1242700 扬I、本案若有化學式時’請揭示最能顯示發明特徵的化華4 · 玖、發明說明： 字式· 【發明所屬之技術領域】 一種具有適應控制及失效偵測之飛控系統，尤指一種智慧型之飛 制系統，其能判別當飛行機構的飛行動作狀態不正確時，究竟為作動狀二 不完全或為該飛行機構相關實體機構已損壞，並依據判別之結果，作進二 步飛行控制之修正，以便更有效提昇飛行之安全。 【先前技術】 傳統上，近代飛行機構的飛控系統至少具有一縱向控制迴路以及一橫 向控制迴路，其中： 、 該縱向控制迴路架構中包含有縱向控制率的轉移函數f (s )，其決定升 降舱所要旋轉的角度，而，該橫向控制迴路則負責處理該飛行機構之 和偏航，主要控制副翼及方_。另，近代飛行機構的飛控系統中更包括 至少一飛控電腦，該飛控電腦需要四類基本輸入訊號，包括：一駕駛員 輸入之命令。二·加速儀、陀螺儀感測器的回授訊號。三·攻角及側滑角 的桌號。四·咼度及速度的訊號。舉例來說，駕駛員的推桿動作，上下推 動乍疋在下達俯仰率的指令;左右推之動作是在下達滾轉率的指令 如橫向控辭_移函數g(s)，其主要咖授職侧身滾轉率 而飛控電腦則根據P⑽和〜，間的誤差，決定副翼所㈣整的角度。 傳駐，不論是縱向迴路或橫向迴路，基本上，其操縱流程如圖一所示， ;圖中 私7 900由駕駛室輸入，經一第八運算單元901運算後輸入至 硬體致動器902，透過該硬體制動器9〇2驅動該硬體9〇3動作，其作紐 :=3器9〇4將量測結果經由一回授感測器905回授(Feed Back)至該 異，1算單疋901 ’此時，若係該硬體903運作結果與實際應動作數據有差 、、“第八運算單$9〇1自動會再依内部設定的數據傳出指令控制硬體 6 1242700 903修正運作直到達一定標準； 惟’上述習知飛控系統僅能感應該硬體903在作動上的數據誤差，並回 授(Feed Back)運算以調整該硬體9〇3作動至正確的狀態，但，該硬體9〇3 在被驅動後產生的作動誤差，有時係該硬體9〇3實際損壞所造成，並非單純 的作動上的數據誤差，然，習技術中用對於該硬體903損壞的部分並無法測 知，造成該硬體903於飛行中之不預期損壞但無法被感測，成為嚴重之飛安 問題之一，故設計一有效而能確實改善該飛安問題者，即為重要之課題。 【發明内容】 本發明之主要目的，在於提供一種具有適應控制及失效偵測之飛控系 統，其可於飛行中確實感測、明辨硬體損壞狀態，並確實回授告知，並透 過修改控制器内部之參數，而使飛行機構迅速恢復穩定控制之狀態，大幅 有效改善飛安問題。 本發明之次一目的，在於提供一種具有適應控制及失效偵測之飛控系 統，其可於飛行中感測相關硬體的損壞，並可回授控制相關硬體作動，而 於第一時間内安全操縱飛行機構。 為解決上述習用技術之缺失，本發明提供一種具有適應控制及失效偵 測之飛控系統，用以提供準確的飛行機構方向控制_功能，其包括· 為料讀取/輸入裝置，用以讀取/輸入至少一俯仰率、一滾轉率以及一 偏航率的指令，以控制-飛行機構之飛行方向；且，該飛行機構亦不斷讀 取至少-高度、-攻角、-側滑角以及_缝的即時數據； 至少-縱向控制器及-橫向控制器，係用以各別接受指令與數據以控 制該飛行機構相關硬體作動；該指令魏括該俯仰率、滾轉率、偏航率、 -理想俯仰率、-理想滾轉率、-理想偏鱗；該數據係包括該高度、攻 角、側滑角、動壓，該縱向控制器及該橫向控制器並於接受該前述指令並 運算前述數據後控制相關裝置； 1242700 一升降舵致動器、一副翼致動器、一方向舵致動器係分別受該縱向控 制器及該橫向控制器之控制而改變該飛行機構的飛行狀態； 一實際飛行狀態模組，用以計算及儲存該飛行機構受該升降舵致動 器、副翼致動器以及方向舵致動器控制後的實際飛行狀態； 一感測裝置，其至少包括一俯仰率感測器，用以偵測該飛行機構於飛 行中之一實際俯仰率、一滾轉率感測器，用以偵測該飛行機構於飛行中之 一實際滾轉率、一偏航率感測器，用以偵測該飛行機構於飛行中之一實際 偏航率、一飛行速度感測器，用以偵測該飛行機構於飛行中之一實際飛行 速度、一俯仰角感測器，用以偵測該飛行機構於飛行中之一實際俯仰角、 一滾轉角感測器，用以偵測該飛行機構於飛行中之一實際滾轉角、一偏航 角感測器，用以偵測該飛行機構於飛行中之一實際偏航角、一攻角感測器， 用以偵測該飛行機構於飛行中之一實際攻角，以及一側滑角感測器，用以 偵測該飛行機構於飛行中之一實際側滑角； 一常態飛行系統模組，用以讀取該縱向控制器及橫向控制器輸出數 據’再配合内定公式，運算出該飛行機構於常態飛行下應產生的理想數據； 一總誤差估測模組，係用以讀取各感測器感測的數據及該常態飛行系 統模組的數據，比對其令的誤差後輸出該誤差； 控制面異常識別模組，係用以接受該總誤差估測模組輸出之誤差以 及輸入之該實際俯仰率、實際滾轉率及實際偏航率，並配合一交越相關係 數以比對該飛行機構之飛行狀態，以判別比對之異常究係硬體運作未達標 準，亦或實係硬體損壞，另，本發明之系統並設有一升降舵燈、一方向舵 燈以及一副翼燈以配合損壞感測之結果； 另，該縱向控制器及橫向控制器並讀取一俯仰率誤差、一滾轉率誤差 及一偏航率誤差，以精確偵測該飛行機構之飛行狀態； 本發明之-種具有適應控做失效_之飛㈣統更包括： -緊急飛行模組，用以接受該俯仰率指令、滾轉率指令、偏航率指 8 1242700 令後，於飛行機構飛行的過程中，不斷讀取該飛行機構即時的高度、攻角 以及動壓，義運算出-即時緊急飛行狀態下的—緊急俯仰率、一緊急滾 轉率以及一緊急偏航率； 一總誤差估算模組，用以讀取該實際俯仰率、實際滾轉率及實際偏 航率後，再依序對應該緊急俯仰率、緊急滾轉率及緊急偏航率進行運算， 得-容錯俯仰率、-容錯轉率及—容錯偏航率，並以—容錯輸出值組形 態輸出； 一縱向參數更新法則運算模組及一橫向參數更新法則運算模組，用 以分別讀取該容錯俯仰率、容錯滾轉率、容錯偏航率以及該容錯輸出值組， 並予運算後輸出一縱向參數更新法則及一橫向參數更新法則； 一第一類神經網路運算模組、一第二類神經網路運算模組及一第三 類神經網路運算模組，該各類神經網路運算模組係用以讀取該容錯俯仰 率、容錯滾轉率、容錯偏航率、高度、攻角(對應於縱向控制)或側滑角(對 應於橫向控制）、動壓及該縱向參數更新法則或該橫向參數更新法則後，運 算得到一增益變數，進而輸入該縱向或橫向控制器； 第一開關、一第二開關及一第三開關依序對應前述該三個類神經網 路’該二個開關再依序對應該升祕異常訊號、副翼異常訊號及方向般異 常訊號的產生而被同時啟動；進而控制相關硬體反應作動，以轉飛安。 【實施方法】 印參閱圖二，由圖二所示，本發明係為一種具有適應控制及失效偵測 之飛控系統，主要用以提供—飛行機構飛行方向鋪侧之功能，其包括： 一資料讀取/輸入裝置08，用以讀取/輸入至少一俯仰率彳彳、一滾轉率12 以及-偏航率13的指令，且，該飛行機構亦不斷讀取至少—高度14、一攻 角15、一側滑角151以及一動壓16的即時數據； 至少-縱向控制器21及-橫向控制器22，用以各別依接受的指令與數 9 1242700 據，以控制飛行機構相關硬體作動，該指令係包括該俯仰率彳彳、滾轉率12、 偏航率13、一理想俯仰率141、一理想滾轉率142、一理想偏航率143，該 數據係包括該南度14、攻角15、側滑角151、動壓16，該縱向控制器21及 該板向控制器22並於接受前述指令及運算前述數據後控制飛行機構之相關 硬體作動； 一升降舱致動器31、一副翼致動器32、一方向蛇致動器33係分別受該 縱向控制器21及該橫向控制器22之控制而改變該飛行機構的飛行狀態； 一實際飛行狀態模組40，用以計算及儲存該飛行機構受該升降舵致動 器31、副翼致動器32以及方向舵致動器33控制後的實際飛行狀態； 一感測裝置50 ’用以偵測該飛行機構在飛行時的實際數據(幻），其至 少包括一俯仰率感測器51，用以偵測該飛行機構於飛行中之一實際俯仰率 ㈠)51、一滾轉率感測器52，用以偵測該飛行機構於飛行中之一實際滾轉率 (户)52、一偏航率感測器53，用以偵測該飛行機構於飛行中之一實際偏航率 (〇53、一飛行速度感測器54，用以偵測該飛行機構於飛行中之一實際飛行 速度(v)54、一俯仰角感測器55，用以偵測該飛行機構於飛行中之一實際俯 仰角0 )55、一滾轉角感測器56，用以偵測該飛行機構於飛行中之一實際滾 轉角（<〇56、一偏航角感測器57，用以偵測該飛行機構於飛行中之一實際偏 航角（V)57、一攻角感測器58，用以偵測該飛行機構於飛行中之一實際攻角 (α)58，以及一側滑角感測器59，用以债測該飛行機構於飛行中之一實際側 滑角Μ)59; 一常態飛行系統模組6〇，用以讀取該縱向控制器21及橫向控制器22 輸出的數據，再配合内定公式，運算出該飛行機構於常態飛行下應產生的 理想數據以㈨）； 一總誤差估測模組70，用以讀取該感測裝置50中各感測器所感測的實 際數據(叩))及該常態飛行系統模組6〇的理想數據(办)），比對其中的誤差 後輸出該誤差；亦即： 10 ^427〇〇 1/21242700 Yang I. If there is a chemical formula in this case, 'please reveal the Huahua 4 that can best show the characteristics of the invention. · 玖, description of the invention: Word type [Technical field to which the invention belongs] A flight