviation; when the current sea state changes, construct the current sea state process, compare it with the healthy response trajectory according to the same sea state process position to obtain the response span deviation and process response deviation. S3. Based on the sea state deviation, and in conjunction with the response span deviation and process response deviation at the time of the current sea state transition, determine the fault type, which includes zero drift, sensitivity drift and dynamic response anomaly. S4. When the fault type is either zero-point drift or sensitivity drift, determine the calibration control quantity based on the corresponding health response and the current response, and update the calibration parameters of the target sensor; when the fault type is abnormal dynamic response, maintain the current calibration parameters and output maintenance information. S5. After the calibration parameters are updated, at least one of the following should be selected from the historical healthy sea state and sea state conversion process for retesting: if the retest passes, the updated calibration parameters should be retained; if the retest fails, the parameters before calibration should be restored.

2. The method for self-diagnosing and calibrating marine sensor faults according to claim 1, characterized in that, In step S1, historical sea condition health is classified based on auxiliary sensor data, including: Extract continuous monitoring windows from historical data where both the target sensor and auxiliary sensors are in a healthy state; The average values ​​and variation ranges of each auxiliary sensor within the continuous monitoring window are used to form sea state characteristics, and the continuous monitoring windows are classified according to the similarity between the sea state characteristics. The marine environmental state corresponding to the same type of continuous monitoring window is determined as the same historical healthy sea state, and the corresponding health response range is determined based on the target sensor data under the historical healthy sea state.

3. The method for self-diagnosing and calibrating marine sensor faults according to claim 1, characterized in that, In step S1, the historical sea state process is constructed based on the changes between different historical healthy sea states, including: Determine the first historical sea state and the second historical sea state, and obtain the status values ​​of each auxiliary sensor under the first historical sea state and the second historical sea state; For the historical monitoring process of transitioning from the first historical sea state to the second historical sea state, the historical individual process value is determined according to the degree of change of the monitoring value of each auxiliary sensor relative to the two state values ​​during the transition process. Historical single-item process values ​​obtained when the direction of change of the corresponding auxiliary sensor during the historical monitoring process is inconsistent with the direction of change determined by its two state values ​​are excluded. The remaining historical single-item process values ​​are then fused to obtain the historical sea state process that characterizes the degree of sea state change during the historical monitoring process.

4. The method for self-diagnosing and calibrating marine sensor faults according to claim 1, characterized in that, In step S1, the establishment of the health response trajectory that changes with the historical sea state process includes: Determine the first and second historical healthy sea states from historical healthy sea states, obtain the historical health monitoring process of multiple transitions from the first historical healthy sea state to the second historical healthy sea state, and align the process positions according to the corresponding historical sea state process. The target sensor health responses corresponding to each sea state process position are statistically analyzed, the process health response value of each sea state process position is determined, and the health response trajectory is formed according to the chronological relationship of the sea state process positions.

5. The method for self-diagnosing and calibrating marine sensor faults according to claim 1, characterized in that, In step S2, the sea state deviation, response span deviation, and process response deviation are obtained, including: When the current response of the target sensor exceeds the health response range corresponding to the matched historical healthy sea state, the sea state deviation is determined based on the directed difference between the current response and the exceeded boundary value. When the current response is within the health response range, the sea state deviation is determined to be zero. The response span deviation is determined based on the difference between the change in the target sensor response before and after the current sea state change and the change in the response at both ends of the health response trajectory. The current response trajectory is formed by combining the current responses of the target sensors at each sea state process position during the current sea state transition. The current response trajectory is then translated and scaled to align its two ends with the two ends of the healthy response trajectory. The response differences between the aligned current response trajectory and the healthy response trajectory at the same sea state process position are compared to obtain the process response deviation corresponding to each sea state process position.

6. The method for self-diagnosing and calibrating marine sensor faults according to claim 5, characterized in that, Step S3 includes: During the current sea state transition, when the process response deviation exceeds the dynamic response allowable range at multiple consecutive sea state process positions, it is determined to be a dynamic response anomaly; During the current sea state transition, if the dynamic response anomaly judgment conditions are not met, when the response span deviation exceeds the allowable span range, it is determined to be sensitivity drift. If the target sensor matches the same historical healthy sea state in multiple consecutive monitoring windows and produces a sea state deviation in the same direction, it is determined to be zero drift when the current sea state has not changed. When the current sea state changes, if the target sensor produces a sea state deviation in the same direction under the historical healthy sea state before and after the current sea state change, and neither the dynamic response anomaly judgment condition nor the sensitivity drift judgment condition is met, it is determined to be zero drift.

7. The method for self-diagnosing and calibrating marine sensor faults according to claim 6, characterized in that, In step S4, determining the calibration control quantity based on the corresponding health response and the current response includes: When zero-point drift is determined, a health response reference value is determined based on the upper and lower boundaries of the health response range corresponding to the matched historical healthy sea state. The zero-point calibration control value is determined based on the offset between the health response reference value and the current response, and the zero-point calibration parameters are updated. When sensitivity drift is identified, the sensitivity calibration control quantity is determined based on the proportional relationship between the target sensor response span during the current sea state transition and the response change amplitude at both ends of the healthy response trajectory, and the sensitivity calibration parameters are updated.

8. The method for self-diagnosing and calibrating marine sensor faults according to claim 6, characterized in that, In step S4, when the fault type is dynamic response anomaly, the current calibration parameters of the target sensor are maintained, the abnormal state of the current output data of the target sensor is marked, and maintenance information characterizing the dynamic response anomaly of the target sensor is generated.

9. A method for self-diagnosing and calibrating faults in a marine sensor according to claim 7, characterized in that, In step S5, at least one type of historical healthy sea state and sea state transition process is selected for re-examination, including: Select subsequent matching data that were not involved in the determination of the calibration control quantities in this study; When the calibration parameters updated this time are zero-point calibration parameters, the sea state deviation is re-determined using the historical healthy sea state of subsequent matching, and the re-determined sea state deviation is used as the zero-point re-examination deviation. When the calibration parameters updated this time are sensitivity calibration parameters, the response span deviation and process response deviation are re-determined using the subsequent matched sea state conversion process, and the re-determined response span deviation and process response deviation are combined into a sensitivity re-examination deviation group. When the updated calibration parameter is the zero-point calibration parameter and the zero-point re-inspection deviation is within the corresponding allowable range, the re-inspection is deemed successful and the updated calibration parameter is maintained. When the updated calibration parameter is the sensitivity calibration parameter and all deviations in the sensitivity retest deviation group are within the corresponding allowable range, the retest is deemed passed and the updated calibration parameter is maintained. Except for cases where the re-inspection passes as described above, if the re-inspection fails, the parameters before calibration should be restored.

10. A method for self-diagnosing and calibrating faults in a marine sensor according to claim 9, characterized in that, After the re-inspection is passed, it also includes: Set the target sensor data and the synchronously collected auxiliary sensor data when the re-inspection is passed as the health data to be confirmed; Select another set of subsequent matching data that did not participate in the previous re-examination and re-examine it. If the re-examination passes, add the health data to be confirmed to the historical health data to update at least one of the health response range and health response trajectory that match the health data to be confirmed.

Images