阅读说明：本技术 一种新型的弹射座椅程序控制方法 (Novel ejection seat program control method ) 是由 毛晓东 冯晓晗 白庭蝶 庞丽萍 于 2019-12-12 设计创作，主要内容包括：本发明公开一种新型的弹射座椅程序控制方法,属于航空弹射救生技术领域,该方法首先根据弹射座椅在低空不利姿态情况下弹射的安全救生要求,设计姿态控制方案,确定不同姿态控制方案的执行机构、控制参数和弹射状态参数；然后建立弹射座椅弹射全过程的数学模型及仿真程序,通过仿真得到使性能最优的样本点集,并用样本点集训练和测试BP神经网络,最后将得到的BP神经网络模型写入弹射座椅程序控制器。本发明可以有效避免对状态参数进行空间划分,不受弹射状态参数及控制参数数量的限制,并且保证在状态空间全局范围内控制参数的最优性。(The invention discloses a novel ejection seat program control method, which belongs to the technical field of aviation ejection lifesaving, and comprises the steps of firstly designing an attitude control scheme according to the safety lifesaving requirement of ejection of an ejection seat under the condition of low-altitude unfavorable attitude, and determining actuating mechanisms, control parameters and ejection state parameters of different attitude control schemes; and then establishing a mathematical model and a simulation program of the whole ejection process of the ejection seat, obtaining a sample point set with optimal performance through simulation, training and testing a BP neural network by using the sample point set, and finally writing the obtained BP neural network model into an ejection seat program controller. The invention can effectively avoid space division of the state parameters, is not limited by the ejection state parameters and the number of the control parameters, and ensures the optimality of the control parameters in the global range of the state space.)

1. A novel ejection seat program control method is characterized by comprising the following steps:

step 1: designing a posture control scheme according to the safety lifesaving requirement of ejection of the ejection seat under the condition of low-altitude unfavorable posture;

step 2: determining actuating mechanisms of different attitude control schemes, and determining control parameters and ejection state parameters under the different attitude control schemes;

and step 3: establishing a mathematical model and a simulation program of the whole ejection process of the ejection seat to realize simulation calculation of the trajectory posture of the whole ejection process;

and 4, step 4: fixing a group of ejection state parameters under different attitude control schemes, and calculating to obtain control parameters meeting the optimal performance according to the optimal calculation model by adopting the simulation program in the step 3;

and 5: continuously changing the ejection state parameters so as to determine the optimal control parameters in the overall range of the ejection state parameters and form point set mapping;

step 6: establishing a BP neural network model, taking point set mapping as a sample point set, performing training and testing on the neural network, and obtaining the BP neural network model which meets the precision requirement after testing;

and 7: and (4) writing the BP neural network model obtained in the step (6) into an ejection seat program controller, and realizing final program control.

2. The novel ejection seat program control method according to claim 1, characterized in that:

the attitude control scheme includes: roll attitude control, pitch attitude control and roll pitch simultaneous control;

the actuating mechanisms of the different attitude control schemes are as follows:

the actuating mechanism for controlling the rolling posture comprises: the left and right rolling postures control the rocket and the main rocket packet to be switched on and off;

the actuating mechanism for pitch attitude control is as follows: a pitching attitude control rocket;

the executing mechanism for roll and pitch simultaneous control comprises: a left-right rolling attitude control rocket, a main rocket packet switch and a pitching attitude control rocket;

the control parameters and the ejection state parameters under different attitude control schemes are as follows:

controlling the rolling posture: the control parameters are the ignition time interval DelayTime of the left and right rolling attitude control rocket and a main rocket packet switch RocketSwitch; the state parameters are ejection speed and roll angle;

pitch attitude control: the control parameter is the ignition time PitchRocktTime of the pitching attitude control rocket; the state parameters are ejection speed and a diving angle;

roll and pitch simultaneous control: the control parameters are the ignition time interval DelayTime of the left-right rolling attitude control rocket, a main rocket packet switch RocktSwitch and the ignition time PitchRocktTime of the pitching attitude control rocket; the state parameters are ejection speed, roll angle and dive angle.

