Free piston engine stability control method based on machine learning

Abstract

The invention relates to the technical field of engine stability control, and discloses a free piston engine stability control method based on machine learning, and the method comprises the following steps: constructing a hydraulic free piston engine simulation model which comprises a thermodynamic model, a hydraulic model and a dynamic model; performing simulation analysis on the hydraulic free piston engine by using the hydraulic free piston engine simulation model to obtain corresponding relations between the output power of the hydraulic free piston engine and the return pressure, the output pressure and the combustion duration, and taking the corresponding relations as input characteristics and the oil injection quantity as an output parameter; a fuel injection quantity prediction model based on a BP neural network is constructed; inputting the return pressure, the output pressure, the combustion duration and the fuel injection quantity obtained based on the fuel injection quantity prediction model into a hydraulic free piston engine simulation model; according to the requirement of the target power, it can be guaranteed that the hydraulic free piston engine conducts high-stability working condition switching.

Description

Technical Field

[0001] The present invention relates to the technical field of engine stability control, and more specifically, to a stability control method for a free piston engine based on machine learning. Background Art

[0002] As a relatively popular new type of power mechanical device in recent years, the hydraulic free piston engine cancels the crankshaft connecting rod mechanism compared with the traditional engine. Therefore, the movement trajectory of the piston is a linear movement in the working state, thereby avoiding the generation of piston side forces, reducing friction losses and greatly shortening the energy transfer chain, and improving the energy conversion efficiency of the engine.

[0003] On the other hand, the structural simplification of the hydraulic free piston engine further reduces the overall weight, increases the power density and reduces the manufacturing cost. In addition, the above characteristics of the hydraulic free piston engine determine that the movement trajectory of its piston is determined by the resultant force. Therefore, it has the advantage of a controllable and flexible compression ratio; since the crankshaft mechanism is cancelled, the mechanical constraint mechanism that can ensure the stable operation of the piston also disappears. Therefore, piston movement control is the key to ensuring the stable operation of the hydraulic free piston engine. At present, the lack of reliable control leading to the inability of the hydraulic free piston engine to operate continuously and stably is the main factor restricting its engineering application. Summary of the Invention

[0004] The object of the present invention is to provide a stability control method for a free piston engine based on machine learning, which identifies the parameters that have a significant impact on the output power through simulation analysis and inputs them into the simulation model of the hydraulic free piston engine to adjust the parameters to achieve stable control of the target power and operating condition switching commands.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A stability control method for a free piston engine based on machine learning, comprising the following steps:

[0007] S1. Construct a simulation model of a hydraulic free piston engine, including: a thermodynamic model, a hydraulic model and a kinetic model;

[0008] S2. Use the simulation model of the hydraulic free piston engine to perform simulation analysis on the hydraulic free piston engine, obtain the corresponding relationships between the output power of the hydraulic free piston engine and the return pressure, the output pressure and the combustion duration respectively, and draw them into a MAP diagram;

[0009] S3. Construct an injection quantity prediction model based on a BP neural network with the output power, the return pressure, the output pressure and the combustion duration as input features and the injection quantity as an output parameter;

[0010] S4. Take the response pressure, output pressure, combustion duration, and fuel injection quantity obtained from the fuel injection quantity prediction model as input parameters, and input them into the hydraulic free piston engine simulation model. Through parameter adjustment, realize the stable control of the target power and operating condition switching command.

[0011] Further, in the above S1, the construction method of the thermodynamic model is specifically as follows:

[0012] Based on the first law of thermodynamics and the variation law of the cylinder volume under the action of the unbalanced pressure, obtain the corresponding relationship between the in-cylinder pressure of the engine and the change in the heat of the in-cylinder working medium.

[0013] Determine the relationship between the change in the heat of the in-cylinder working medium, the heat generated by fuel combustion, and the heat loss of heat exchange between the working medium in the combustion chamber and the environment.

