Fuel sensor credibility estimation method

Abstract

The invention relates to the field of sensor credibility estimation, in particular to a fuel sensor credibility estimation method, which comprises the following steps of: firstly, judging the effectiveness of an aircraft fuel sensor, and designing a time dimension credibility function and a space dimension credibility function of the sensor aiming at the sensor of which the judgment result is effective; the credibility of the sensor in the time dimension and the credibility of the sensor in the space dimension are calculated, and then the credibility of the two dimensions is fused by using an improved D-S evidence theory fusion formula to complete evaluation of the credibility of the aircraft fuel sensor. Compared with the prior art, the problem that the measurement result is inaccurate, unstable and unreliable due to the fact that aircraft fuel oil measurement is easily influenced by factors such as flight attitude and sensor failure is solved.

Description

Technical Field

[0001] The present invention relates to the field of sensor credibility estimation, and particularly to a method for estimating the credibility of a fuel sensor. Background Art

[0002] With the rapid development of aviation technology, the aircraft fuel measurement system, as a core component for flight safety and performance evaluation, has received high attention in the industry and the scientific research field. However, due to the complex structure, irregular shape of the aircraft fuel tank, errors existing in the sensor itself, fluctuations in the oil level surface, and dynamic changes in the flight state of the aircraft, the data obtained by the sensor is not completely reliable. Therefore, how to determine the credibility of the current sensor has become an important problem that urgently needs to be solved to improve the stability of the aircraft fuel measurement system. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for estimating the credibility of a fuel sensor. By designing a credibility function in the time dimension and a credibility function in the space dimension, the credibility in the time dimension and the credibility in the space dimension are calculated, and then the improved D-S evidence theory fusion rule is used to fuse the obtained credibility in the time dimension signal and the credibility in the space dimension to evaluate the state of the aircraft fuel sensor, thereby improving the accuracy of the fuel measurement system.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] A method for estimating the credibility of a fuel sensor, comprising the following steps:

[0006] S1. Based on the characteristic of the reading time consistency of the aircraft fuel sensor, judge the state of the current aircraft fuel sensor; the state of the aircraft fuel sensor includes two types: an effective state and a failure state. The effective state means that the aircraft fuel sensor is in normal operation, that is, the reading is stable; the failure state means that the reading of the aircraft fuel sensor remains unchanged or has a violent fluctuation.

[0007] S2. Calculate the credibility of the sensor determined to be effective in S1. The credibility includes the credibility in the time dimension and the credibility in the space dimension;

[0008] S3. Use the improved D-S evidence theory fusion formula to perform a fusion calculation on the credibility in the time dimension and the credibility in the space dimension calculated in step S2;

[0009] S4. Compare the fusion calculation result obtained in step S3 with a preset credibility threshold. If it exceeds the preset credibility threshold, the sensor is considered credible; if it does not exceed, the sensor is considered not credible. The fusion calculation result is used as the final credibility output.

[0010] Further, the method for determining the status of the current aircraft fuel sensor in S1 based on the time consistency characteristic of the aircraft fuel sensor readings includes the steps:

[0011] S1.1. Calculate the difference between the readings at each moment and the reading at the previous moment within the set window length to obtain a difference sequence;

[0012] S1.2. Calculate the variance based on the difference sequence, and compare the variance with the set threshold. If it is equal to or exceeds the set threshold, it is determined that the sensor data at the current moment is in a state of severe fluctuation, and the sensor is considered to have failed. If it does not exceed the set threshold, the sensor is considered valid;

[0013] S1.3. Count the number of zeros in the difference sequence. If the number of zeros exceeds half of the window length, it is determined that the sensor output remains unchanged, and the sensor is considered to have failed; and set the historical data penalty of all failed sensors and the cross - reading penalty of other sensors for this sensor to 0.9.