control system with adaptive control and failure detection, Especially an intelligent flight control system, which can determine whether the flight state is incomplete or the related entity of the flight mechanism is damaged when the flight mechanism's flight action status is incorrect, and make progress based on the result of the judgment. Correction of two-step flight control to improve flight safety more effectively. [Previous technology] Traditionally, the flight control system of modern flight mechanisms has at least a longitudinal control loop and a lateral control loop, wherein: The longitudinal control loop architecture includes a transfer function f (s) of the longitudinal control rate, which determines The angle of rotation of the elevator cabin, and the lateral control loop is responsible for handling the sum of yaw of the flight mechanism, mainly controlling the ailerons and sides. In addition, the flight control system of modern flight agencies includes at least one flight control computer. The flight control computer requires four types of basic input signals, including a command input by a pilot. 2. Feedback signal from accelerometer and gyroscope sensor. 3. Table number of angle of attack and sideslip. 4. Signals of degrees and speed. For example, the driver ’s putter action pushes up and down to give the pitch rate command; the left and right push action is to give the roll rate command, such as the lateral rhetoric_shift function g (s), which mainly gives The side-by-side roll rate and the flight control computer determine the angle adjusted by the aileron according to the error between P⑽ and ~. Basically, its operation flow is shown in Figure 1, whether it is a vertical loop or a horizontal loop. The private 7 900 in the figure is input by the cab, and is input to the hardware actuator by an eighth arithmetic unit 901. 902. The hardware brake 903 is driven by the hardware brake 902, and its operation is as follows: = 3 device 904 sends the measurement result to a different sensor via a feedback sensor 905. ， 1 calculation sheet 901 ”At this time, if the operation result of the hardware 903 is different from the actual response data,“ the eighth operation sheet $ 9〇1 will automatically control the hardware according to the internally set data transmission instruction 6 1242700 903 amended operation until a certain standard was reached; except that the above-mentioned conventional flight control system can only sense data errors in the operation of the hardware 903 and feed back calculations to adjust the hardware 903 operation to the correct state However, the operating error of the hardware 903 after being driven is sometimes caused by the actual damage of the hardware 903. It is not simply a data error in operation. The damaged part of the body 903 cannot be detected, causing the hardware 903 to be unpredictable in flight. If it is damaged but cannot be sensed, it becomes one of the serious flying safety problems, so designing an effective and surely improving the flying safety problem is an important subject. SUMMARY OF THE INVENTION The main purpose of the present invention is to provide a kind of flying safety problem. Flight control system with adaptive control and failure detection, which can accurately detect and clearly identify the state of hardware damage during flight, and indeed feedback and inform, and by modifying the parameters inside the controller, the flight mechanism can quickly return to stability. The state of control greatly improves the flight safety problem. A second object of the present invention is to provide a flight control system with adaptive control and failure detection, which can detect the damage of related hardware during flight and can provide feedback. Control the related hardware to operate the flight mechanism safely in the first time. In order to solve the above-mentioned shortcomings in the conventional technology, the present invention provides a flight control system with adaptive control and failure