CN105928695A - Fault diagnosis system and method for mechanical parts of small-size unmanned helicopter - Google Patents

Abstract

A small unmanned helicopter mechanical parts fault diagnosis system and a fault diagnosis method, which belong to the mechanical parts fault diagnosis system fault diagnosis method. Utilizing the corresponding relationship between the faults of small UAVs and vibration signals, it is possible to judge whether some mechanical parts of small unmanned helicopters are faulty by analyzing the vibration signals; using the self-learning of multi-layer neural network machine learning algorithms to realize the vibration signal Learning of artificially specified features; through the supporting wireless receiving device, the vibration signal of the small unmanned helicopter measured by the three-way acceleration sensor is transmitted to the host computer in real time, and the host computer then inputs the data into the trained neural network to judge Whether there are: main propeller wear, aileron wear, tail pipe bending, tail rudder tie rod bending, tail rudder tie rod falling off, tail wear and main shaft bending faults. Real-time understanding of the health status of parts and components of unmanned helicopters, to avoid major accidents, this method is highly reliable and has a very high promotion value.

Description

A small unmanned helicopter mechanical parts fault diagnosis system and fault diagnosis method

technical field

The invention relates to a fault diagnosis system and a fault diagnosis method for mechanical parts, in particular to a fault diagnosis system and a fault diagnosis method for small helicopter mechanical parts.

Background technique

With more and more research on small unmanned helicopters, the real-time fault diagnosis of small unmanned helicopters is an inevitable problem to be solved in its development. Today's small unmanned helicopter fault diagnosis technology mainly has the following problems: firstly, it lacks real-time performance, The fault diagnosis of small unmanned helicopters is mainly to check on the ground when returning to the voyage, and lacks real-time judgment on whether a fault occurs; in addition, today's fault diagnosis of small unmanned helicopters mainly focuses on the fault diagnosis of sensors and transmission systems. Fault diagnosis of other mechanical components of small unmanned helicopters is lack of research. The probability of failure of other parts of the small unmanned helicopter is not low, and the damage caused to the small unmanned helicopter after a failure is very fatal, which will lead to a crash to a large extent and may cause considerable losses. Today, small unmanned helicopters are used more and more widely, and the fields used also include fields that require particularly high self-diagnosis performance, such as the military field. Failure, resulting in unexpected situations during the execution of tasks, the losses caused by that will be unimaginable.

Contents of the invention

The purpose of the present invention is to provide a small-sized helicopter mechanical parts fault diagnosis system and fault diagnosis method to realize real-time and intelligent detection of whether some mechanical parts of small-sized drones fail.

In order to achieve the above object, the present invention adopts the following technical scheme: the small UAV fault detection system includes: a vibration measurement device, a display module, an alarm module, a host computer and a lower computer; The signal collected by the device, and each group of vibration signals are used as the input of the trained neural network, the data is processed in real time, and it is displayed in real time whether the small unmanned helicopter fails, and if there is a failure, the unmanned helicopter will take off and land Control; the upper computer connects an external buzzer and flashing indicator light through the alarm module to realize hardware alarm; the power supply module is used for power supply of the entire system; the upper computer is the ground controller, and communicates wirelessly with the vibration measurement device and the lower computer through radio frequency transmission technology. Receive and process the vibration signal of the small helicopter and send corresponding control commands to the lower computer; the lower computer is the controller of the small unmanned helicopter, which is installed on the small helicopter and communicates directly with the upper computer. The command is sent to the lower computer, and the lower computer realizes the action corresponding to the command.

The vibration measuring device includes 2 acceleration sensors, an acceleration acquisition module, a wireless sending device and a wireless receiving device, and the wireless sending device sends the state data of the unmanned helicopter to the wireless receiving device in real time; the 2 acceleration The sensors are the first three-direction accelerometer and the second three-direction accelerometer. The acceleration sensor is pasted close to the transmission system of the unmanned helicopter. The acceleration acquisition module and the wireless transmission device are installed at the landing gear of the unmanned helicopter. The vibration situation of the helicopter; the wireless receiving device uses radio frequency transmission to communicate, the input end of the wireless sending module is connected with the acceleration acquisition module, the wireless sending device and the wireless receiving device communicate through the antenna, and the acceleration acquisition module can read the vibration signal data.