[0014] Further, the time - dimension credibility in S2 is calculated using the time - dimension trust function based on the difference between the historical readings and the current readings of each sensor within the set time; the trust function of the time dimension is:

[0015] m i-i (credible)=1 - p i-i m i-i (uncredible)=1 - m u-i (credible)

[0016]

[0017] Where p i-i is the historical reading penalty, p′ i-i is the weighted sum of the normalized results of the differences between the historical readings and the current readings of sensor i within the set time, e is the natural logarithm, n is the window length, w is the weight within the window, and h i (t) is the reading of sensor i at time t.

[0018] Further, the calculation method of the space - dimension credibility in S2 includes the steps:

[0019] S2.1. Calculate the current fuel liquid level based on the current reading points of each sensor. If the number of sensors whose current readings are within the range and neither zero nor full - scale is less than or equal to 3, execute step S2.1.1. If the number of sensors whose current readings are within the range and neither zero nor full - scale is greater than 3, execute step S2.1.2;

[0020] S2.1.1. Set the three-axis accelerations of the aircraft at the current moment as \(a_x\), \(a_y\), and \(a_z\). Let the normal vector of the oil level surface be \(n\). According to the current reading point of sensor \(i\) and the normal vector of the oil level, use the point-normal form to fit the oil level surface at the current moment.

[0021] S2.1.2. According to the current reading point coordinates of each sensor, use the least squares method to fit the oil level surface at the current moment.

[0022] S2.2. Substitute the \(x\) and \(y\) coordinates of the sensor spatial coordinates into the oil level surface equation to determine the intersection points of the oil level surface at the current moment and each sensor. Calculate the difference between each intersection point and the current reading of the sensor, and normalize it to be used as the self-reading penalty. Fit a plane with each sensor and the normal vector of the oil level surface. Use the intersection points of the fitted plane and other sensors as the projection results of this sensor onto other sensors. Calculate the difference between the projection point and the current reading point and normalize it to be used as the cross-reading penalty.

[0023] S2.3. Define the trust function in the spatial dimension as: \(m\) i-j (credible) = 1 - \(p\) i ×\(p\) i-j ; where \(p\) i is the self-reading penalty, and \(p\) i-j is the cross-reading penalty. According to the self-reading penalty and cross-reading penalty calculated in S2.2, use the trust function in the spatial dimension to calculate the spatial dimension credibility.

[0024] Further, the improved D-S evidence theory fusion formula used in S3 is:

[0025]

[0026] \(k\) i = \(m\) i-1 (credible) × … × \(m\) i-n (credible) + \(m\) i-1 (incredible) × … × \(m\) i-n (incredible)

[0027]

[0028] \(k\) ab-i = 1 - [\(m\) a-i (credible) × \(m\) b-i (credible) + \(m\) a-i (incredible) × \(m\) b-i (incredible) ]

[0029] where \(k\) i is the evidence weight, \(q(A)\) is the average support degree of evidence, \(\epsilon\) is the evidence credibility, and \(k\) ab-i is the conflict coefficient. is the average sum of the conflict magnitudes for each pair of evidence sources among all evidence sources, k ab-i is the conflict magnitude of the degree of trust support of sensors a and b for sensor i.

[0030] Furthermore, step S3 further includes performing exponential weighting based on the spatial distances between the sensors before fusion to update the credibility of each sensor in the spatial dimension; the exponential weighting formula is:

[0031]

[0032] The formula for updating the credibility of a sensor in the spatial dimension is expressed as:

[0033]

[0034] where i represents the i-th sensor, j represents the j-th sensor, and w j represents the weight of the j-th sensor.

[0035] Further, the preset credibility threshold in step S4 is obtained by calculating and statistically analyzing the credibility distribution of each attitude of the aircraft based on a large amount of real data during the actual operation of the aircraft.

[0036] The fuel sensor credibility estimation method of the present invention completes the evaluation of the credibility of the aircraft fuel sensor by designing the trust degree functions in the time dimension and the spatial dimension of the sensor, calculating the credibility in the time dimension and the spatial dimension of the sensor, and then using the improved D-S evidence theory fusion formula to fuse the credibility of these two dimensions.