detection to provide accurate direction control of the flight mechanism. _Function, which includes: is a material reading / input device for reading / inputting at least one pitch rate, one roll rate, and one yaw rate. To control the flight direction of the flight mechanism; and the flight mechanism also continuously reads real-time data of at least -altitude, -attack angle, -slip angle, and sew; at least -longitudinal controller and -transverse controller, It is used to separately receive instructions and data to control the relevant hardware actions of the flight mechanism; the instructions include the pitch rate, roll rate, yaw rate, -ideal pitch rate, -ideal roll rate, -ideal scale; The data includes the height, angle of attack, sideslip angle, dynamic pressure, the longitudinal controller and the lateral controller, and controls the related devices after receiving the foregoing instruction and calculating the foregoing data; 1242700 an elevator actuator, a pair The wing actuator and a rudder actuator are controlled by the longitudinal controller and the lateral controller respectively to change the flight status of the flight mechanism; an actual flight status module is used to calculate and store the flight mechanism The actual flight status after the elevator actuator, aileron actuator and rudder actuator are controlled; a sensing device including at least a pitch rate sensor for detecting the flight mechanism in flight An actual pitch rate and roll rate sensor for detecting an actual roll rate of the flying mechanism in flight and a yaw rate sensor for detecting one of the flying mechanisms in flight Actual yaw rate, a flight speed sensor to detect an actual flight speed of the flight mechanism in flight, and a pitch angle sensor to detect an actual pitch angle of the flight mechanism in flight A roll angle sensor to detect an actual roll angle of the flight mechanism in flight, a yaw angle sensor to detect an actual yaw angle, An angle of attack sensor is used to detect an actual angle of attack of the flight mechanism during flight, and a side slip angle sensor is used to detect an actual side slip angle of the flight mechanism during flight; a normal state The flight system module is used to read the output data of the longitudinal controller and the lateral controller, and then cooperate with the built-in formula to calculate the ideal data that the flight mechanism should generate under normal flight. A total error estimation module is used To read the data sensed by each sensor and the constant The data of the flight system module is compared with the error of the order and the error is output; the control surface abnormality identification module is used to accept the error of the total error estimation module output and the input actual pitch rate and actual roll Rate and actual yaw rate, combined with a cross-correlation coefficient to compare the flight status of the flight mechanism, to determine the abnormality of the comparison is that the hardware operation is not up to standard, or the actual hardware is damaged. The invented system is also provided with an elevator light, a rudder light and an aileron light to match the result of damage detection; in addition, the longitudinal controller and the lateral controller read a pitch rate error, a roll rate error and a The yaw rate error to accurately detect the flight status of the flight mechanism; The invention of a flying system with adaptive control failure includes:-an emergency flight module to accept the pitch rate command, roll The rate command and yaw rate refer to 8 1242700 command. During the flight of the flight mechanism, the real-time altitude, angle of attack, and dynamic pressure of the flight mechanism are continuously read. Elevation rate, an emergency roll rate, and an emergency yaw rate; a total error estimation module for reading the actual pitch rate, actual roll rate, and actual yaw rate, and then sequentially respond to the emergency pitch rate , The emergency roll rate and the