3. The novel ejection seat program control method as claimed in claims 1 and 2, characterized in that the process of step 4 is as follows:

step 4.1: judging the type of the executed control scheme, and if the control scheme is a rolling attitude control scheme, executing the step 4.2 to the step 4.5; if the pitch attitude control scheme is adopted, executing the step 4.6 to the step 4.9; if the roll and pitch are controlled simultaneously, executing the step 4.10 to the step 4.13;

step 4.2: determining decision variables as DelayTime and RocktSwitch according to a rolling attitude control scheme, wherein the DelayTime is a rolling attitude control rocket ignition time interval, and the RocktSwitch is a main rocket packet cut-off switch;

step 4.3: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(DelayTime,RocketSwitch)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.4: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time interval of the left and right rolling attitude control rocket is defined as 0.1 s; the parameter value of the main rocket packet disconnecting switch is defined as 0 or 1, wherein 0 represents disconnection, and 1 represents normal work; the ParachuteTime represents the time interval between the free flight and the launching of the escape parachute, and the specific numerical value can be obtained by subtracting the finish time of the cabin exiting stage from the parachute launching delay time in the simulation calculation;

when the DelayTime is 0, the left and right attitude rockets are simultaneously ignited, or the attitude rockets do not work, and the attitude correction is not needed;

when DelayTime is ParachuteTime +0.1, indicating that the ignition time of the left attitude rocket is later than the parachute shooting time, namely only the right attitude rocket is ignited, and finishing the maximum rolling attitude correction by the attitude rocket at the moment;

step 4.5: the fixed state parameters are ejection speed EjectionVelocity and transverse turning attitude angle RollAngle (+), and the decision variables which enable the track height to be the highest when the lifesaving parachute is fully opened under the state parameters meeting the constraint conditions are obtained by adopting an optimization algorithm and are the values of DelayTime and RocktSwitch;

step 4.6: determining a decision variable as PitchRocktTime according to the pitch attitude control scheme, wherein the PitchRocktTime is the ignition time of the pitch attitude control rocket;

step 4.7: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(PitchRocketTime)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.8: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time of the pitching attitude rocket is defined as 0.1 s; and its ignition time must be earlier than the parachute firing time, otherwise it does not need to work;

step 4.9: the fixed state parameters are ejection speed EjectionVelocity and a diving angle PitchAngle, and an optimization algorithm is adopted to obtain a decision variable which enables the track height when the lifesaving parachute is fully opened to be the value of PitchRocktTime under the state parameters meeting constraint conditions;

step 4.10: determining decision variables as DelayTime, RocktSwitch and PitchRockTime according to the roll and pitch attitude control schemes, wherein DelayTime is a roll attitude control rocket ignition time interval, RocktSwitch is a main rocket packet cut-off switch, and PitchRockTime is a pitch attitude control rocket ignition time;

step 4.11: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(DelayTime,RocketSwitch,PitchRocketTime)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.12: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time interval of the left and right rolling attitude control rocket is defined as 0.1 s; the parameter value of the main rocket packet disconnecting switch is defined as 0 or 1, wherein 0 represents disconnection, and 1 represents normal work; the ParachuteTime represents the time interval between the free flight and the launching of the escape parachute, and the specific numerical value can be obtained by subtracting the finish time of the cabin exiting stage from the parachute launching delay time in the simulation calculation;

when the DelayTime is 0, the left and right attitude rockets are simultaneously ignited, or the attitude rockets do not work, and the attitude correction is not needed;

when DelayTime is ParachuteTime +0.1, indicating that the ignition time of the left attitude rocket is later than the parachute shooting time, namely only the right attitude rocket is ignited, and finishing the maximum rolling attitude correction by the attitude rocket at the moment;

the precision of the ignition time of the pitching attitude rocket is defined as 0.1 s; and its ignition time must be earlier than the parachute firing time, otherwise it does not need to work;

step 4.13: the fixed state parameters are ejection speed EjectionVelocity, a yaw attitude angle RollAngle (+) and a diving angle PitchAngle, and the optimization algorithm is adopted to obtain the decision variables which enable the track height when the lifesaving parachute is full to be the highest under the state parameters meeting the constraint conditions, wherein the decision variables are DelayTime, RocktSwitch and PitchRocktTime.

