Active fault-tolerant reliability evaluation method for liquid level monitoring sensor of liquefied gas carrier

Abstract

The invention relates to an active fault-tolerant reliability evaluation method for a liquid level monitoring sensor of a liquefied gas carrier. Comprising the following steps: acquiring required sensor monitoring data and evaluation parameters; establishing a semi-Markov process reliability model of the sensor active fault-tolerant system based on the sensor monitoring data and the evaluation parameters, defining a state space of the sensor active fault-tolerant system, including a normal operation state, a fault state and a fault diagnosis state, and determining a state transition probability according to the evaluation parameters; calculating a reliability function R (t) of the active fault-tolerant system of the sensor based on the semi-Markov process reliability model; according to the reliability function R (t), calculating an MTBF (mean time of failure) of the active fault-tolerant system of the sensor; and outputting a reliability evaluation result of the sensor active fault-tolerant system, wherein the reliability evaluation result comprises the reliability function R (t) and the mean time to failure (MTBF). According to the invention, synchronous improvement of safety redundancy, maintenance cost and operation economy of the ship cargo liquid level monitoring system is realized.

Description

Technical Field

[0001] This invention belongs to the field of shipbuilding and marine engineering technology, specifically relating to an active fault-tolerant reliability assessment method for a liquefied gas carrier level monitoring sensor. Background Technology

[0002] Ship information perception refers to the process by which ships use various sensors and information processing devices to acquire information about their own status and the surrounding environment in order to improve the safety and reliability of navigation. The ship's own status information (such as the status of engine room equipment, speed, and heading) is mainly acquired through sensors such as pressure, temperature, speed, and liquid level; while the surrounding environment information (such as nearby ships, weather, and sea conditions) is perceived by equipment such as AIS, radar, and surveillance cameras.

[0003] Liquid level sensors are one of the core means of ship condition monitoring, playing a crucial role in the safe operation of liquefied petroleum gas (LPG) carriers. Liquid level, temperature, and pressure measurement systems are widely used on ships for cargo level monitoring, high-level alarms, ballast tank level monitoring, draft measurement, and engine room and bilge water level alarms. These systems are essential for ensuring ship safety and stability, helping crew members to monitor liquid conditions in real time and address potential hazards promptly. Taking an LPG carrier as an example, its cargo tank level monitoring system consists of sensors, signal acquisition and transmission, monitoring and alarm, and signal output control units, and is integrated with subsystems such as cargo tank pumps, compressors, cargo pipelines, inert gas systems, fire alarms, ESD emergency shutdown, and ship-to-shore connections. The entire liquid level monitoring system involves numerous devices and a complex control chain, placing extremely high demands on system reliability and safety.

[0004] Currently, commonly used liquid level measurement methods include mechanical buoys / gauges, ultrasonic, capacitive, differential pressure, optical, and microwave radar. Mechanical buoys and immersion probes are simple in structure and have a long history of application, but their measurement accuracy is low and reliability is poor. Ultrasonic ranging utilizes the time-of-flight of sound waves for non-contact measurement; however, changes in liquid temperature gradients and gas phase conditions significantly affect measurement accuracy. Capacitive level sensors calculate liquid level by monitoring changes in capacitance within the liquid, but are susceptible to interference from fluctuations in ambient temperature and liquid dielectric properties. Differential pressure level gauges calculate liquid level based on the hydrostatic pressure of the liquid column; fluctuations in liquid density and temperature introduce errors, affecting measurement reliability. Optical methods, such as camera measurement, require transparent containers, and fiber optic sensors have limited stability in harsh environments; therefore, the applicability of these methods in ship liquid level monitoring is limited.