[0014] Further, the obtaining of the corresponding relationship between the in-cylinder pressure of the engine and the change in the heat of the in-cylinder working medium based on the first law of thermodynamics and the variation law of the cylinder volume under the action of the unbalanced pressure is specifically as follows:

[0015] Based on the first law of thermodynamics and the variation law of the cylinder volume under the action of the unbalanced pressure, obtain:

[0016] δQ = dU + δW (1)

[0017] P C V C = m a RT (2)

[0018] In the formula, δQ represents the heat absorbed by the thermodynamic system from the outside world, dU is the increased internal energy of the system, and δW represents the work done by the system to the outside; P C is the gas pressure acting on the piston, V C represents the cylinder volume under the action of the unbalanced pressure, m a represents the mass of the ideal gas, R represents the air gas constant, and T represents the in-cylinder gas temperature;

[0019] For further derivation and solution of each parameter in formula (1), obtain:

[0020] δW = P C × dC C (3)

[0021] dU = m a C V dT (4)

[0022]

[0023] In the above formula, C V represents the specific heat capacity at constant volume;

[0024] By simultaneously calculating the above formulas (1)-(5), the corresponding relationship between the in-cylinder pressure and the change in the thermal energy of the in-cylinder working fluid is obtained:

[0025]

[0026] The relationship for determining the change in the heat of the in-cylinder working fluid, the heat generated by fuel combustion, and the heat loss due to heat exchange between the working fluid in the combustion chamber and the environment is specifically as follows:

[0027] The change in the heat of the in-cylinder working fluid is affected by two parts: the heat generated by fuel combustion and the heat loss due to heat exchange between the working fluid in the fuel chamber and the environment. Among them, the heat generated by combustion is solved by the Weber function, and the heat loss due to heat exchange between the working fluid in the combustion chamber and the environment is related to the working fluid temperature and the environment temperature. The expression is as follows:

[0028] Q = Q C + Q W (7)

[0029]

[0030] In the formula, Q represents the change in the heat of the in-cylinder working fluid, Q C represents the heat generated by fuel combustion, Q W represents the heat loss due to heat exchange between the working fluid in the combustion chamber and the environment, H u represents the low calorific value of diesel, g f is the cyclic fuel supply, η u represents the combustion efficiency, m c represents the combustion quality index, t c represents the combustion duration, α is the instantaneous heat transfer coefficient, T and T amb represent the working fluid temperature and the environment temperature respectively.

[0031] Furthermore, in the S1, the method for constructing the hydraulic model is specifically as follows:

[0032] At three pump chamber positions of the hydraulic free piston engine simulation model, an oil inlet hole and an oil outlet hole are respectively set; and by using the fluid continuity equation, the actual oil intake, the actual oil output, and the leakage loss flow rate of the engine are obtained.

[0033] Furthermore, the setting of an oil inlet hole and an oil outlet hole respectively at three pump chamber positions of the hydraulic free piston engine simulation model; and obtaining the actual oil intake, the actual oil output, and the leakage loss flow rate of the engine by using the fluid continuity equation is specifically as follows:

[0034] At three pump chamber positions of the hydraulic free piston engine, oil holes are respectively set to generate pressure, and the fluid equations of the three pump chambers in the working state are derived by using the fluid continuity equation, specifically as follows:

[0035]

[0036] In the above formula, P h represents the output pressure of the pump chamber, K represents the elastic modulus, Q in and Q out respectively represent the actual oil inlet volume of the oil inlet hole and the actual oil outlet volume of the oil outlet hole, A p represents the piston contact area, and △q represents the leakage loss flow rate;

[0037] Expanding the flow parameters in formula (10), we get:

[0038] △q = P h ×λ (11)

[0039]

[0040] In the above formula, λ represents the leakage coefficient, c0 represents the flow coefficient, A0 represents the flow-through area, ρ represents the density of the hydraulic oil, and P c represents the pressure constant.

[0041] Furthermore, in S1, the construction method of the kinetic model is specifically as follows:

[0042] Based on Newton's second law, analyze the kinematic law of the piston of the hydraulic free piston engine under specific working conditions to obtain the change laws of the piston's displacement, velocity, and acceleration.