emergency yaw rate are calculated to obtain -fault-tolerant pitch rate, -fault-tolerant turn rate, and -fault-tolerant yaw rate, and output in the form of -fault-tolerant output value groups; a longitudinal parameter update rule calculation module and a Lateral parameter update rule calculation module is used to read the fault-tolerant pitch rate, fault-tolerant roll rate, fault-tolerant yaw rate, and the fault-tolerant output value group, and output a longitudinal parameter update rule and a lateral parameter update rule after calculation. ; A first type of neural network operation module, a second type of neural network operation module, and a third type of neural network operation module, the various types of neural network operation module are used to read the fault tolerance After pitch rate, fault-tolerant roll rate, fault-tolerant yaw rate, altitude, angle of attack (corresponding to longitudinal control) or sideslip angle (corresponding to lateral control), dynamic pressure and the longitudinal parameter update rule or the lateral parameter update rule, Operation To a gain variable, and then input the vertical or horizontal controller; the first switch, a second switch, and a third switch sequentially correspond to the aforementioned three types of neural networks; The abnormal signal, the aileron abnormal signal, and the direction-like abnormal signal are activated at the same time; and then the related hardware response is controlled to fly to safety. [Implementation method] Please refer to Figure 2. As shown in Figure 2, the present invention is a flight control system with adaptive control and failure detection, which is mainly used to provide the function of laying the flight direction of the flight mechanism, including: The data reading / input device 08 is used for reading / inputting at least one command of the pitch rate 彳 彳, a roll rate 12 and-yaw rate 13, and the flight mechanism also continuously reads at least-altitude 14, one Real-time data of angle of attack 15, side slip angle 151, and dynamic pressure 16; at least-longitudinal controller 21 and-lateral controller 22, used to control the relevant hardware of the flight mechanism according to the received instructions and data 9 1242700, respectively. The command system includes the pitch rate 彳 彳, roll rate 12, yaw rate 13, an ideal pitch rate 141, an ideal roll rate 142, and an ideal yaw rate 143. The data includes the south degree 14. Angle of attack 15, sideslip angle 151, dynamic pressure 16, the longitudinal controller 21 and the plate to the controller 22 and control the relevant hardware of the flight mechanism after receiving the aforementioned instructions and calculating the aforementioned data; Actuator 31, an aileron actuator 32, a directional snake The actuator 33 is controlled by the longitudinal controller 21 and the lateral controller 22 to change the flight status of the flight mechanism; an actual flight status module 40 is used to calculate and store the flight mechanism actuated by the elevator The actual flight status after being controlled by the aircraft 31, aileron actuator 32, and rudder actuator 33; a sensing device 50 'for detecting actual data (magic) of the flight mechanism during flight, which includes at least a pitch Rate sensor 51 for detecting an actual pitch rate of the flight mechanism in flight ㈠) 51, a roll rate sensor 52 for detecting an actual roll rate of the flight mechanism in flight (Household) 52. A yaw rate sensor 53 is used to detect an actual yaw rate of the flight mechanism in flight (53, a flight speed sensor 54 is used to detect the An actual flight speed (v) 54 during flight, a pitch angle sensor 55 for detecting an actual pitch angle 0 of the flight mechanism during flight 55), a roll angle sensor 56 for detecting Measure one of the flight's actual roll angles in flight (< 〇56, a yaw angle sensing Device 57 for detecting an actual yaw angle (V) 57 of the flying mechanism during flight, and an angle of attack sensor 58 for detecting an actual