[0005] In contrast, microwave radar level gauges are unaffected by changes in medium density, dielectric constant, and conductivity, and are also unaffected by smoke, foam, or vapor on the liquid surface. They enable high-precision remote level measurement and are adaptable to a wide range of temperature and pressure conditions, thus becoming an important means of measuring liquid levels in liquefied cargo tanks. However, even radar measurement faces unique challenges in extreme conditions such as those of liquefied gas carriers. Taking LPG cargo tanks as an example, radar level gauges place extremely high demands on the design of the waveguide (measuring cavity); the diameter, length, and arrangement of vents in the waveguide all affect the propagation and echo quality of the microwave signal, while research on LPG conditions is relatively limited. Furthermore, the frequent occurrence of gas-liquid two-phase flow and violent sloshing on the LPG surface further increases the difficulty of measuring device design and signal processing. In addition, the harsh environment of LPG cargo tanks and stringent safety monitoring requirements necessitate long-term reliable operation of sensors under low temperature, high pressure, and flammable conditions, placing extremely high demands on sensor reliability. Therefore, timely monitoring of sensor malfunctions and effective reconstruction of level information are essential for ensuring the safe operation of ships.

[0006] Although liquid level monitoring systems have a foundation in shipboard applications, they still fall short in terms of measurement accuracy and operational reliability, especially on LPG carriers where existing technology struggles to handle rapid fluctuations in liquid levels. On one hand, research on the failure mechanisms of liquid level monitoring systems is weak, often resulting in slow and inadequate early warning and response once sensors fail. On the other hand, single sensors are ill-suited to the complex and ever-changing conditions at sea. Therefore, anomaly detection of liquid level signals and multi-sensor information fusion have become key research areas. Current research has attempted to improve measurement reliability through virtual sensor redundancy and data fusion, but the deployment of these algorithms on actual ships and their stability and effectiveness under various environments still require further verification and refinement. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of the aforementioned background technology and provide an active fault-tolerant reliability assessment method for liquefied gas ship level monitoring sensors.

[0008] The technical solution adopted in this invention is: an active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors, including... Acquire the required sensor monitoring data and evaluation parameters, including physical sensor failure rate, virtual sensor failure rate, fault diagnosis false alarm rate, self-calibration rate, and fault diagnosis delay distribution parameters; A semi-Markov process reliability model of the sensor active fault-tolerant system is established based on the sensor monitoring data and evaluation parameters. The state space of the sensor active fault-tolerant system is defined, including normal operation state, fault state and fault diagnosis state. The state transition probability is determined according to the evaluation parameters. The reliability function R(t) of the sensor active fault-tolerant system is calculated based on the semi-Markov process reliability model. The mean time between failures (MTBF) of the sensor active fault-tolerant system is calculated based on the reliability function R(t). Output the reliability assessment results of the sensor active fault-tolerant system, including the reliability function R(t) and the mean time between failures (MTBF).

[0009] Preferably, the semi-Markov process reliability model includes a fault state model and a fault diagnosis state model. The fault state model represents the process by which a physical sensor transitions from a normal operating state to a fault state. The fault diagnosis state model represents the transition process of the fault diagnosis result between a correct detection state and a false alarm state, and considers the influence of the fault diagnosis delay distribution parameter and the fault diagnosis false alarm rate on the state transition.

[0010] Preferably, the reliability function R(t) and mean time between failures (MTBF) of the sensor active fault-tolerant system are calculated using the semi-Markov process reliability model, where R(t) represents the probability that the sensor active fault-tolerant system operates without failure in the time interval (0,t), and MTBF is obtained by integrating the reliability function R(t) over the time interval [0, ∞).

[0011] Preferably, the sensor active fault-tolerant system adopts a hybrid redundancy sensor configuration, including one physical liquid level sensor and two virtual liquid level sensors for monitoring the same cargo liquid level information, thereby forming a three-signal redundancy architecture with one main and two backups.

[0012] Preferably, the two virtual liquid level sensors are generated by different algorithms: one is based on multi-sensor collaborative regression prediction to estimate the cargo liquid level using monitoring data from relevant sensors; the other is based on time series prediction to predict the cargo liquid level using historical monitoring data.