[0043] Furthermore, based on Newton's second law, analyze the kinematic law of the piston of the hydraulic free piston engine under specific working conditions to obtain the change laws of the piston's displacement, velocity, and acceleration, specifically as follows:

[0044] The power of the hydraulic free piston engine includes: in-cylinder pressure, hydraulic pressure, and frictional force; among them, the in-cylinder pressure is generated on the left side of the hydraulic free piston engine, and the hydraulic pressure is generated on the right side of the hydraulic free piston engine;

[0045] Based on Newton's second law, the motion mode of the hydraulic free piston is as follows:

[0046]

[0047] In the formula, x, and are respectively the displacement, velocity, and acceleration of the piston, M represents the piston mass; P represents the in-cylinder pressure, P1 represents the return pressure, P2 represents the periodically changing pressure, P3 represents the output pressure, S, S1, S2, S3 respectively represent the contact areas of each piston, indicates that the frictional force received by the piston is opposite to the direction of motion, Ff Represents the reciprocating frictional force.

[0048] Further, in step S3, taking the output power, the return pressure, the output pressure, and the combustion duration as input features and the fuel injection quantity as the output parameter, a fuel injection quantity prediction model based on a BP neural network is constructed. Specifically:

[0049] Based on the hydraulic free piston engine simulation model, a data set constructed from the output power, the return pressure, the output pressure, the combustion duration, and the fuel injection quantity is obtained, and the data set is divided into a training set and a test set according to a ratio of 500:100.

[0050] The training set is used to train the network parameters of the fuel injection quantity prediction model.

[0051] The test set is used to evaluate the generalization ability of the fuel injection quantity prediction model.

[0052] Further, in step S4, taking the return pressure, the output pressure, the combustion duration, and the fuel injection quantity obtained from the fuel injection quantity prediction model as input parameters, inputting them into the hydraulic free piston engine simulation model, and through parameter adjustment, realizing the stable control of the target power and the working condition switching instruction. Specifically:

[0053] Judging the optimal switching time according to the movement displacement and movement speed of the piston.

[0054] The piston motion state of the hydraulic free piston engine is collected in real time through the control end. After receiving the demand power change instruction, the optimal switching point that reaches the switching condition for the first time after this moment is obtained according to the piston motion state, so as to realize the smooth switching of the hydraulic free piston engine.

[0055] Further, after receiving the demand power change instruction, the optimal switching point that reaches the switching condition for the first time after this moment is obtained according to the piston motion state, so as to realize the smooth switching of the hydraulic free piston engine. The determination of the demand power is specifically as follows:

[0056] The demand power needs to compare the difference between the real power at this moment and the demand power at the previous moment. When the difference between the two is less than the constraint range, the target power at this moment is directly determined; when the difference between the two is greater than the constraint range, the demand power at this moment is the sum of the demand power at the previous moment and the constraint maximum value, and the unmet small part of the power is continuously compared in the next cycle. Finally, the hydraulic free piston engine completes the output of the demand power ability within several cycles.

[0057] According to the specific embodiments provided by the present invention, the following technical effects of the present invention are disclosed:

[0058] By analyzing the simulation model of the hydraulic free piston engine, the present invention obtains four parameters that have the greatest influence on the output power, namely: fuel injection quantity, return pressure, output pressure, and combustion duration. The corresponding relationships between the output power and the return pressure, output pressure, and combustion duration are plotted as MAPs, which can intuitively show the influence of each parameter on the output power. Combining the fuel injection quantity predicted by the fuel injection quantity prediction model, it is possible to select appropriate control parameters according to the power demand under steady-state conditions to achieve the energy output of the hydraulic free piston engine. Under dynamic operating conditions, the present invention can also apply a stability control strategy according to the power demand to ensure high responsiveness and high stability condition switching of the hydraulic free piston engine. Brief Description of the Drawings

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0060] The following further illustrates the method for controlling the stability of a free piston engine based on machine learning according to the present invention with reference to the drawings;

[0061] Figure 1 It is a schematic diagram of the overall process in the method for controlling the stability of a free piston engine based on machine learning provided by the present invention;

[0062] Figure 2 It is a simulation model of a hydraulic free piston engine in the method for controlling the stability of a free piston engine based on machine learning provided by the present invention;

[0063] Figure 3 It is a schematic diagram of the thermodynamic process in the method for controlling the stability of a free piston engine based on machine learning provided by the present invention;