4. The novel ejection seat program control method of claim 3, wherein the optimization algorithms are of a wide variety, including: a series of optimization algorithms which can be used for optimizing control parameters under constraint conditions, such as enumeration method, particle swarm optimization, genetic algorithm, fish swarm optimization, ant swarm optimization and the like, so that the optimization algorithms meet the maximization of the objective function, and any one of the optimization algorithms is selected for optimization calculation.

Technical Field

The invention relates to the technical field of aviation ejection lifesaving, in particular to a novel ejection seat program control method.

Background

The ejection seat is a main lifesaving device for providing emergency departure for pilots. Along with the continuous improvement of the performance of the fighter, the ejection state is more and more complex, and new index requirements are provided for the lifesaving performance of the ejection seat, namely ejection in a low-altitude unfavorable posture and high-speed ejection lifesaving. In order to meet the new lifesaving performance requirement, more devices (attitude rockets, thrust vectors and the like) are applied to the ejection seats, so that more controlled parameters are needed, and the program control design is more and more complex.

According to the traditional ejection seat program control, corresponding ejection modes are divided according to different ejection states (ejection speed and ejection height), namely, each state parameter is divided into a plurality of continuous subintervals in a global multidimensional space, and each interval adopts different control parameters to control a related executing mechanism, so that safe lifesaving is realized. The ejection seats with main models in service are all designed by adopting the program control method, and a typical dual-mode program control method is shown in figure 1. When the number of state parameters is small (two or less), the method is easy to implement and has high reliability, as shown in fig. 1. However, when the number of state parameters is large and the values of each parameter overlap greatly, the method for partitioning the mode is extremely complex, which mainly reflects that the state space is difficult to partition, the critical value of the parameter is difficult to determine, and the optimization degree of the control parameter in each partition region cannot be guaranteed.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present invention provides a novel ejection seat program control method.

In order to solve the technical problems, the technical scheme adopted by the invention is as follows: a novel ejection seat program control method, the flow of which is shown in fig. 2, includes the following steps:

step 1: designing a posture control scheme according to the safety lifesaving requirement of ejection of the ejection seat under the condition of low-altitude unfavorable posture;

the attitude control scheme includes: roll attitude control, pitch attitude control and roll pitch simultaneous control;

step 2: determining actuating mechanisms of different attitude control schemes, and determining control parameters and ejection state parameters under the different attitude control schemes;

the actuating mechanisms of the different attitude control schemes are as follows:

the actuating mechanism for controlling the rolling posture comprises: the left and right rolling postures control the rocket and the main rocket packet to be switched on and off;

the actuating mechanism for pitch attitude control is as follows: a pitching attitude control rocket;

the executing mechanism for roll and pitch simultaneous control comprises: a left-right rolling attitude control rocket, a main rocket packet switch and a pitching attitude control rocket.

The control parameters and the ejection state parameters under different attitude control schemes are as follows:

controlling the rolling posture: the control parameters are the ignition time interval DelayTime of the left and right rolling attitude control rocket and a main rocket packet switch RocketSwitch; the state parameters are ejection speed and roll angle;

pitch attitude control: the control parameter is the ignition time PitchRocktTime of the pitching attitude control rocket; the state parameters are ejection speed and a diving angle;

roll and pitch simultaneous control: the control parameters are the ignition time interval DelayTime of the left-right rolling attitude control rocket, a main rocket packet switch RocktSwitch and the ignition time PitchRocktTime of the pitching attitude control rocket; the state parameters are ejection speed, roll angle and dive angle.