[0013] Preferably, the fault diagnosis state model sets the fault diagnosis false alarm rate and self-correction rate as model parameters to characterize the probability of a diagnostic error alarm and the probability of recovering from a false alarm state to a correct diagnosis state.

[0014] A preferred approach includes assessing the reliability impact of changes in evaluation parameters, which include the false alarm rate of fault diagnosis and the virtual sensor failure rate. By reducing the false alarm rate of fault diagnosis or changing the virtual sensor failure rate and recalculating the reliability function and MTBF, the reliability indices under different parameter conditions are compared to determine the degree of influence of each parameter on the reliability of the sensor active fault-tolerant system.

[0015] Preferredly, the state transition matrix of the semi-Markov process reliability model is generated as follows: the transition probability of the fault state is determined based on the physical sensor failure rate and the virtual sensor failure rate, and the transition probability of the fault diagnosis state is determined based on the fault diagnosis false alarm rate and the self-correction rate, thereby forming a joint state transition probability matrix.

[0016] Preferably, the time interval between the occurrence of a fault and its diagnosis is modeled as a random variable of fault diagnosis delay, which is a fault diagnosis delay distribution parameter τ. The distribution function and probability density function of τ are used to describe the characteristics of this diagnosis delay, and this diagnosis delay model is incorporated into the calculation of the semi-Markov process reliability model.

[0017] Preferredly, reliability assessment is performed by combining the sensor fault state process and the fault diagnosis state process into a joint process. The semi-Markov process reliability model is used to simultaneously describe the sensor hardware fault state and the fault diagnosis state, thereby comprehensively calculating the reliability index of the sensor active fault-tolerant system.

[0018] This invention comprehensively considers the timing characteristics of sensor failures and the incompleteness of sensor fault diagnosis and isolation. By comparing and analyzing with the Double Redundancy Sensor Fault Tolerant (DRSFT) system, it explores the reliability of the HRSFT method and analyzes the impact of the number of redundant sensors on system reliability, thereby providing support for the design and optimization of sensing systems.

[0019] This invention incorporates hardware failures, false alarms, and diagnostic delays into a semi-Markov process model. This method accurately quantifies the entire-link evolution of the LNG ship cargo level monitoring system within a "one primary, two backup" hybrid redundancy architecture, significantly improving both the reliability function R(t) and MTBF. Introducing false alarm rate and self-correction rate reduces unplanned downtime and false alarms, while the three-signal heterogeneous redundancy further covers latent fault scenarios. Parameter sensitivity analysis provides quantitative decisions for operation and maintenance optimization, and software-level deployment requires no additional hardware modifications. Overall, this invention simultaneously improves the safety redundancy, maintenance costs, and operational economy of the LNG ship cargo level monitoring system. Attached Figure Description

[0020] Figure 1 Here is the state transition diagram for X(t); Figure 2 A diagram illustrating the analytical process of a semi-Markov kernel; Figure 3 Reliability curves for the structure of the active fault-tolerant method for hybrid redundancy sensors; Figure 4 From A probability diagram of state transitions to each state; Figure 5 Reliability curves for different fault diagnosis false alarm rates; Figure 6 Reliability curves for different fault diagnosis false alarm rates and self-correction rates; Figure 7 Reliability curves for different failure rate functions; Figure 8 A comparative analysis chart for 16 different working conditions; Figure 9 This is a flowchart of an active fault-tolerant reliability assessment method for a liquefied gas ship level monitoring sensor according to the present invention. Detailed Implementation

[0021] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0022] like Figure 9 As shown in this embodiment, an active fault-tolerant reliability assessment method for a liquefied gas carrier level monitoring sensor includes... Acquire the required sensor monitoring data and evaluation parameters, including physical sensor failure rate, virtual sensor failure rate, fault diagnosis false alarm rate, self-calibration rate, and fault diagnosis delay distribution parameters; A semi-Markov process reliability model of the sensor active fault-tolerant system is established based on the sensor monitoring data and evaluation parameters. The state space of the sensor active fault-tolerant system is defined, including normal operation state, fault state and fault diagnosis state. The state transition probability is determined according to the evaluation parameters. The reliability function R(t) of the sensor active fault-tolerant system is calculated based on the semi-Markov process reliability model. The mean time between failures (MTBF) of the sensor active fault-tolerant system is calculated based on the reliability function R(t). Output the reliability assessment results of the sensor active fault-tolerant system, including the reliability function R(t) and the mean time between failures (MTBF).