A small UAV component fault diagnosis system, comprising the following steps:

a) Carry out failure simulation of mechanical parts for small unmanned helicopters, including: main propeller wear, aileron wear, tail pipe bending, tail rudder rod bending, tail rudder rod falling off, tail wear and main shaft bending; Correspondence after simulation During the vibration data collection process, the position of the acceleration sensor cannot be changed;

b) The collected vibration signals corresponding to various faults are used as the input training samples of the multi-layer neural network, and then the fault type is defined as the classification mode label of the training samples, and the program corresponding to the multi-layer neural network is called to train in the server ; Finally, it can be realized: According to the data of a group of sampling samples, it is possible to distinguish whether the mechanical parts of the unmanned helicopter are malfunctioning? What's wrong? Then judge which level the fault severity belongs to among the three levels;

c) Copy the algorithm obtained after training to the program of the host computer, and use the vibration data of each group of samples of the unmanned helicopter as the input of the neural network to obtain the operating status of the unmanned helicopter, and judge whether there are faults and errors in the test. Fault level; and through an external display, real-time display of the operating status of the unmanned helicopter;

d) If there is a failure, the software will issue an error warning, and the buzzer connected to the host computer will emit a warning sound, the blinking indicator light will flash, and the host computer will control the return and landing of the drone;

e) If the system makes a wrong judgment or misreports, it needs to be corrected in time and let the machine learn again to continuously improve the algorithm.

Beneficial effects and advantages: By adopting the above-mentioned scheme, the machine learning algorithm based on multi-layer neural network can be applied to the fault vibration signal processing of small unmanned helicopters, which allows the controller to learn the existing data by itself and perform induction on its own , find the corresponding relationship between signals and faults, realize fault prevention, and meet the fault diagnosis requirements of small unmanned helicopter mechanical parts; using this fault diagnosis system can avoid accidents caused by damage to unmanned helicopter parts, which is convenient for everyone. Replace the parts that are about to fail. According to the configuration of the flight, the ground state and the air state are divided; and according to the control law involved, the control channels of the vertical and horizontal directions of the aircraft are reasonably divided to ensure that the test items are fully covered, and the test items are minimized, saving a lot of money. human and material resources.

Description of drawings

Fig. 1 is a diagram of the overall detection system of the diagnostic system of the present invention.

Fig. 2 is a flow chart of the preparation of the algorithm required by the present invention.

Fig. 3 is a flow chart of the fault diagnosis of the small unmanned aerial vehicle of the present invention.

Fig. 4 is a fault simulation position diagram of the present invention.

Fig. 5 is a sensor installation diagram of the present invention.

In the figure, 1. Main propeller; 2. Aileron; 3. Tail pipe; 4. Tail servo rod; 5. Empennage; 6. Main shaft; 7. The first three-way accelerometer; 9. Wireless sending device; 10. Acceleration acquisition module.

detailed description

The small UAV fault detection system includes: a vibration measurement device, a display module, an alarm module, an upper computer and a lower computer; the upper computer displays the signals collected by the vibration measurement device in real time through an external display through the display module, and uses each group of vibration signals as The input of the trained neural network is used for data processing in real time, and it is displayed in real time whether the small unmanned helicopter fails. If a failure occurs, the takeoff and landing control of the unmanned helicopter will be carried out; the host computer is connected to an external buzzer and Flashing indicator lights to realize hardware alarm; power supply module is used for power supply of the whole system; the upper computer is the ground controller, which communicates wirelessly with the vibration measurement device and the lower computer through radio frequency transmission technology, receives and processes the vibration signal of the small helicopter and sends the corresponding Control command; the lower computer is the controller of a small unmanned helicopter, which is installed on the small helicopter and communicates directly with the upper computer. The upper computer sends the control command to the lower computer through radio frequency transmission technology, and the lower computer realizes the action corresponding to the command.

The vibration measuring device includes 2 three-way acceleration sensors, an acceleration acquisition module, a wireless sending device and a wireless receiving device, and the wireless sending device sends the state data of the unmanned helicopter to the wireless receiving device in real time; the described The two acceleration sensors are the first three-direction accelerometer and the second three-direction accelerometer. The acceleration sensor is pasted near the transmission system of the unmanned helicopter. The acceleration acquisition module and the wireless transmission device are installed at the landing gear of the unmanned helicopter. Measure the vibration of the unmanned helicopter; the wireless receiving device uses radio frequency transmission for communication, the input end of the wireless sending device is connected to the acceleration acquisition module, the wireless sending device and the wireless receiving device communicate through the antenna, and the acceleration acquisition module can read the vibration signal data .