And step 3: establishing a mathematical model and a simulation program of the whole ejection process of the ejection seat to realize simulation calculation of the trajectory posture of the whole ejection process;

and 4, step 4: fixing a group of ejection state parameters under different attitude control schemes, and calculating to obtain control parameters meeting the optimal performance according to the optimal calculation model by adopting the simulation program in the step 3;

step 4.1: judging the type of the executed control scheme, and if the control scheme is a rolling attitude control scheme, executing the step 4.2 to the step 4.5; if the pitch attitude control scheme is adopted, executing the step 4.6 to the step 4.9; if the roll and pitch are controlled simultaneously, executing the step 4.10 to the step 4.13;

step 4.2: determining decision variables as DelayTime and RocktSwitch according to a rolling attitude control scheme, wherein the DelayTime is a rolling attitude control rocket ignition time interval, and the RocktSwitch is a main rocket packet cut-off switch;

step 4.3: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(DelayTime,RocketSwitch)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.4: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time interval of the left and right rolling attitude control rocket is defined as 0.1 s; the parameter value of the main rocket packet disconnecting switch is defined as 0 or 1, wherein 0 represents disconnection, and 1 represents normal work; the ParachuteTime represents the time interval between the free flight and the launching of the escape parachute, and the specific numerical value can be obtained by subtracting the finish time of the cabin exiting stage from the parachute launching delay time in the simulation calculation;

when the DelayTime is 0, the left and right attitude rockets are simultaneously ignited, or the attitude rockets do not work, and the attitude correction is not needed;

when DelayTime is ParachuteTime +0.1, the ignition time of the left attitude rocket is later than the parachute shooting time, namely only the right attitude rocket is ignited, and the attitude rocket finishes maximum rolling attitude correction at the moment.

Step 4.5: the fixed state parameters are ejection speed EjectionVelocity and transverse turning attitude angle RollAngle (+), and the decision variables which enable the track height when the lifesaving parachute is fully opened under the state parameters meeting the constraint conditions are obtained by adopting an optimization algorithm and are the values of DelayTime and RocktSwitch.

Step 4.6: determining a decision variable as PitchRocktTime according to the pitch attitude control scheme, wherein the PitchRocktTime is the ignition time of the pitch attitude control rocket;

step 4.7: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(PitchRocketTime)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.8: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time of the pitching attitude rocket is defined as 0.1 s; and its ignition time must be earlier than the parachute firing time, otherwise it does not need to work.

Step 4.9: the fixed state parameters are ejection speed EjectionVelocity and a diving angle PitchAngle, and an optimization algorithm is adopted to obtain a decision variable which enables the track height when the lifesaving parachute is fully opened to be the value of PitchRocktTime under the state parameters meeting constraint conditions;

step 4.10: determining decision variables as DelayTime, RocktSwitch and PitchRockTime according to the roll and pitch attitude control schemes, wherein DelayTime is a roll attitude control rocket ignition time interval, RocktSwitch is a main rocket packet cut-off switch, and PitchRockTime is a pitch attitude control rocket ignition time;

step 4.11: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(DelayTime,RocketSwitch,PitchRocketTime)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.12: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time interval of the left and right rolling attitude control rocket is defined as 0.1 s; the parameter value of the main rocket packet disconnecting switch is defined as 0 or 1, wherein 0 represents disconnection, and 1 represents normal work; the ParachuteTime represents the time interval between the free flight and the launching of the escape parachute, and the specific numerical value can be obtained by subtracting the finish time of the cabin exiting stage from the parachute launching delay time in the simulation calculation;

when the DelayTime is 0, the left and right attitude rockets are simultaneously ignited, or the attitude rockets do not work, and the attitude correction is not needed;

when DelayTime is ParachuteTime +0.1, the ignition time of the left attitude rocket is later than the parachute shooting time, namely only the right attitude rocket is ignited, and the attitude rocket finishes maximum rolling attitude correction at the moment.