[0023] Specifically, it includes the following steps: Reliability assessment based on a semi-Markov process model 1) Reliability Indicators This embodiment uses a reliability function and Mean Time Between Failures (MTBF) to evaluate the sensor active fault-tolerant system. The reliability function R(t) represents the reliability index of the sensor active fault-tolerant system within the interval (0, t). If no failure occurs within a specified time T, the system meets the reliability requirements during that time period.

[0024] Assume the system performance at time t is Fault Model i The system performance boundary in is For each fault model, the system performance satisfies... Due to factors such as hardware and environment, the system may momentarily experience [a certain condition]. The state is not satisfied. The system will revert to its previous state after a period of time. The state. Therefore, in this embodiment, this instant is defined as... Assuming in time Internal system performance meets The requirements are as follows: Also known as a hard deadline. The system's reliability model is as follows: (1) (2) Where R(t) is the reliability function; P {} represents the probability of an event occurring; MTBF is the mean time between failures. t 1 represents the starting point of a certain time interval, indicating the moment when the system's performance begins to decline; t 2 represents the end time of the interval, indicating that the system performance continues to decline until that moment.

[0025] 2) State description of a semi-Markov process The semi-Markov process reliability model is a statistical model with the same structure as the hidden Markov model, but the difference lies in that the unobservable process it describes is a semi-Markov process rather than a Markov process. Specifically, the main difference between a semi-Markov process and a Markov process is that the former is an actual stochastic process that evolves over time, while the latter is defined as a pair of states and time, and the process has defined states at any given time.

[0026] The behavior of a sensor-based active fault-tolerant system can be divided into a fault state process and a fault diagnosis state process, each employing a stochastic process. and Describe it.

[0027] Assuming the lifespan of the sensor units installed on the dredging vessel follows an exponential distribution, then... It is a homogeneous Markov process with a finite state space. This is used to describe the system's working model (state 0) and fault state model (state 1, ..., state 2). N 1).

[0028] Definition for sufficiently small ,state i arrive j The transition probability is The formula is as follows: (3) in, and express The transition probability, This indicates that the sensor has changed from a normal state to a fault state; This indicates that the sensor state remains unchanged; and ; Represents higher-order infinitesimal terms; Indicates a unit of time; The transition matrix is N1 is the number of states.

[0029] Fault diagnosis state models are conditional stochastic processes used to describe diagnostic results. , The state is tracked with a certain random time delay and error probability to track the fault state. Assuming The state space is Then for sufficiently small ,state i arrive j The transition probability is The formula is as follows: (4) in, Indicates a given hour i The transition probability that remains unchanged. Indicates a given hour i arrive j The transition probability, and ; The generator matrix composed of transition probabilities is used Indicates that N2 is the number of states; if and only if Use a negative sign in the case of a specific condition, and a positive sign in all other cases. kThis indicates the fault diagnosis process.

[0030] Since the design purpose of an active fault-tolerant sensor system is to ensure the continuous and normal operation of the sensor, the fault diagnosis result of such a system is only twofold: the physical sensor is functioning normally or the physical sensor has failed. This is represented by the symbol... , where 0 represents the fault-free model of the physical sensor and 1 represents the effectiveness loss of the physical sensor.

[0031] transition probability parameters Depends on the failure rate of the sensing element Information can be obtained through reliability manuals, manufacturers, and accelerated life tests simulating dredger operating conditions. transition probability parameters This depends on the fault diagnosis delay distribution parameters.