[0037] For the calculation of the credibility in the time dimension, the historical readings within a certain window length are subtracted from the current sensor readings, and weighting is performed according to the time length from the current moment, increasing the weight of the closer moment and weakening the influence of the farther moment on the current moment. Finally, the weighted result is normalized by tanh to obtain the historical reading penalty, and the credibility in the time dimension of the sensor is obtained by using 1 minus the obtained historical reading penalty.

[0038] For the calculation of the credibility in the spatial dimension, the distance between the intersection point of the plane fitted by the reading points of each sensor and the reading point of each sensor is normalized as the self-reading penalty of the sensor, and the distance between the projection point obtained by projecting the reading points of each sensor along the direction of the fitted plane to each sensor and the reading point of the sensor is used, and after normalization processing, it is used as the cross-reading penalty between sensors. The cross-reading penalty is weighted by the self-reading penalty, and the credibility in the spatial dimension is obtained by using 1 minus the weighted result.

[0039] In the fusion stage, an improved D-S evidence theory fusion formula is used to fuse and calculate the credibility in the time dimension and the credibility in the space dimension, so as to obtain the credibility of the current reading of the sensor. In addition, before fusion, exponential weighting is first performed according to the spatial distance between sensors to update the credibility of each sensor in the space dimension.

[0040] In summary, in view of the special working conditions of aircraft fuel sensors, the present invention makes full use of information such as its time variation law and spatial layout characteristics, calculates its credibility in the time and space dimensions respectively, and uses the D-S evidence theory for fusion, solving the problems that the aircraft fuel measurement is easily affected by factors such as flight attitude and sensor failure, resulting in inaccurate, unstable and unreliable measurement results. Brief Description of the Drawings

[0041] Figure 1 is a schematic flow chart of the fuel sensor credibility estimation method;

[0042] Figure 2 is a schematic diagram of the spatial credibility calculation method. Detailed Embodiment

[0043] Next, the technical solution of the present application will be clearly and completely described in conjunction with the drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0044] As Figure 1 shown, a fuel sensor credibility estimation method provided in this embodiment includes the following steps:

[0045] S1. Judge the effectiveness of the sensor.

[0046] When the aircraft fuel sensor is working normally, the change of its reading is usually in a relatively stable state. When it fails, the output of the aircraft fuel sensor often has violent fluctuations or remains unchanged. This phenomenon indicates that the aircraft fuel sensor has the characteristic of time consistency. Therefore, the time consistency characteristic of the sensor reading can be used to judge whether the sensor is in an effective state or a failure state. The implementation method is as follows:

[0047] S1.1. Within the set window length, calculate the difference between the reading at each moment and the reading at the previous moment to obtain a difference sequence; the calculation formula of the difference sequence is:

[0048] sub i (t)=[h i (t)-h i (t - 1),h i(t - 1) - h i (t - 2), h i (t - 2)

[0049] -h i (t - 3), ……, h i (t - w) - h i (t - w - 1)]

[0050] Among them, sub i (t) represents the difference sequence of the i-th sensor at time t, h i (t) represents the reading value of the i-th sensor at time t, n represents the window length, and in this embodiment, the window length n = 8 is set.

[0051] S1.2. Calculate the variance according to the difference sequence, compare the calculated variance with the set threshold. If it exceeds the set threshold, it is determined that the sensor data at the current moment is in a state of severe fluctuation and the sensor is considered to fail. If it does not exceed the set threshold, the sensor is considered effective. In this embodiment, the formula for calculating the variance according to the difference sequence is:

[0052]

[0053] Among them, s i (t) represents the variance, mean((sub i (t)) represents the mean of the difference sequence. k represents each moment within the window.

[0054] S1.3. Count the number of 0s in the difference sequence. If the number of 0s exceeds half of the window length, it is determined that the sensor output remains unchanged and the sensor is considered to fail. And set the historical data penalty of all failed sensors and the cross-reading penalty of other sensors for this sensor to 0.9.