The precision of the ignition time of the pitching attitude rocket is defined as 0.1 s; and its ignition time must be earlier than the parachute firing time, otherwise it does not need to work.

Step 4.13: the fixed state parameters are ejection speed EjectionVelocity, a yaw attitude angle RollAngle (+) and a diving angle PitchAngle, and the optimization algorithm is adopted to obtain the decision variables which enable the track height when the lifesaving parachute is full to be the highest under the state parameters meeting the constraint conditions, wherein the decision variables are DelayTime, RocktSwitch and PitchRocktTime.

The variety of optimization algorithms is many, including: a series of optimization algorithms which can be used for optimizing control parameters under constraint conditions, such as enumeration, particle swarm optimization, genetic algorithm, fish swarm optimization, ant swarm optimization and the like, so that the objective function is maximized are selected to perform optimization calculation, and the principle of the optimization algorithm is schematically shown in fig. 3.

And 5: continuously changing the ejection state parameters so as to determine the optimal control parameters in the global range of the ejection state parameters and form point set mapping, as shown in fig. 4;

step 6: establishing a BP neural network model, taking point set mapping as a sample point set, performing training and testing on the neural network, and obtaining the BP neural network model which meets the precision requirement after testing;

and 7: and (4) writing the BP neural network model obtained in the step (6) into an ejection seat program controller, and realizing final program control.

Adopt the produced beneficial effect of above-mentioned technical scheme to lie in:

1. the novel ejection seat program control method provided by the invention can effectively avoid space division of the state parameters, is not limited by the ejection state parameters and the number of the control parameters, and ensures the optimality of the control parameters in the global range of the state space;

2. the invention establishes a mathematical model and a simulation program, and obtains a data set between the state parameters and the control parameters through calculation of the optimal calculation model, wherein the data set can accurately reflect the corresponding relation between the ejection state parameters and the optimal control parameters. In addition, due to the advantages of computer simulation, the number of samples of the data set is not limited, the cost is low, and the period is short;

3. the invention utilizes the BP neural network model to complete the nonlinear mapping process from the state parameters to the control parameters, has small error and meets the engineering application requirements;

4. the ejection state parameters can be obtained by an airplane data bus or a sensor arranged on the ejection seat, and the BP neural network model can be directly written into a singlechip of the ejection seat program controller, so that the novel ejection seat program control method is easy to realize, and the structure of the existing program controller is not required to be changed.

Drawings

FIG. 1 is a diagram illustrating an exemplary method for bimodal process control in the context of the present invention;

FIG. 2 is a flow chart of the program control method of the present invention;

FIG. 3 is a schematic diagram of the optimization algorithm of the present invention;

FIG. 4 is a diagram illustrating a mapping relationship between status parameters and control parameters according to the present invention;

FIG. 5 is a schematic view of a pair of roll attitude adjusting rockets installed in a position in accordance with an embodiment of the present invention;

FIG. 6 is a graph of a comparison of trace heights using the method of the present invention and other methods in an embodiment of the present invention;

fig. 7 is a diagram of a BP neural network structure in the embodiment of the present invention.

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

Because the existing third-generation catapult seat in active service in China does not have the function of attitude control, according to the actual situation, considering engineering application and implementation, the simulation model takes an HTY-8 type seat as a prototype.

Step 1: the present embodiment takes the control of the roll attitude as an example. The control of the rolling posture is realized by installing a pair of rolling posture adjusting rockets at the rear part of the headrest umbrella box, and the schematic diagram of the installation position is shown in figure 5.

In addition, the model also provides a main rocket packet cut-off switch, and corresponding control measures are as follows:

1) when a forward rolling angle exists, the right rolling attitude rocket is ignited, and after a certain delay time, the left rolling attitude rocket is ignited, so that the attitude correction is avoided from passing through the head; on the contrary, when a negative roll angle exists, the left attitude rocket is ignited firstly;

2) when the rolling attitude angle is too large, so that the attitude correction cannot be carried out by depending on the attitude rocket, the main rocket packet is cut off, and the loss of the safety lifesaving altitude caused by the thrust of the main rocket packet is avoided.