[0032] The fault diagnosis delay distribution parameter is defined as the time during which the fault diagnosis process remains in a certain state without a state transition, and is denoted by the symbol [symbol missing]. express. and They represent The distribution function and probability density function. , and The following relationship must be satisfied: (5) in, and They represent The distribution function and probability density function. and Indicates a given hour The transition probability, ij Indicates a change in state. ii This indicates that the state remains unchanged.

[0033] The distribution pattern of fault diagnosis delay parameters can be determined through Monte Carlo simulation of the fault diagnosis algorithm. Common multi-sample test fault diagnosis delay distribution parameters... Satisfies Gamma distribution The specific formula for the conditional transition probability is as follows: (6) in, : No. k Next, from state i to state i j The rate parameter of the Gamma distribution; : No. k Next, from state i to state ij The shape parameter of the Gamma distribution, i.e., the shape parameter of the Gamma distribution. r ; t Time variable; m The summation variable is an integer used to calculate the normalization factor in the denominator; : indicates that under a given Gamma distribution, the th k Next from i arrive j In time t The conditional transition probability density.

[0034] Due to the unique working environment of dredging vessel sensors, they frequently experience short-term overloads followed by a return to their original performance after a period of time. This situation is represented by the symbol [symbol missing]. And introduce performance functions The performance of a sensor active fault-tolerant system at time t is evaluated. If the sensor active fault-tolerant system fails to meet performance requirements for a short period but recovers its performance within a specified time t, then it is defined as... This indicates that the system is resilient; if it cannot resume operation, it is determined to be fault F.

[0035] Two state transition diagrams are as follows Figure 1 As shown, this includes eight failure modes. (For the sake of brevity, the self-transition of each state is not shown here.) Accordingly, X(t) has a total of 17 states, namely: (7) In this table, "F" represents the unique absorption failure state, while other functional states are represented by symbols consisting of numbers and letters. Numbers represent failure modes, the letter "N" indicates satisfactory performance, and "F" indicates unsatisfactory performance within the hard deadline. The failure models and their corresponding actual failure categories are shown in Table 1. , and Represented as: (8) (9) in, This indicates that the system is in fault mode i and its performance is still acceptable (N state); The performance metric function is reflected in the pattern. i Below, state variables Performance; This indicates the fault diagnosis process; For the threshold; This indicates that the system is in fault mode i and the performance is unsatisfactory (F state). This indicates that the performance indicators have fallen below the threshold, and the system performance has deteriorated. Indicates the duration of the current state F; This indicates the maximum tolerance time (hard deadline) for state F. If this time is exceeded, the system enters the complete failure state F.

[0036] Table 1 Fault Model Coding

[0037] (Note: √ Represents fault-free mode ， (╳ represents the failure mode) Several predefined features of the model Definition 1: Model parameters Based on parameters and Establish a system model The state transition matrix of this model is For specific fault modes and fault diagnosis and identification modes, the probabilistic performance of the sensor active fault-tolerant system is defined as follows: (10) in, The fault mode is i And the FDI mode is j The probabilistic performance of the system can be estimated using Monte Carlo simulation algorithms; Fault mode i The minimum acceptable value (performance threshold) for the following performance indicators; and The system's performance index function is determined by the actual fault modes and fault diagnosis results, and measures whether the system is currently within the normal performance range. This indicates a probability calculation.

[0038] Definition 2: Stationary distribution of the Fault Diagnosis and Identification (FDI) model For a specific failure mode, the stationary distribution of the FDI model is: (11) in, This indicates that the system is actually in a fault mode. i At that time, the FDI model determines that it is a recognition pattern. j The stationary probability distribution value.

[0039] when hour, It can be generated by matrix This was calculated. Especially because the accuracy of prediction algorithms is limited, The prediction results cannot achieve 100% reliability; a scaling function must be used to measure the results of fault diagnosis. Therefore, all... Both provide a probability measure of the imperfections of FDI. Because It is a Markov process The distribution is stationary, therefore it can be calculated using typical Markov theory processes. This process only requires... The generator performs basic matrix operations.