The installation method of the acceleration sensor is pasted. This method is not easy to cause damage to the UAV itself and affect the performance of the UAV. After installation, a dynamic and static balance test should be carried out to adjust the initial state of the UAV propeller. Consider the relationship between the vibration signal and the fault. The installation position should be close to the transmission system of the unmanned helicopter, and it should be easy to install and fix; the acceleration acquisition module and the wireless transmission device are installed at the landing gear of the unmanned helicopter.

Run mathematical analysis software on the server, write a machine learning algorithm based on a multi-layer neural network, and let the algorithm learn the measured vibration data. Each fault corresponds to 90 sets of vibration data at two locations, and each of the 90 sets of data The 30 groups correspond to different degrees of faults, such as main pulp wear, and are divided into 3 levels according to the degree of wear, and 30 sets of data are measured for each level.

The machine learning algorithm based on the multi-layer neural network is used to learn the vibration signals corresponding to various faults of the drone, so as to train a multi-layer neural network that can judge whether the mechanical parts of the drone are faulty; The device realizes the real-time reading of the vibration signal of the UAV, and judges in real time whether the small UAV has a fault in the experiment.

The small unmanned helicopter is modified based on the 450 helicopter model; the present invention faces the main propeller 1, the aileron 2, the tail pipe 3, the tail servo pull rod 4, the empennage 5 and the main shaft 6; the fault types are divided into main propeller wear , Aileron wear, tail pipe bending, tail servo rod bending, tail servo rod falling off, tail wear and main shaft bending.

The acceleration sensor is installed on the fuselage of the small drone. When the small drone is performing tasks, the wireless chassis with the acceleration acquisition module transmits the vibration signal collected by the acceleration sensor to the host computer in real time through radio frequency transmission.

Through experiments, a large number of vibration signals corresponding to different mechanical parts faults are collected, and the fault degree of each fault is divided into three levels. The vibration signals and fault types corresponding to various faults are used as training samples and models for multi-layer neural networks. label, train a multi-layer neural network that can distinguish the above faults. When the UAV performs a mission, a sampling of the UAV's vibration signal is used as an input of the trained multi-layer neural network, and the multi-layer neural network will output the diagnosis result.

The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings, which non-limitatively disclose that the present invention combines vibration-based fault diagnosis theory with neural network-based machine learning theory and applies it to small UAV mechanical parts Troubleshooting.

Embodiment 1: As shown in Figure 5, there are a first three-way accelerometer 7, a second three-way accelerometer 8, a wireless transmitter 9, an acceleration acquisition module 10 and a lower computer on the UAV model. There is a wireless receiving device, a display module, an alarm module and a host computer in the ground control system.

The fault diagnosis system for small unmanned helicopter components includes a wireless transmission device, a vibration measurement device, a display module, an alarm device, and a control module; the vibration measurement device, that is, a three-way acceleration sensor is pasted on the On the acceleration acquisition module, the acceleration acquisition module is inserted into the chassis with a 2.4GHz omnidirectional antenna, that is, the wireless sending device, which sends information to the ground controller in real time, thereby realizing wireless communication; after solving the driving problem of the acceleration acquisition module , the upper computer can identify the acceleration acquisition module and read the vibration signal measured by the three-way accelerometer, and then input the vibration signal into the trained multi-layer neural network to realize real-time fault diagnosis; The display will display in real time; through the software code, the fault alarm on the program interface is realized, and the host computer is connected with a buzzer and a flashing indicator light to realize the hardware alarm; if a fault occurs, the host computer will send corresponding instructions to control the drone to continue flying It is still the return trip and landing; the power supply module is used to supply power to the entire system.