[0064] Figure 4 It is a schematic diagram of the hydraulic system in the method for controlling the stability of a free piston engine based on machine learning provided by the present invention;

[0065] Figure 5 It is a schematic diagram of the dynamic process in the method for controlling the stability of a free piston engine based on machine learning provided by the present invention;

[0066] Figure 6It is the verification curve graph of the engine piston displacement and cylinder pressure model in the stability control method of the free piston engine based on machine learning provided by the present invention; among which (a) is the comparison graph of the dynamic operation curves of the predicted value and the true value; (b) is the comparison graph of the dynamic operation curves of the fuel injection quantity and the sampling points; (c) is the comparison graph of the dynamic operation curves of the error value and the sampling points; (d) is the comparison graph of the dynamic operation curves of the proportion of the prediction error and the prediction error.

[0067] Figure 7 It is the schematic diagram of the stability control logic in the stability control method of the free piston engine based on machine learning provided by the present invention.

[0068] Figure 8 It is the schematic diagram of the prediction step of the optimal switching moment in the stability control method of the free piston engine based on machine learning provided by the present invention.

[0069] Figure 9 It is the flow chart of the stability control strategy for the dynamic process of the hydraulic free piston engine in the stability control method of the free piston engine based on machine learning provided by the present invention.

[0070] Figure 10 It is the graph of the variation law of the real-time dynamic switching parameters of the hydraulic free piston engine in the stability control method of the free piston engine based on machine learning provided by the present invention; among which (a) is the graph of the displacement variation law of the piston in a good state; (b) is the graph of the speed variation law of the piston in a good state; (c) is the graph of the acceleration variation law of the piston in a good state; (d) is the graph of the cylinder pressure variation law of the piston in a good state, (e) is the graph of the compression ratio variation law of the piston; (f) is the real-time power of the dynamic switching of the engine working conditions. Specific Embodiments

[0071] The following combines the drawings and embodiments to further describe in detail the specific embodiments of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0072] To better understand the purpose, structure and function of the present invention, the following further describes the present invention in detail with reference to the drawings.

[0073] As Figure 1 shown, the present invention provides a stability control method for a free piston engine based on machine learning, including the following steps:

[0074] As Figure 2 shown, S1. Establish a simulation model of the hydraulic free piston engine, including: a thermodynamic model, a hydraulic model and a kinetic model;

[0075] It should be noted that in the thermodynamic model, the combustion process simulation is carried out based on the piston displacement and piston speed obtained from the simulation calculation of the previous cycle of the hydraulic free piston engine. This part of the model outputs the in-cylinder pressure. The hydraulic model solves the pressure based on the displacement speed obtained from the simulation calculation of the previous cycle of the hydraulic free piston engine, and respectively obtains the output pressure, return pressure, and reciprocating pressure. Based on the above-obtained output pressure, return pressure, and reciprocating pressure, they are coupled into the dynamic model, and the piston motion state of the hydraulic free piston engine is obtained by applying Newton's second law, so as to further calculate the output power of the engine. This simulation model of the hydraulic free piston engine can successfully simulate and calculate the variation laws of parameters such as the acceleration, speed, displacement, motion frequency, and in-cylinder pressure of the piston.

[0076] And the piston displacement data and in-cylinder pressure data of the hydraulic free piston engine model are collected. The dynamic operation curves of the above two parameters within one cycle during the engine operation process are mainly compared and verified. The verification results are as Figure 6 shown; through Figure 6 (a) in Figure 6 (b) in, it can be seen that the dynamic operation curve comparison of the piston displacement and in-cylinder pressure in the hydraulic free piston engine model with the experimental results of the physical prototype is respectively shown. The overall trends of the dynamic operation curves of the above two parameters are consistent with the test data. The difference part is mainly in the early stage of this cycle operation. The main reason is that the constant hydraulic pressure controls the movement of the piston, thereby affecting the piston compression speed, and the phenomenon in the in-cylinder pressure is more obvious in the physical prototype. However, the peaks and operation periods of the dynamic operation curves of the above two parameters are the same as the test data. Generally speaking, the piston displacement and in-cylinder pressure curves of the simulation data basically overlap with the experimental data. Therefore, the above comparison results prove that the built hydraulic free piston engine can replace the physical prototype to support further research work.