[0040] Definition 3: and Competition probability For the elements of the transition probability matrix, set Indicates state transition The length of stay yes The dwell time. Based on Markov process theory, and It follows an exponential distribution, and its parameters in the generating matrix are defined by Equation 12. and Let represent the corresponding transition probabilities. Before the state transition, assume... and .

[0041] (12) in, The symbol represents probability, indicating the likelihood of an event occurring under given conditions.

[0042] Based on the above properties (13) (14) From formulas 12 to 14, we can see that... State transition probability It can be obtained.

[0043] (15) (16) in, represent The process occurs first; express The process occurs first; Indicates two parallel processes and In the process of real state transition Duration of stay Prior to the diagnosis of state transition process Duration of stay The probability of arrival; Conversely.

[0044] Definition 4: FDI Process Model Given The probability of an FDI process is defined as: (17) in, The calculation can be performed using Bayes' theorem. Based on the background of this embodiment, .

[0045] (18) Given The probability of an FDI process is defined as: (19) in, The solution process is similar to that of (4-17).

[0047] Definition 5: Semi-Markov kernel Define a new static procedure: for any , ,in This represents a semi-Markov process, signifying... Continuous state sequence transitions, For conversion time. Semi-Markov kernel Using matrix functions Let be the representation, and the one-step state transition probability of this function matrix is ​​shown in formula (4-20): (20) in, , and . Indicates from state to state A one-step state transition, and the state transition time No more than time t .

[0048] like ,but The calculation method is as follows: Formula 21: (twenty one) like ,but The calculation method is shown in Formula 4-22: (twenty two) For clarity, this embodiment provides a graphical representation of the semi-Markov analytical process and a visual representation of the computational logic, such as... Figure 2 As shown. In all assumptions, the initial state is assumed to be... .

[0050] Therefore, the following formula can be used to determine all semi-Markov nuclei. .

[0051] (twenty three) (twenty four) (25) (26) (27) (28) (29) in , state space , and Each by , and Indicates. When At that time, ,otherwise .

[0052] 3) Reliability model of sensor active fault-tolerant system According to the previous definition, the initial state is: Then the reliability of the sensor active fault-tolerant system at time t can be expressed as: (30) Among them, from state to state F The transition probability function can be expressed as: (31) Figure 1 Table 1 shows that, from state The transition to state F involves transitions between multiple states, therefore... The specific formula is as follows: (32) Similarly, The calculation formula is as follows: (33) Based on the characteristics of the sensing system and the active fault-tolerant sensor system of the cutter suction dredger, the parameters involved in the calculation process are given. Among them, Values ​​are taken from The fault model is shown in Table 1. Then from The value is set to 0, where 0 and 1 represent no fault and physical sensor failure, respectively. Specifically, the state transition rate is measured in hours, and based on actual equipment maintenance records of dredgers, the typical sensor replacement frequency is three years. Therefore, the physical sensor failure rate can be set to 0. In this embodiment, the failure rates of the two resolution redundancy sensors are set as follows: and Furthermore, this embodiment assumes that the fault is irreversible. The generation matrices of various Markov processes describing the fault occurrence and FDI results are as follows: (34) In the simulation test, the false alarm probability of the system is assumed to be 0.01, and the probability of the system transitioning from a false alarm state to a correct detection state is assumed to be 0.19, reflecting the system's self-correcting capability. Therefore, the transfer matrix of the fault diagnosis model is: (35) (36) in, The superscript "3" in the text indicates This indicates that the fault model only assumes that physical sensors are the source of the fault. H The interpretation of other matrices is similar, so details are omitted.

[0053] Next, four key parameter values. , , and They were introduced separately. Indicates when the fault mode is i and FDI mode is j The system's probabilistic performance at that time.

[0054] (37) A quantitative indicator representing the degree of imperfection in an FDI system.

[0055] (38) It is a static probabilistic model of the FDI process, in which .