angle of attack (α) of the flying mechanism during flight 58 and a side slip angle sensor 59 for measuring the actual side slip angle of the flight mechanism during flight 59) 59; a normal flight system module 60 for reading the longitudinal controller 21 And the data output by the horizontal controller 22, and then cooperate with the built-in formula to calculate the ideal data that the flight mechanism should generate under normal flight: i); a total error estimation module 70 for reading the sensing device 50 The actual data (叩)) sensed by each of the sensors in the sensor and the ideal data of the normal flight system module 60 (do)), the error is output after comparing the errors; that is: 10 ^ 427〇〇1 /2
En==2{ Y(k)-Y{k) I 其中， Y(k)^[v(k) p(k) q{k) r(k) a{k) 9{k) fi(k) ψ{Κ) ^)]7 响十(幻一 $⑷；㈨S⑷〜）一）了 另，該縱向控制器21及橫向控制器22並讀取一俯仰率誤差17、一滾轉 率誤差18及一偏航率誤差19，以偵測該飛行機構的飛行狀態； 一控制面異常識別模組80，用以接收該總誤差估測模組70之輸出以及 該實際俯仰率511、實際滾轉率521、實際偏航率531的輸入，並以一交 越相關係數比對飛行機構的飛行狀態，其可測知，若比對有異常反應時， 係何種因素(例如：副翼損壞、方向蛇損壞、升降舱損壞等)所造成，當然， 如無異常現象則保持常態飛行。 其中’該交越相關係數計算方法為： 其中五為膽值運算元。且，其中之Y、X可被P(滾辨）、Q(俯仰率)或 P(滾轉率）、R(偏航率)取代而分別得到RPQ或Rpr。 又ε3"、£PQ、£pR分別為預設之總警戒值、縱向控制異常警戒值及 橫向控制異常警戒值。 並利用以下之公式17?~0-1，〇卜7^(^-2,〇)|>〜>〇(縱向控制異常警戒 值)成立與否’來判斷是否為縱向控制面的升降舵異常訊號81 ; 利用|、(n，〇)—叫卜〜>0(橫向控制異常警戒值)成立與否，來 判斷是否為橫向控制面的副翼異常訊號82或方向舵異常訊號83 ; 而如果判定為升降舵異常訊號81時，即連動啟亮圖四中所示的升降舵 異常燈811，_，當戦異常訊號82或方她異常減83發生時，其相對 應之-副翼異常燈821及一方向艇異常燈831將被連動啟亮（本發明之系統 另0又有警報燈，惟該警報燈未顯示於圖示中），以告知操縱者；並同時將 異常Λ號回授至相難置以進行其他運算及操縱，而該彳貞錯的運算流程方 1242700 塊圖如附件一所示，請參照。 容錯」操 方一8_中之任，，為避免=== 發生危險，本發明之飛«社刻啟動—容錯操倾式以雜中 作，以改變相關硬體的動作，以維持飛安。 飞T二可知，本發明之飛控系統之容錯操作模式又包括： 緊心氣灯模組9G，用以接收該俯仰率指令彳彳、該滾轉率指令η、該 偏航率指令13，並於賴行機構飛行的過財，领·㈣.Μ、攻 角15、侧滑角爾及動壓16 ’進而運算出—即時緊急飛行狀態下的一緊 急俯仰率91、—緊急轉率咖及-緊急偏鮮93 ; 、 —-總誤差估算模組1GG ’肋讀取該實際俯仰率511、實際滾轉率521 及實際偏航率531 ’再依序對應該緊急俯仰率91、緊急滾轉率犯及緊急偏 航率93進行運算，得-容錯俯仰率94、—容錯滾轉率的及―容錯偏 96，並以一容錯輸出值組彳01形態輸出； -縱向參數更新法則運算模組111及—橫向參數更新法則運算模組 112，用以分別讀取該容錯俯仰率94、容錯滚轉率95、容錯偏航率96以及 該容錯輸出值組101，並予運算(運算係以習知的梯度演算法進行參數更 新，其過程如附件三所示)後輸出一縱向參數更新法則彳彳糾及一橫向參數更 新法則1121 ; 一第一類神經網路121、一第二類神經網路122及一第三類神經網路 123，該各任一類神經網路之内部運算方式如附件二所示。該各類神經網路 運算模組係用以讀取該容錯俯仰率94、容錯滾轉率95、容錯偏航率96、高 度14、攻角15(對應於縱向控制)或側滑角151(對應於橫向控制）、動壓16及 該縱向參數更新法則1111或該橫向參數更新法則1121後，運算得到一增益 變數K(其包含一PID參數：KP (比例參數）、Κ!(微分參數）、KD (積分參 12 1242700 數）），進而輸入至該縱向控制器21或橫向控制器22; -第-開關131、-第二開關132及-第三開關133依序對應前述之該第 一類神經網路12]、第二類神經網路122及第三類神經網路123，該第一開 關131、第二開關132及第三開關133再依序對應該升輸異常訊、副 翼異常訊號82及方向航異常訊號83的產生而被同時啟動，且由於任一開關 皆為改變控制器内部之參數進而改變飛行硬體的操作，故僅以第一開關131 為例敘明如下： 當第-開關131因升降艘異常訊號81產生(此時升降舱異常燈811啟 亮）、或副翼異常訊號82或方向航異常訊觸產生而開啟時，新的增益變數 K(其包含PID參數：KP、K!、KD)寫入並取代該控制縱向(橫向)控制器21(22) 中原有之增益變數K，即控制縱向(橫向)控制！!21(22)重新對升降紐動器 31、副翼致動器32及該方向航致動器33發出改變作動的指令，直到該飛行 機構實際飛行的數據接近（或符合）緊急飛行模式下的數據為止。 疋以’該飛賴構接錄織令後’該齡經由層層數據運算，最後 控制該飛行機構飛行的狀態，實務上，由圖五中，可_看出，由於縱向 指令的執行，使飛行機構在縱向軸有所動作，而第一至第三類神經網路(121 〜123)也隨之輕，並具鱗優獅娜縣，符合各觀行鋪的要求。 圖五為本發明之-實施例，說明交越相關係數曲線圖中之上部份為 氣行機構於1行時’其飛行動作的轉輪俯仰率_交越糊係數曲 線，圖中之中間部份為當升降航(水平尾翼)於t=14秒時發生毀壞時交越相 關係數曲線’由圖中之中間部份可發現，t=14秒時存在喔的曲線變化， 圖中之下部份則疋經由容錯控制器調整後的交越相關曲線，已無圖中之中 間部份的鴨變化，其飛行效果與圖中之上部份類似。 圖六係本發明之另-實施例，說明一飛行機構於飛行時，其飛行動作 的俯仰率變化情形，其中正常飛行狀況下其飛行動作的俯仰率以粗實線表 不，戰機於10秒作抬頭動作，於20秒時作回復動作，之後飛機處於平飛 13 1242700 :狀當升降納—ι4秒發生毀壞”俯仰率將侧待般動作(如圖 所^ ’表7F其飛行狀況不正常，但加人容錯控舰，其仍可作飛 值羞，如圖Γ粗虛線所7^，雖不如正常之即時正常之飛行動作且有些微 ’但至少能轉平穩飛行姿態，不致於墜毀。 4由上文可知’本發明能在第-時間，感測被驅動的硬體是為故 早S :、、作動未達標準，並確實軸魏告知収直接作峨回授(Feed ，再讀據運算成指令驅動相關硬體動作，以彌補 有效提昇磐。 、斤述係利用車义佳實施例詳細說明本發明，而非限制本發明之範 圍。大凡熟知此類技藝人士皆能日膽，適#而作些微的改變及調整，仍將 不失本I明之要義所在，亦不脫離本發明之精神和細。綜上所述，本發 明實施之具雜，誠已符合專利法中所規定之㈣專利要件 查委員惠予審視，並賜准專稿禱。 14 1242700 【附件】 附件·係本㈣之錯誤_流程方塊圖 附件二··係本發明之類神經網路架構 附件三··梯度演算法的大部内容 【圖式簡單說明】 圖一係習知技術中飛控系統流程之示意圖。 圖二係本發明之系統操作流程圖。 圖三係本發明容錯處理的流程示意圖。 圖四係本發明的燈號顯示之示意圖。 圖五係本發明之交越相關曲線之示意圖。 圖六係本發明之一實施例說明一飛行機構作飛行動 — 率(粗實線）、升降舵毀壞下的俯仰率(細虛線)金 時’其實際俯仰 的比較示意圖。 