[0077] S2. According to the simulation model of the hydraulic free piston engine, the engine is simulated and analyzed to obtain the corresponding relationships between the output power of the engine and the return pressure, output pressure, and combustion duration respectively, and they are plotted into a MAP diagram;

[0078] It should be noted that: Since the control problem of the hydraulic free piston engine in the present invention is mainly divided into two parts. The first part is how to select appropriate control parameters according to the power demand under steady-state conditions so as to achieve the energy output of the engine. The second part is that under the actual dynamic conditions where the power demand changes continuously, control strategies are applied to ensure high responsiveness and high stability of the hydraulic free piston engine during working condition switching. Therefore, by comparing the correlation between the input parameters and output parameters of the hydraulic free piston engine model, four parameters that have a greater impact on the engine output power are finally determined, namely: fuel injection quantity, return pressure, output pressure, and combustion duration. In order to obtain control parameters more accurately through machine learning prediction algorithms; as Figure 7 shown, the present invention determines the corresponding relationships between the output power and the return pressure, output pressure, and combustion duration respectively, and plots them into MAP diagrams; and inputs the output power, return pressure, output pressure, combustion duration, and the fuel injection quantity predicted by the fuel injection quantity prediction model into the hydraulic free piston engine model, and controls the stability of the dynamic process of the hydraulic free piston engine by changing these four parameters.

[0079] S3. Use the output power, return pressure, output pressure, and combustion duration as input features to construct a fuel injection quantity prediction model based on a BP neural network;

[0080] S4. Input the return pressure, output pressure, combustion duration, and the fuel injection quantity obtained based on the fuel injection quantity prediction model as input parameters into the hydraulic free piston engine simulation model, and achieve stable control of the target power and working condition switching command through parameter adjustment.

[0081] It should be noted that: After obtaining the return pressure, output pressure, combustion duration, and fuel injection quantity, according to the change in the demand of the target power, the optimal switching point is found in real time to ensure high responsiveness, high stability, and high accuracy of the hydraulic free piston engine during switching. Specifically: First, judge the working state of the hydraulic free piston engine and predict the optimal switching time for dynamic switching. The main judgment condition for predicting the optimal switching time is based on the movement displacement and movement speed of the piston. The control end collects the piston running state of the hydraulic free piston engine in real time. After receiving the instruction of the change in demand power, the optimal switching point, that is, the optimal moment point before the piston reaches the top dead center, which is the first to reach the switching condition after this moment, is obtained according to the piston running state. The schematic diagram of the steps of the judgment process is as Figure 8 shown.

[0082] This process simultaneously receives the demand power command and monitors the piston state. The change in demand power is the starting point for operating condition switching. Based on the change in demand power, combined with the MAP diagram and the machine learning prediction algorithm, the corresponding control parameters can be obtained. The monitoring of the piston's motion state includes real-time information feedback on the piston's motion displacement and velocity. Based on the piston's motion displacement and velocity, the combustion moment in the thermodynamic process can be obtained. Subsequently, the time characteristics of the demand power change point are extracted, and the optimal switching moment is determined according to the piston's motion displacement and velocity. The determination of this moment ensures the smooth switching of the hydraulic free piston engine, and at the same time, this moment is the point closest to the demand power change moment, meeting the high responsiveness characteristics required by the control strategy. After the above series of processes are completed, the engine's smooth operating condition switching is finally achieved according to the predicted control parameters and switching moment.

[0083] The thermodynamic model in S1 is specifically as follows:

[0084] Based on the first law of thermodynamics and the law of change in cylinder volume under unbalanced pressure, the corresponding relationship between the in-cylinder pressure and the change in heat of the in-cylinder working medium of the engine is obtained;

[0085] Analyze the corresponding relationship between the in-cylinder pressure and the change in heat of the in-cylinder working medium to determine the main factors affecting the heat exchange loss between the working medium in the combustion chamber and the environment.