[0056] (39) It is also a static probabilistic model of the FDI process, in which .

[0057] (40) The reliability curve of the HRSFT structure is as follows: Figure 3 As shown, the curve indicates that when the initial state is set to... At that time, the system's reliability is expected to remain at a high level for a considerable period of time. Specifically, the system's reliability will still exceed 90% after 11 hours of operation.

[0058] Figure 4 Showing from The probability of transitioning from one state to another. As shown in the diagram, the probability of transitioning to each state... The probability of a state increases over time, while the probability of transitioning to other states is relatively small. This result is highly consistent with real-world engineering scenarios; therefore, this model can accurately simulate transition states in practical engineering applications.

[0059] Since the performance of data-driven resolution redundancy sensors improves with increasing data volume, this study explores the impact of this improvement on the reliability of the sensor's active fault-tolerant system. Specifically, it analyzes the impact of different fault diagnosis false alarm rates on system reliability while keeping the failure rate and self-correction rate constant. Assuming that the system's fault diagnosis false alarm rate decreases with increasing data volume, this embodiment sets five different fault diagnosis false alarm rates (0.01, 0.0075, 0.005, 0.0025, and 0.0001) and evaluates their corresponding impact on system reliability. Table 2 lists the specific parameter settings used in the analysis of this embodiment.

[0060] Table 2 shows the change in false alarm rate for fault diagnosis as system performance increases.

[0061] like Figure 5 As shown, the system's reliability increases with the decrease in the false alarm rate of fault diagnosis, further demonstrating a negative correlation between the false alarm rate and system reliability. Simultaneously, the impact of reducing the system's false alarm rate and improving its self-correction capability on reliability was analyzed. While keeping the resolution redundancy sensor failure rate constant, reducing the false alarm rate increased the probability of recovering from a false alarm state to a correct diagnosis state. Specific parameters are shown in Table 3.

[0062] Table 3 shows the changes in fault diagnosis false alarm rate and self-correction rate as system performance increases.

[0063] Depend on Figure 6 As shown, the system's self-calibration capability increases with the amount of training data, and its reliability also increases, thus verifying a positive correlation between the self-calibration rate and the system's reliability. Specifically, this embodiment assumes that as the amount of effective learning data increases, the prediction accuracy of the resolution redundancy sensor also improves, thereby reducing its failure rate. A comparative analysis was conducted by reducing the failure rates of the two resolution redundancy sensors by an order of magnitude and keeping them at the same level as the actual sensors. Table 4 lists the details of each parameter.

[0064] Table 4 shows the change in the failure rate of the resolution redundancy sensor as system performance increases.

[0065] Table 4 (continued) shows how the failure rate of the resolution redundancy sensor changes with the increase in system performance.

[0066] like Figure 7 As shown, when the resolution redundancy sensor failure rate is reduced by an order of magnitude to At that time, with a failure rate Compared to the previous system, this one has higher reliability. This result indicates that the failure rate of the resolution redundancy sensor has a significant impact on system reliability; therefore, reducing the failure rate of the resolution redundancy sensor and improving its performance is crucial for enhancing system reliability.

[0067] like Figure 8 As shown, this embodiment compares and analyzes 16 different operating conditions, finding that increasing the amount of effective learning data can improve system performance. Specifically, the resolution redundancy sensor failure rate and fault diagnosis false alarm rate of the FDI system decrease, while the system's self-correction capability is enhanced. Compared to a constant failure rate, the system's reliability curve is higher when the resolution redundancy sensor failure rate is reduced by an order of magnitude. Furthermore, it was found that the resolution redundancy sensor failure rate, the fault diagnosis false alarm rate of the FDI system, and the self-correction rate are the most important parameters for evaluating the model. Overall, the model demonstrates good adaptability in accurately describing changes in system reliability.

[0068] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Contents not described in detail in this specification belong to prior art known to those skilled in the art.