[0055] S2. Calculate the credibility of the sensors determined to be effective in S1. The credibility includes the credibility in the time dimension and the credibility in the space dimension.

[0056] The calculation method of the credibility in the time dimension is:

[0057] Calculate the difference between the historical readings and the current readings of the sensor at each moment within the window, and perform weighted summation, and then use the tanh function for normalization processing to obtain the historical reading penalty p i-i , and finally calculate the credibility in the time dimension using the credibility function in the time dimension. The credibility function in the time dimension is:

[0058] m i-i (credible) = 1 - p i-i m i-i(Unreliable) = 1 - m i-i (Reliable)

[0059]

[0060] Where p i-i is the historical reading penalty, p' i-i is the weighted sum of the normalized results of the differences between the historical readings and the current readings of sensor i within the set time, e is the natural logarithm, n is the window length, n = 8, w is the weight within the window, and h i (t) is the reading of the i-th sensor at time t. In this embodiment, the window length n = 8, and the weights w within the window are w = [0.2, 0.2, 0.15, 0.15, 0.1, 0.1, 0.05, 0.05].

[0061] The calculation method of the spatial dimension credibility is as follows Figure 2 shown:

[0062] S2.1. Calculate the current oil level surface at the current moment according to the current reading points of each sensor. If the number of sensors whose current readings are within the range and neither 0 nor full scale is less than or equal to 3, execute step S2.1.1; if the number of sensors whose current readings are within the range and neither 0 nor full scale is greater than 3, execute step S2.1.2;

[0063] S2.1.1. Let the three-axis acceleration of the aircraft at the current moment be ax, ay, az as the normal vector of the oil level surface, that is, let the normal vector of the oil level surface at the current moment be v = (ax, ay, az). Using the point-normal form, and using the reading points of the sensors, that is, (x i-Bottom 、y i-Bottom 、z i-Bottom + h(t)), where x i-Bottom 、y i-Bottom 、z i-Bottom are the three-dimensional coordinates of the bottom point of sensor i respectively, fit an oil level surface, and take the average value of the fitted oil level surface to obtain the finally fitted oil level surface;

[0064] S2.1.2. According to the current reading point coordinates of each sensor, use the least squares method to fit the oil level surface at the current moment;

[0065] S2.2. After obtaining the oil level surface, substitute the x and y coordinates of the sensor spatial coordinates into the oil level surface equation, calculate the intersection points of the current oil level surface and each sensor, calculate the difference between each intersection point and the current reading of the sensor, and normalize it to be used as its own reading penalty; the calculation formula of the self-reading penalty is:

[0066]

[0067] Where pi Indicates the self-reading penalty, h i-i (t) represents the intersection point of the oil liquid level and each sensor at time t, ih i (t) represents the reading value of sensor i at time t, p′ i Represents the normalization result of the distance between the intersection point of the oil liquid level and sensor i and the reading point of sensor i.

[0068] Perform plane fitting on each sensor and the normal vector of the oil liquid level, take the intersection point of the fitted plane and other sensors as the projection result of this sensor onto other sensors, calculate the difference between the projection point and the current reading point and perform normalization processing to serve as the cross-reading penalty. The calculation method of the cross-reading penalty is:

[0069]

[0070]

[0071] Among them, p i-j Is the cross-reading penalty, p′ i-j Is the normalization result of the distance between the projection point of sensor i on sensor j and the reading point of sensor j.

[0072] S2.3. Define the trust function of the spatial dimension as: m i-j (credible) = 1 - p i ×p i-j ; Among them, p i Is the self-reading penalty, p i-j Is the cross-reading penalty; According to the self-reading penalty and cross-reading penalty calculated in S2.2, use the trust function of the spatial dimension to calculate the spatial dimension credibility.

[0073] S3. Use the improved D-S evidence theory fusion formula to perform fusion calculation on the time dimension credibility and spatial dimension credibility calculated in step S2.