Step 2: as can be seen from the above control measures, there are two control parameters in this embodiment, which are the ignition time interval DelayTime of the left and right rolling attitude control rocket and the cut-off switch rockswitch of the main rocket packet. Through the analysis of the seat, the ignition time interval of the left-and-right posture rocket is the main control parameter of the posture control method, and the correction effect on the posture and the track of the human seat can be directly influenced. The determination of the state parameter critical value corresponding to whether the main rocket packet is cut off is also a difficult point of the traditional multi-mode control law.

As can be seen from the foregoing description, the present embodiment only considers unfavorable attitude launch conditions with roll angle, and therefore the condition parameters include launch velocity and roll attitude angle. Although the ejection height is also one of the state parameters when the ejection is started, the attitude control mainly aims at improving the lifesaving performance during low-altitude ejection, and when the ejection height is higher, the lifesaving performance cannot be greatly influenced even if the trajectory attitude control is not adopted, so the ejection height is not taken as one of the state parameters in the optimized design. Due to the axisymmetric nature of aircraft and human chair systems, only the forward roll angle needs to be calculated, from which the state parameter space can be expressed as,

wherein EjectionVelocity is the ejection speed, and RollAngle (+) is the transverse turning attitude angle;

and step 3: establishing a mathematical model and a simulation program of the whole ejection process of the ejection seat to realize simulation calculation of the trajectory posture of the whole ejection process;

and 4, step 4: fixing a group of ejection state parameters under different attitude control schemes, and calculating to obtain control parameters meeting the optimal performance according to the optimal calculation model by adopting the simulation program in the step 3;

step 4.1: judging the type of the executed control scheme, and if the control scheme is a rolling attitude control scheme, executing the step 4.2 to the step 4.5; if the pitch attitude control scheme is adopted, executing the step 4.6 to the step 4.9; if the roll and pitch are controlled simultaneously, executing the step 4.10 to the step 4.13; the implementation of the present embodiment is roll attitude control, so step 4.2 to step 4.5 are implemented,

step 4.2: determining decision variables as DelayTime and RocktSwitch according to a rolling attitude control scheme, wherein the DelayTime is a rolling attitude control rocket ignition time interval, and the RocktSwitch is a main rocket packet cut-off switch;

step 4.3: taking the highest track height when the lifesaving parachute is fully opened as an objective function of the optimal calculation model:

max InflationHeight(DelayTime,RocketSwitch)

wherein, the function InflationHeight represents the track height of the parachute system when the lifesaving parachute is full;

step 4.4: and (3) considering the value range and the control precision of the decision variables, and constraining the decision variables:

the precision of the ignition time interval of the left and right rolling attitude control rocket is defined as 0.1 s; the parameter value of the main rocket packet disconnecting switch is defined as 0 or 1, wherein 0 represents disconnection, and 1 represents normal work; the ParachuteTime represents the time interval between the free flight and the launching of the escape parachute, and the specific numerical value can be obtained by subtracting the finish time of the cabin exiting stage from the parachute launching delay time in the simulation calculation;

when the DelayTime is 0, the left and right attitude rockets are simultaneously ignited, or the attitude rockets do not work, and the attitude correction is not needed;

when DelayTime is ParachuteTime +0.1, the ignition time of the left attitude rocket is later than the parachute shooting time, namely only the right attitude rocket is ignited, and the attitude rocket finishes maximum rolling attitude correction at the moment.

Step 4.5: the fixed state parameters are ejection speed EjectionVelocity and transverse turning attitude angle RollAngle (+), and the decision variables which enable the track height when the lifesaving parachute is fully opened under the state parameters meeting the constraint conditions are obtained by adopting an optimization algorithm and are the values of DelayTime and RocktSwitch.