Claims

1. A method for evaluating the active fault-tolerant reliability of a liquefied gas carrier level monitoring sensor, characterized in that: include Acquire the required sensor monitoring data and evaluation parameters, including physical sensor failure rate, virtual sensor failure rate, fault diagnosis false alarm rate, self-calibration rate, and fault diagnosis delay distribution parameters; A semi-Markov process reliability model of the sensor active fault-tolerant system is established based on the sensor monitoring data and evaluation parameters. The state space of the sensor active fault-tolerant system is defined, including normal operation state, fault state and fault diagnosis state. The state transition probability is determined according to the evaluation parameters. The reliability function R(t) of the sensor active fault-tolerant system is calculated based on the semi-Markov process reliability model. The mean time between failures (MTBF) of the sensor active fault-tolerant system is calculated based on the reliability function R(t). Output the reliability assessment results of the sensor active fault-tolerant system, including the reliability function R(t) and the mean time between failures (MTBF).

2. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 1, characterized in that: The semi-Markov process reliability model includes a fault state model and a fault diagnosis state model. The fault state model represents the process by which a physical sensor transitions from a normal operating state to a fault state. The fault diagnosis state model represents the transition process of the fault diagnosis result between a correct detection state and a false alarm state, and considers the influence of the fault diagnosis delay distribution parameter and the fault diagnosis false alarm rate on the state transition.

3. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 1, characterized in that: The reliability function R(t) and mean time between failures (MTBF) of the sensor active fault-tolerant system are calculated using the semi-Markov process reliability model. R(t) represents the probability that the sensor active fault-tolerant system will operate without failure in the time interval (0, t), and MTBF is obtained by integrating the reliability function R(t) over the time interval [0, ∞).

4. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 1, characterized in that: The sensor active fault-tolerant system adopts a hybrid redundancy sensor configuration, including one physical liquid level sensor and two virtual liquid level sensors for monitoring the same cargo liquid level information, thus forming a three-signal redundancy architecture with one main and two backups.

5. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 4, characterized in that: The two virtual liquid level sensors are generated using different algorithms: one is based on multi-sensor collaborative regression prediction to estimate the cargo liquid level using monitoring data from relevant sensors; the other is based on time series prediction to predict the cargo liquid level using historical monitoring data.

6. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 2, characterized in that: The fault diagnosis state model sets the fault diagnosis false alarm rate and self-correction rate as model parameters to characterize the probability of a diagnostic error alarm and the probability of recovering from a false alarm state to a correct diagnosis state.

7. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 1, characterized in that: This includes assessing the reliability impact of changes in evaluation parameters, which include the false alarm rate of fault diagnosis and the virtual sensor failure rate. By reducing the false alarm rate of fault diagnosis or changing the virtual sensor failure rate and recalculating the reliability function and MTBF, the reliability indices under different parameter conditions are compared to determine the degree of impact of each parameter on the reliability of the sensor active fault-tolerant system.

8. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 2, characterized in that: The state transition matrix of the semi-Markov process reliability model is generated as follows: the transition probability of the fault state is determined based on the physical sensor failure rate and the virtual sensor failure rate, and the transition probability of the fault diagnosis state is determined based on the fault diagnosis false alarm rate and the self-correction rate, thereby forming a joint state transition probability matrix.

9. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 2, characterized in that: The time interval between the occurrence of a fault and its diagnosis is modeled as a random variable for fault diagnosis delay, and the distribution parameter τ for fault diagnosis delay is used to describe the characteristics of this diagnosis delay. This diagnosis delay model is then incorporated into the calculation of the semi-Markov process reliability model.

10. The active fault-tolerant reliability assessment method for liquefied gas carrier level monitoring sensors according to claim 1, characterized in that: Reliability assessment is performed by combining the sensor fault state process and the fault diagnosis state process into a joint process. The semi-Markov process reliability model is used to simultaneously describe the sensor hardware fault state and the fault diagnosis state, thereby comprehensively calculating the reliability index of the sensor active fault-tolerant system.

Images