[0086] The hydraulic model in S1 is specifically as follows:

[0087] At the three pump chamber positions of the hydraulic free piston engine simulation model, an oil inlet hole and an oil outlet hole are respectively set; and using the fluid continuity equation, the actual oil inlet volume, actual oil outlet volume, and leakage loss flow rate of the engine are obtained.

[0088] The dynamic model in S1 is specifically as follows:

[0089] Based on Newton's second law, analyze the kinematic law of the piston of the hydraulic free piston engine under specific working conditions to obtain the change laws of the piston's displacement, velocity, and acceleration.

[0090] As Figure 3 shown, based on the first law of thermodynamics and the law of change in cylinder volume under unbalanced pressure, the corresponding relationship between the in-cylinder pressure and the change in heat of the in-cylinder working medium of the engine is obtained;

[0091] Analyze the corresponding relationship between the in-cylinder pressure and the change in heat of the in-cylinder working medium to determine the main factors affecting the heat exchange loss between the working medium in the combustion chamber and the environment, specifically as follows:

[0092] Based on the first law of thermodynamics and the law of change in cylinder volume under unbalanced pressure, it is obtained that:

[0093] δQ=dU+δW (1)

[0094] P C V C =m a RT (2)

[0095] In the formula, δQ represents the heat absorbed by the thermodynamic system from the outside, dU is the increased internal energy of the system, and δW represents the work done by the system to the outside; P C is the gas pressure acting on the piston, V C Represents the cylinder volume driven by unbalanced pressure, m a represents the mass of the ideal gas, R represents the air gas constant, and T represents the temperature of the gas in the cylinder;

[0096] Further deduction and solution of various parameters in formula (1) yield:

[0097] δW=P C ×dV C (3)

[0098] dU=m a C V dT (4)

[0099]

[0100] In the above formula, C V represents the specific heat at constant volume;

[0101] The above formulas (1)-(5) are calculated together to obtain the corresponding relationship between the cylinder pressure and the change of the thermal energy of the working fluid in the cylinder:

[0102]

[0103] The change of working fluid heat is mainly affected by the heat generated by fuel combustion and the heat lost by heat exchange between the working fluid and the environment in the fuel cavity. The heat generated by combustion is solved by the Weibull function, while the heat lost by heat exchange between the working fluid and the environment in the combustion cavity is related to the working fluid temperature and the ambient temperature. The expression is as follows:

[0104] Q=Q C +Q W (7)

[0105]

[0106] In the formula, Q C Indicates the heat generated by fuel combustion, Q W It represents the heat loss caused by the heat exchange between the working fluid in the combustion chamber and the environment, H u Represents diesel lower calorific value, g fis the cyclic fuel supply, η u represents the combustion efficiency, m c represents the combustion quality index, t c represents the combustion duration, α is the instantaneous heat transfer coefficient, T and T amb represent the working fluid temperature and the ambient temperature respectively.

[0107] It should be noted that: as Figure 3 shown, four assumptions are satisfied when establishing the thermodynamic model of the hydraulic free piston engine:

[0108] ① The state of the working fluid in the cylinder is uniform, that is, the pressure, temperature, and concentration at each point in the cylinder are equal at the same instant;

[0109] ② The uniformity assumption is adopted. The thermodynamic states and chemical compositions at each point in the system are the same, and the parameters do not change with the spatial coordinates. Therefore, the effects of factors such as the combustion chamber shape, in-cylinder air flow organization, and flame propagation on the actual performance are ignored, and it is assumed that the mixture always follows the ideal gas state equation;

[0110] ③ The actual replacement and leakage losses of the working fluid are not considered. The total mass of the working fluid remains unchanged, and the thermodynamic cycle is carried out with a fixed amount of working fluid. The intake and exhaust flow losses and their effects are ignored;

[0111] ④ The actual combustion process of the fuel is replaced by a surrogate combustion heat release law.

[0112] Based on the above assumptions, in order to study the variation law of the pressure of the working fluid in the cylinder, the boundary formed by the cylinder head, cylinder wall, and piston top is taken as the control surface, and the space volume it encloses is taken as the control volume.