、、’4錯控制後俯仰率(粗虛線) 圖號說明 08 — 資料讀取/輸入裝置 11 - 俯仰率指令 12 — 滾轉率指令 13- 偏航率指令 14 — 高度 15- 攻角 151 -側滑角 16 — 動壓 17- 誤差偏航率 18- 誤差滾轉率 19 一 誤差俯仰率 21 - 縱向控制器 22- 橫向控制器 31 — 升降舵致動器 32 — 副翼致動器 33 — 方向舵致動器 40 — 實際飛行系統 50 - 感測裝置 15 1242700 51 —俯仰率感測器 511 52 -滾轉率感測器 521 53 —偏航率感測器 531 54 -飛行速度感測器 541 55 —俯仰角感測器 551 56 -滾轉角感測器 561 57 —偏航角感測器 571 58 —攻角感測器 581 59 —側滑角感測器 591 60 —常態飛行系統模式 70 -總誤差估測 80 -控制面異常識別機制 81 -升降舵異常訊號 811 82 -副翼異常訊號 821 83 -方向舵異常訊號 831 90 -緊急飛行模式 91 - 92 —緊急滾轉率 93- 94 -容錯俯仰率 95- 96 -容錯偏航率 100 101 -容錯總輸出值組 111 -縱向參數更新法則運算模組 112 -橫向參數更新法則運算模組 1111 --縱向參數更新法則 1121 --橫向參數更新法則 121 —第一類神經網路 122-第二類神經網路 -實際俯仰率 -實際滾轉率 -實際偏航率 -實際飛行速度 實際俯仰角 實際滾轉肖 實際偏航角 實際攻角 實際側滑角 升降舵異常燈 •副翼異常燈 方向舵異常燈 緊急俯仰率 緊急偏航率 容錯滾轉率 總誤差估算 16 1242700 123-第三類神經網路 132 —第二開關 141 —理想俯仰率 143-理想偏航率 131—第一開關 133 —第三開關 142—理想滾轉率 17En == 2 {Y (k) -Y (k) I where Y (k) ^ [v (k) p (k) q (k) r (k) a (k) 9 (k) fi (k ) ψ {Κ) ^)] 7 ring ten (magic one $ ⑷; ㈨S⑷ ~) one) In addition, the vertical controller 21 and lateral controller 22 read a pitch rate error 17, a roll rate error 18 And a yaw rate error 19 to detect the flight status of the flight mechanism; a control surface abnormality identification module 80 to receive the output of the total error estimation module 70 and the actual pitch rate 511, actual roll Input 521, actual yaw rate 531, and compare the flight status of the flight mechanism with a cross-correlation coefficient. It can be determined what factors (such as aileron damage, Damage to the snake, damage to the elevator cabin, etc.) Of course, if there is no abnormal phenomenon, keep flying normally. Among them, the calculation method of the cross-correlation coefficient is: where five is a bile value operand. Moreover, Y and X can be replaced by P (rolling rate), Q (pitch rate) or P (roll rate), R (yaw rate) to obtain RPQ or Rpr, respectively. Also ε3 ", £ PQ, £ pR are preset total alert value, vertical control abnormal alert value and horizontal control abnormal alert value, respectively. And use the following formulas 17? ~ 0-1, 0b 7 ^ (^-2, 0) | > ~ > 〇 (Vertical control abnormal alert value) is true or not 'to determine whether it is an elevator with a longitudinal control surface Anomaly signal 81; Use |, (n, 〇) —called Bu ~> 0 (transverse control abnormal alert value) is established or not to determine whether it is an aileron abnormal signal 82 or a rudder abnormal signal 83 on the lateral control surface; and If it is judged that the elevator abnormality signal 81 is activated, the elevator abnormality light 811, _ shown in Figure 4 is activated, and when the abnormality signal 82 or Fang abnormality minus 83 occurs, its corresponding-aileron abnormality light 821 And the direction boat abnormal light 831 will be activated and activated (the system of the present invention has another alarm light, but the alarm light is not shown in the figure) to inform the operator; and at the same time, the abnormal Λ number will be returned to It is difficult to perform other calculations and manipulations, and the block diagram of this erroneous calculation is shown in Annex 1, please refer to it. "Fault-tolerant" operator of any of the eighth, in order to avoid === danger, the fly of the present invention «Social moment start-fault-tolerant operation is done in a miscellaneous manner to change the action of related hardware to maintain Fei'an . It can be seen from Flight T2 that the fault-tolerant operation mode of the flight control system of the present invention further includes: a tight heart gas module 9G for receiving the pitch rate command 彳 彳, the roll rate command η, the yaw rate command 13, and After earning money from the flight agency, the leader, 领 .M, angle of attack 15, sideslip angle, and dynamic pressure 16 'were calculated—an emergency pitch rate 91 in the immediate emergency flight state, and an emergency turn rate Café. -Emergency yaw 93 ;, --- Total error estimation module 1GG 'The rib reads the actual pitch rate 511, actual roll rate 521, and actual yaw rate 531', and then responds to the emergency pitch rate 91, emergency roll The rate offender and emergency yaw rate 93 are calculated to obtain-fault-tolerant pitch rate 94,-fault-tolerant roll rate, and-fault-tolerant bias 96, and output as a fault-tolerant output value group 彳 01;-longitudinal parameter update law operation module 111 and—The lateral parameter update rule calculation module 112 is used to read the fault-tolerant pitch rate 94, the