[0074] In the spatial dimension, the credibility provided by sensors closer to sensor i is more reliable than that provided by sensors farther away. Therefore, before fusion, this embodiment adopts a method of exponentially weighting the spatial credibility provided by other sensors according to the distance between other sensors and sensor i, increasing the influence of sensors closer in distance and weakening the influence of sensors farther away. Specifically:

[0075] S3.1. Calculate the spatial distance d between sensor i and other sensors i-j , then the weight of other sensors is Based on the other sensors after exponential weighting, the credibility update formula of sensor j on sensor i in the spatial dimension is expressed as:

[0076]

[0077] Among them, i represents the i-th sensor, j represents the j-th sensor, and w j represents the weight of the j-th sensor.

[0078] S3.2. Use the improved D-S evidence theory fusion formula to perform fusion calculation on the credibility in the time dimension and space dimension. The calculation process is as follows:

[0079] The traditional D-S evidence theory fusion formula is:

[0080]

[0081] k i = m i-1 (credible) ×... × m i-n (credible0 + m u-1 (incredible0 ×... × m i-n (incredible)

[0082] Since there is a problem that when the credibility of a certain evidence source is very low, it will dominate the entire synthesis result in the traditional D-S evidence theory fusion formula. Therefore, in this embodiment, parameters such as the conflict coefficient, evidence credibility, and evidence average support degree are introduced to balance the evidence conflict problem, and the traditional improved D-S evidence theory fusion formula is improved to obtain the improved D-S evidence theory fusion formula as:

[0083]

[0084] Among them, k i is the evidence weight, q(A) is the evidence average support degree, ε is the evidence credibility, and k ab-i is the conflict coefficient. The calculation method of k ab-i is as follows:

[0085] k i = m i-1 (credible) ×... × m i-n (credible0 + m u-1 (incredible0 ×... × m i-n (incredible)

[0086]

[0087] k ab-i = 1 - [m a-i (credible) × m b-i (credible) + m a-i (incredible) × m b-i (incredible)].

[0088] S4. Calculate and statistically analyze the credibility distribution of each aircraft attitude based on a large amount of real data during actual aircraft operation to obtain a credibility threshold; compare the fusion calculation result obtained in step S3 with the preset credibility threshold. If it exceeds the preset credibility threshold, the sensor is considered credible; otherwise, it is considered non-credible. The fusion calculation result is used as the final credibility output.

[0089] As described above, only the specific implementation manners of the present invention are concerned. Any feature disclosed in this specification, unless specifically described, can be replaced by other equivalent or similar-purpose alternative features; all the disclosed features, or all the steps in any method or process, except for mutually exclusive features and / or steps, can be combined in any manner.

Claims

1. A fuel sensor credibility estimation method, characterized in that: The following steps are involved: S1. Based on the reading time consistency characteristic of the aircraft fuel sensor, determine the current state of the aircraft fuel sensor; the aircraft fuel sensor state includes a valid state and a failed state. The valid state means that the aircraft fuel sensor is in normal working state, i.e., the reading is stable; Failure is a state where the aircraft fuel sensor reading remains constant or fluctuates drastically; S2, calculate the credibility of the sensor judged as valid by S1, the credibility includes the credibility of time dimension and the credibility of space dimension; S3, using the improved DS evidence theory fusion formula, the time dimension credibility and space dimension credibility calculated in step S2 are fused and calculated; S4. Compare the fusion calculation result obtained in step S3 with a preset credibility threshold. If the preset credibility threshold is exceeded, the sensor is considered to be credible. If not, the sensor is considered to be untrustworthy. The fusion calculation result is used as the final credibility output.

2. A fuel sensor credibility estimation method according to claim 1, characterized in that: The method S1 for determining the current state of the aircraft fuel sensor based on the time consistency characteristics of the aircraft fuel sensor readings comprises the following steps: S1.

1. Within the set window length, calculate the difference between the reading at each moment and the reading at the previous moment to obtain the difference sequence; S1.