[0113] An oil inlet hole and an oil outlet hole are respectively arranged at three pump chamber positions of the hydraulic free piston engine simulation model; and by using the fluid continuity equation, the actual fuel intake, actual fuel output, and leakage loss flow of the engine are obtained, specifically:

[0114] As Figure 4 shown, oil holes are respectively arranged at three pump chamber positions of the hydraulic free piston engine to generate pressure, and the fluid equations of the three pump chambers in the working state are derived by using the fluid continuity equation, specifically:

[0115]

[0116] In the above formula, P h represents the output pressure of the pump chamber, K represents the elastic modulus, Q in and Q out respectively represent the actual fuel intake of the oil inlet hole and the actual fuel output of the oil outlet hole, A p represents the piston contact area, and △q represents the leakage loss flow;

[0117] Expand the flow parameter in formula (10) to obtain:

[0118] △q = P h ×λ (11)

[0119]

[0120] In the above formula, λ represents the leakage coefficient, c0 represents the flow coefficient, A0 represents the flow-through area, ρ represents the density of hydraulic oil, and P c represents the pressure constant.

[0121] Based on Newton's second law, analyze the kinematic law of the piston of the hydraulic free piston engine under specific working conditions to obtain the changes in the displacement, velocity, and acceleration of the piston, specifically:

[0122] As Figure 5 shown, the power of the hydraulic free piston engine includes: in-cylinder pressure, hydraulic pressure, and frictional force; among them, the in-cylinder pressure is generated on the left side of the hydraulic free piston engine, and the hydraulic pressure is generated on the right side of the hydraulic free piston engine to make the piston continuously reciprocate;

[0123] Based on Newton's second law, the motion mode of the hydraulic free piston is as follows:

[0124]

[0125] In the formula, x, and are the displacement, velocity, and acceleration of the piston respectively, M represents the piston mass; P represents the in-cylinder pressure, P1 represents the return pressure, P2 represents the periodically varying pressure, P3 represents the output pressure, and S, S1, S2, S3 respectively represent the contact areas of each piston, indicates that the frictional force received by the piston is opposite to the direction of motion, and F f represents the reciprocating frictional force.

[0126] In step S3, the output power, return pressure, output pressure, and combustion duration are used as input parameters to establish an injection quantity prediction model based on the BP neural network, specifically:

[0127] Divide the output power, return pressure, output pressure, and combustion duration into a training set and a test set according to a ratio of 500:100;

[0128] The training set is used to train the network parameters of the injection quantity prediction model;

[0129] The test set is used to evaluate the generalization ability of the injection quantity prediction model.

[0130] As Figure 6As shown, the correlation coefficient (R2) of the calculated complete set in the present invention reaches 0.9999, and the root mean square error (RMSE) is 0.0157; from Figure 6 the relationship diagram between the sampling points and the fuel injection quantity and the relationship diagram between the sampling points and the error value in Figure 6 , it can be seen that the error comparison between the true value and the predicted value of 100 sampling data points is presented, and the absolute values of the errors of the above two are both within the range of 0.04; observing Figure 7 the proportion of the prediction error in Figure 7 , it can be known that the overall error of the prediction result of the BP neural network basically follows a normal distribution. Based on the above analysis, the BP neural network prediction model established in the present invention has a good prediction function and can accurately predict the fuel injection quantity of the engine.

[0131] In step S4, the return pressure, the output pressure, the combustion duration, and the fuel injection quantity predicted by the fuel injection quantity prediction model are input into the hydraulic free piston engine simulation model to perform stability control on the dynamic process of the hydraulic free piston engine. Specifically:

[0132] Judge the optimal switching time according to the movement displacement and movement speed of the piston;

[0133] The piston motion state of the hydraulic free piston engine is collected in real time through the control end. After receiving the target power change command, the optimal switching point that first reaches the switching condition after this moment is obtained according to the piston motion state, so as to realize the smooth switching of the hydraulic free piston engine.

[0134] After receiving the demand power change command, the optimal switching point that first reaches the switching condition after this moment is obtained according to the piston motion state, so as to realize the smooth switching of the hydraulic free piston engine. Among them, the determination of the demand power is specifically:

[0135] The determination of the target power needs to compare the difference between the demand power at this moment and the demand power at the previous moment. When the difference between the two is less than the constraint range, the demand power value at this moment is directly determined; when the difference between the two is greater than the constraint range, the demand power at this moment is the sum of the demand power at the previous moment and the constraint maximum value, and the unmet small part of the power is continuously compared in the next cycle. Finally, the hydraulic free piston engine completes the output of the demand power capacity within several cycles.