fault-tolerant roll rate 95, the fault-tolerant yaw rate 96, and the fault-tolerant output value group 101, respectively, and perform calculations (the calculation system is to learn Parameters update using the known gradient algorithm, the process is as shown in Annex III ) Outputs a vertical parameter update rule and a horizontal parameter update rule 1121; a first-type neural network 121, a second-type neural network 122, and a third-type neural network 123, each of which The internal operation of the neural network is shown in Annex 2. The various types of neural network computing modules are used to read the fault-tolerant pitch rate 94, fault-tolerant roll rate 95, fault-tolerant yaw rate 96, altitude 14, angle of attack 15 (corresponding to longitudinal control) or side slip angle 151 ( Corresponds to lateral control), dynamic pressure 16 and the longitudinal parameter update rule 1111 or the lateral parameter update rule 1121, and a gain variable K (which includes a PID parameter: KP (proportional parameter), κ! (Differential parameter) is calculated after operation. , KD (integration parameter 12 1242700 number)), and then input to the vertical controller 21 or horizontal controller 22;-the first switch 131,-the second switch 132 and-the third switch 133 sequentially correspond to the aforementioned first Neural network 12], the second neural network 122, and the third neural network 123. The first switch 131, the second switch 132, and the third switch 133 respond to the abnormal signal and the aileron in sequence. The abnormal signal 82 and the directional abnormal signal 83 are activated at the same time, and since any switch changes the parameters inside the controller and thus changes the operation of the flight hardware, only the first switch 131 is described as an example: When the-switch 131 is generated due to the abnormal signal 81 When the cabin abnormality lamp 811 is turned on), or when the aileron abnormality signal 82 or the directional abnormality signal is triggered and turned on, a new gain variable K (which includes PID parameters: KP, K !, KD) is written and replaces the longitudinal control. (Horizontal) The original gain variable K in the controller 21 (22), that is, controlling the vertical (horizontal) control! ! 21 (22) Re-issue instructions to the elevator button 31, aileron actuator 32, and directional actuator 33 until the actual flight data of the flight mechanism approaches (or meets) the emergency flight mode. Data so far.疋 The age of the flight mechanism is calculated after layered data is used to control the flight status of the flight mechanism. In practice, as shown in Figure 5, it can be seen that due to the execution of vertical instructions, The flight mechanism moves on the longitudinal axis, and the first to third types of neural networks (121 to 123) are also lighter, and have a scale of Sina County, which meets the requirements of various shops. FIG. 5 is an embodiment of the present invention, illustrating the cross-correlation coefficient curve. The upper part of the graph is the pitching rate of the runner's flight action_cross-over paste coefficient curve when the air moving mechanism is in one row. Part is the cross-correlation coefficient curve when the lift (horizontal tail) is damaged at t = 14 seconds. From the middle part of the figure, it can be found that there is a curve change at t = 14 seconds. In part, the cross-correlation curve adjusted by the fault-tolerant controller has no duck changes in the middle part of the figure, and its flight effect is similar to the upper part of the figure. FIG. 6 is another embodiment of the present invention, illustrating the change of the pitch rate of a flight mechanism during flight. The pitch rate of the flight action under normal flight conditions is indicated by a thick solid line, and the fighter aircraft is in 10 seconds. Perform a head-up action and a recovery action at 20 seconds, after which the plane is in a level flight 13 1242700: when the elevator is up and down-4 seconds of destruction, the pitch rate will act sideways (as shown in the figure ^ 'Table 7F The flight status is abnormal However, adding a fault-tolerant control ship can still be used as a fly value, as shown by the thick dashed line in Figure Γ. Although it is not as normal as normal and normal flight movements and slightly different, it can at least turn into a stable flight attitude and not crash. 