2. Calculate the variance based on the difference sequence and compare the variance with the set threshold. If it is equal to or exceeds the set threshold, it is judged that the sensor data at the current moment is in a state of violent fluctuation and the sensor is considered to be invalid. If it does not exceed the set threshold, the sensor is considered to be valid. S1.

3. Count the number of 0s in the difference sequence. If the number of 0s exceeds half of the window length, the sensor output is determined to remain unchanged and the sensor is considered to be failed. The historical data penalty for all failed sensors and the cross-reading penalty for the sensor by other sensors are set to 0.

9.

3. A fuel sensor credibility estimation method according to claim 1, characterized in that: The time dimension credibility in S2 is calculated based on the difference between the historical readings and the current readings of each sensor within a set time using a time dimension credibility function; the time dimension credibility function is: m i-i (credible) = 1-p i-i m i-i (untrustworthy) = 1-m i-i (Reliable) Among them, p i-i is the historical reading penalty, p′ i-i is the weighted sum of the normalized difference between the historical readings and the current readings of sensor i within the set time, e is the natural logarithm, n is the window length, w is the weight within the window, and h i (t) is the reading of sensor i at time t.

4. A fuel sensor credibility estimation method according to claim 3, characterized in that: The method for calculating the spatial dimension credibility in S2 comprises the steps of: S2.

1. Calculate the oil level at the current moment according to the current reading point of each sensor. If the number of sensors whose current readings are within the range and neither 0 nor full scale is less than or equal to 3, execute step S2.1.

1. If the number of sensors whose current readings are within the range and neither 0 nor full scale is greater than 3, execute step S2.1.

2. S2.1.1, assuming that the three-axis acceleration of the aircraft at the current moment is ax, ay, and az are the normal vectors of the oil surface. According to the current reading point of sensor i and the oil normal vector, the oil surface at the current moment is obtained by using the point method fitting; S2.1.

2. According to the coordinates of the current reading points of each sensor, the oil level at the current moment is fitted using the least square method; S2.2, Substitute the x and y coordinates of the sensor space coordinates into the oil level equation, determine the intersection of the oil level and each sensor at the current moment, calculate the difference between each intersection point and the current reading of the sensor, and normalize it as the self-reading penalty; perform plane fitting on each sensor and the oil level normal vector, use the intersection of the fitted plane and other sensors as the result of the projection of the sensor to other sensors, calculate the difference between the projection point and the current reading point and normalize it as the cross-reading penalty; S2.

3. Define the trust function of the spatial dimension as: m i-j (credible) = 1-p i ×p i-j ; Among them, p i is the penalty for its own reading, p i-j is the cross-reading penalty; according to the self-reading penalty and cross-reading penalty calculated in S2.2, the spatial dimension credibility is calculated using the trust function of the spatial dimension.

5. A fuel sensor credibility estimation method according to claim 4, characterized in that: The improved DS evidence theory fusion formula used in S3 is: k i =m i-1 (Trustworthy)×…×m i-n (credible)+m i-1 (Untrustworthy)×…×m i-n (Unreliable) k ab-i =1-[m a-i (Reliable)×m b-i (credible)+m a-i (Untrustworthy)×m b-i (Not credible)] Among them, k i is the weight of evidence, q(A) is the average support of evidence, ε is the credibility of evidence, k ab-i is the conflict coefficient, is the average sum of the conflict sizes of each pair of evidence sources among all evidence sources, k ab-i is the conflict degree of trust support between sensors a and b on sensor i.

6. A fuel sensor credibility estimation method according to claim 5, characterized in that: The S3 also includes exponential weighting according to the spatial distance between the sensors before fusion to update the credibility of each sensor in the spatial dimension; the exponential weighting formula is: The formula for updating the sensor's credibility in the spatial dimension is expressed as: Where i represents the i-th sensor, j represents the j-th sensor, and w j represents the weight of the j-th sensor.

7. A fuel sensor credibility estimation method according to claim 1, characterized in that: The preset credibility threshold in S4 is obtained by calculating and counting the credibility distribution of each posture of the aircraft based on a large amount of real data in the actual operation of the aircraft.

Images