A small unmanned helicopter component fault diagnosis system, comprising the following steps:

a) Carry out failure simulation of mechanical parts for small unmanned helicopters, including: main propeller wear, aileron wear, tail pipe bending, tail rudder rod bending, tail rudder rod falling off, tail wear and main shaft bending; Correspondence after simulation During the vibration data collection process, the position of the acceleration sensor cannot be changed;

b) The collected vibration signals corresponding to various faults are used as the input training samples of the multi-layer neural network, and then the fault type is defined as the classification mode label of the training samples, and the program corresponding to the multi-layer neural network is called to train in the server ; Finally, it can be realized: According to the data of a group of sampling samples, it is possible to distinguish whether the mechanical parts of the unmanned helicopter are malfunctioning? What's wrong? Then judge which level the fault severity belongs to among the three levels;

c) Copy the algorithm obtained after training to the program of the host computer, and use the vibration data of each group of samples of the unmanned helicopter as the input of the neural network to obtain the operating status of the unmanned helicopter, and judge whether there are faults and errors in the test. Fault level; and through an external display, real-time display of the operating status of the unmanned helicopter;

d) If there is a failure, the software will issue an error warning, and the buzzer connected to the host computer will emit a warning sound, the blinking indicator light will flash, and the host computer will control the return and landing of the drone;

e) If the system makes a wrong judgment or misreports, it needs to be corrected in time and let the machine learn again to continuously improve the algorithm.

Claims (3)

1. a small-sized depopulated helicopter component of machine fault diagnosis system, is characterized in that: this SUAV fault is examined Examining system includes: vibration measurement device, display module, alarm module, host computer and slave computer；Host computer is by display mould Block external-connection displayer, signal that in real time display vibration measurement device gathers, and will often organize vibration signal as having trained The input of neutral net, carries out data process in real time, and shows whether small-sized depopulated helicopter breaks down in real time, if sent out Raw fault, controls the takeoff and landing carrying out depopulated helicopter；Host computer is referred to by alarm module external buzzer and flicker Show lamp, it is achieved hardware alarms；Supply module is powered for whole system.

Small-sized depopulated helicopter component of machine fault diagnosis system the most according to claim 1, is characterized in that: institute The vibration measurement device stated comprises 2 acceleration transducers, acceleration acquisition module, a wireless base station apparatus and a nothing Line receives device, real-time being sent on radio receiver by depopulated helicopter status data of wireless base station apparatus；Described 2 acceleration transducers are respectively the first three-dimensional accelerometer and the second three-dimensional accelerometer, and acceleration transducer is pasted onto and leans on At the drive system of nearly depopulated helicopter, acceleration acquisition module and wireless base station apparatus are arranged on the undercarriage of depopulated helicopter Place, measures the Vibration Condition of SUAV；Radio receiver uses radio frequency transmission to communicate, wireless sending module Input is connected with acceleration acquisition module, and wireless base station apparatus is communicated by antenna with radio receiver, acceleration Acquisition module can read vibration signal data.

3. the method for diagnosing faults of the small-sized depopulated helicopter component of machine fault diagnosis system described in claim 1, its Feature is: fault diagnosis system comprises the following steps:

A) small-sized depopulated helicopter is carried out component of machine fault simulation, including: main slurry abrasion, aileron abrasion, tail pipe Bending, tail actuator pull bar bend, tail actuator pull bar comes off, empennage abrasion bends with main shaft；The vibration of correspondence is carried out after simulation Data acquisition, during vibrating data collection, the position of acceleration transducer can not become；

B) using vibration signal corresponding for the various faults collected as the input training sample of multilayer neural network, then by event Barrier type definition is the classification mode label of training sample, calls the program that multilayer neural network is corresponding, carries out in the server Training；May finally realize: whether the component of machine telling depopulated helicopter according to the data of one group of sample occurs Fault？There occurs any fault？Which rank of during the failure judgement order of severity belongs to three ranks again；

C) algorithm obtained after training is copied in the program of host computer, by the vibration number of each for depopulated helicopter group of sample According to the input as neutral net, the running status of depopulated helicopter can be obtained, it may be judged whether occur test in fault with Fault rank；And by external-connection displayer, display depopulated helicopter running status in real time；

D) if there is fault, software will send false alarm, and the buzzer that host computer is connected will give a warning Ringing, flashing indicator light will flash, and host computer will control making a return voyage and stopping fall of unmanned plane；

If e) system misjudgment, report by mistake, need to correct in time, allow machine learn again, constantly improve this algorithm.