4 It can be known from the above that the present invention can sense that the driven hardware is early for the first time, S: ,, and the action is not up to the standard, and it is indeed informed that the receipt is directly used for feedback. It is calculated to drive related hardware actions according to the instructions to make up for the effective improvement. The description is based on the use of Che Yijia's embodiments to explain the present invention in detail, but not to limit the scope of the invention. Anyone who is familiar with this type of technology can dared, Suitable Changes and adjustments will still not lose the essence of this document, nor deviate from the spirit and details of the present invention. In summary, the implementation of the present invention is complex and has already met the requirements of the Patent Elements Examination Committee stipulated in the Patent Law. Thank you for your review and prayers. 14 1242700 [Attachment] Attachment · This is the error of this book_Appendix 2 of the flow block diagram ·· It is the annex of the neural network architecture like the present invention ·· Most of the gradient algorithm Content [Simplified description of the drawings] Figure 1 is a schematic diagram of the flight control system flow in the conventional technology. Figure 2 is a flowchart of the system operation of the present invention. Figure 3 is a flowchart of the fault tolerance process of the present invention. Figure 4 is a lamp of the present invention Fig. 5 is a schematic diagram of the cross-correlation curve of the present invention. Fig. 6 is an embodiment of the present invention illustrating a flight mechanism's flight-rate (thick solid line) and pitch rate (fine (Dotted line) Jin Shi's actual pitch comparison diagram. ", 4" Pitch rate after control (thick dashed line) Figure number description 08 — Data reading / input device 11-Pitch rate command 12-Roll rate command 13- Rate Command 14 — Altitude 15-Angle of Attack 151-Side Slip Angle 16 — Dynamic Pressure 17-Error Yaw Rate 18-Error Roll Rate 19-Error Pitch Rate 21-Longitudinal Controller 22-Lateral Controller 31-Elevator Actuator 32 — aileron actuator 33 — rudder actuator 40 — actual flight system 50 — sensing device 15 1242700 51 — pitch rate sensor 511 52 — roll rate sensor 521 53 — yaw rate sensor Sensor 531 54-Flight speed sensor 541 55-Elevation angle sensor 551 56-Roll angle sensor 561 57-Yaw angle sensor 571 58-Attack angle sensor 581 59-Side slip angle sensor Detector 591 60 — Normal flight system mode 70-Total error estimation 80-Control surface abnormality identification mechanism 81-Elevator abnormal signal 811 82-Aileron abnormal signal 821 83-Rudder abnormal signal 831 90-Emergency flight mode 91-92 — Emergency roll rate 93- 94-fault-tolerant pitch rate 95- 96-fault-tolerant yaw rate 100 101-fault-tolerant total output value group 111-longitudinal parameter update rule calculation module 112-lateral parameter update rule calculation module 1111-longitudinal parameter Update method Rule 1121-Lateral parameter update rule 121-Type 1 neural network 122-Type 2 neural network-Actual pitch rate-Actual roll rate-Actual yaw rate-Actual flight speed Actual pitch angle Actual roll Shaw Actual Yaw angle Actual angle of attack Actual sideslip angle Elevator abnormality light • Aileron abnormality light Rudder abnormality light Emergency pitch rate Emergency yaw rate Fault tolerant roll rate Total error estimate 16 1242700 123- Type III neural network 132 — Second switch 141 —ideal pitch 143—ideal yaw rate 131—first switch 133—third switch 142—ideal roll rate 17