The variety of optimization algorithms is many, including: a series of optimization algorithms which can be used for optimizing control parameters under constraint conditions, such as enumeration method, particle swarm optimization, genetic algorithm, fish swarm optimization, ant swarm optimization and the like, so that the optimization algorithms meet the maximization of the objective function, and any one of the optimization algorithms is selected for optimization calculation.

And 5: continuously changing the ejection state parameters so as to determine the optimal control parameters in the overall range of the ejection state parameters and form point set mapping;

in order to ensure the global optimality of the control parameters and meet the training requirements of the neural network, 164 groups of optimal control parameters corresponding to the ejection state parameters are calculated in total. In this embodiment, the state parameters: the ejection speed is 0km/h, the roll angle is 45 degrees for example, the optimal control calculation result is respectively compared with the multi-mode control rule simulation result and the calculation result without attitude control, the comparison result is shown in figure 6, and the effectiveness of the optimal control parameter obtained by the method is verified. In addition, the simulation result is compared with the performance of the HTY-8 type seat, the lowest safe lifesaving height is compared according to the launching condition required by the state army, and the improvement of the lifesaving performance of the launching lifesaving system by adopting the rolling attitude control rule obtained by the text is quantitatively verified.

Step 6: establishing a BP neural network model, taking point set mapping as a sample point set, performing training and testing on the neural network, and obtaining the BP neural network model which meets the precision requirement after testing;

in order to make the neural network have stronger adaptability and fault tolerance, the present embodiment effectively reduces the network scale, and reduces the calculation amount when the final algorithm is implemented, so the above problem is decomposed into two neural networks to be implemented respectively. The network 1 is responsible for completing the mode recognition of a main rocket packet cut-off switch RocketSwitch, and the network 2 is responsible for completing the nonlinear mapping of a left-right rolling attitude control rocket ignition time interval DelayTime, which is shown as the following formula:

when the value obtained by the judgment of the network 1 is 0, namely the main rocket packet is cut off, the processing calculation of the network 2 is not needed.

The network 1 and the network 2 adopt the same neural network structure, and the number of network layers is 2. Layer 1 is the input layer neurons, with a neuron number of 50. Layer 2 is the output layer neurons, with a neuron number of 1. Since the number of network layers is only 2, the input layer is also called an intermediate layer or hidden layer. The input vector is a two-dimensional vector and respectively corresponds to the ejection speed and the rolling attitude angle. The output vector is 1 dimension, the network 1 outputs the state of a main rocket packet cut-off switch, and the network 2 outputs the left and right rolling postures to control the ignition time interval of the rocket. The structure of the neural network is shown in fig. 7, wherein p is an input vector; IW is an input layer weight matrix; LW is a network layer weight matrix; b is a threshold vector; n is a weighted sum vector; a is the output vector.

After the training is completed, the fit of the neural network 1 to the sample point set is 100%. Namely, any input state vector in the sample point set can be processed by the neural network to obtain the rocket packet cutting state which is the same as the optimal control parameter.

The neural network 2 has better goodness of fit for the sample point set, the error value for more than 90% of the sample points is less than 0.1s, the error values for all the sample points are less than 0.2s, and the result completely meets the engineering requirement. In order to further improve the matching degree of the state parameters to the ignition time interval of the left-right rolling attitude control rocket, the structure of the neural network needs to be improved, and the accuracy is improved by increasing the number of network layers and the number of neurons.

And (3) randomly selecting 12 groups of non-sample point ejection states with low, medium and high ejection speeds and different roll angles to test the neural network. The output results of the neural network were compared to the values of the theoretically optimal control parameters (found by the optimal calculation model alone), and the results are shown in table 1:

TABLE 1 neural network test results

As can be seen from table 1, the output result of the neural network and the value of the theoretically optimal control parameter have an error of 0.1s in the ignition time interval of the roll attitude control rocket only in group 10.

And 7: and (4) writing the BP neural network model obtained in the step (6) into an ejection seat program controller, and realizing final program control.