[0136] It should be noted that: in addition to predicting the control strategy at the switching moment of working conditions, a certain amount of constraint is also required for the demand power itself. This is mainly because the system of the hydraulic free piston engine is relatively complex and the coupling degree of each component is relatively high. Therefore, a large-span change of working conditions cannot be achieved during dynamic switching. This will cause the working state of the hydraulic free piston engine to misfire and collapse when the demand power jumps from a small value to a high value. Therefore, when the above situation occurs, the control strategy will first select an intermediate target power value for transition and then jump from this power value to the high power point, thus ensuring the smooth switching of the hydraulic free piston engine. Summarize the above control strategy, and finally obtain the schematic diagram of the dynamic switching strategy of the hydraulic free piston engine, as shown in Figure 9 shown below.

[0137] This is divided into two cases. ① When the difference between the two is less than the constraint range, the demand power can be directly determined, as shown in the "No" arrow part in the upper left of Figure 9 ; ② Figure 9 In the "Yes" part in the upper left of , when it is greater than the constraint range, the current moment power and the maximum value of the constraint range will be added to obtain the target power.

[0138] Figure 9 The upper part of shows the flow chart of the stability control strategy for the dynamic process of the hydraulic free piston engine. Based on the above-mentioned content, the control process is mainly divided into two parts: demand power determination and switching moment judgment, and the two are carried out simultaneously in real-time control. Among them, for demand power determination, it is necessary to compare the difference between the demand power at this moment and the demand power at the previous moment. When the difference between the two is less than the constraint range, the demand power value at this moment is directly determined; when the difference is greater than the constraint range, the demand power at this moment is the sum of the demand power at the previous moment and the maximum constraint value, and the unmet part of the power is continuously compared in the next cycle. Finally, the hydraulic free piston engine completes the energy output of the demand power within several cycles.

[0139] For the determination of the switching moment, it is necessary to combine the piston displacement and piston speed data feedback in real-time, judge the motion state of the piston based on these two parameters, and then judge the switching point closest to the nominal switching moment, perform a switching moment delay operation, and finally complete the dynamic switching at the actual switching moment. The demand power and the switching moment obtained from this step are transmitted to Figure 9 the machine learning control model in the lower part to complete the parameter regression prediction of the hydraulic free piston engine. Finally, each parameter is input into the combustion process, thermodynamic process and hydraulic process of the engine model, and the coupling dynamics principle is used to finally achieve the fast and smooth dynamic switching of the hydraulic free piston engine model.

[0140] Finally, applying the above-mentioned control strategy for the dynamic process stability of the hydraulic free piston engine, the verification work of the control strategy for the real-time switching control of the engine model is carried out. For the engine model, the stability control of the dynamic process with the output power ranging from 10 kW to 30 kW is carried out, and the parameter switching results of the hydraulic free piston engine model before and after the switching of the two working conditions are as Figure 10 shown.

[0141] Figure 10 is the operating curve of the main parameters of the hydraulic free piston engine model when the required power is switched from 10 kW to 30 kW at the 2nd second. The red area represents the transition time interval of the engine working condition switching under the change of the required power. Obviously, with the help of the proposed control strategy, the engine can achieve a rapid working condition switching. Figure 10 In (a)- Figure 10 In (f) shows that the engine only needs two strokes of piston movement to successfully complete the full switching of the working condition, and the displacement, velocity, acceleration of the piston and the in-cylinder pressure all remain in good condition. It shows the real-time power output of the dynamic switching of the engine working condition. As mentioned in the control strategy, when the engine switches to a working condition with a larger power, there will be a progressive increase. At the same time, the compression ratio of the engine is well controlled within 10 - 30, indicating that the engine is outputting energy efficiently and stably.

[0142] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A free piston engine stability control method based on machine learning, characterized in that: The following steps are involved: S1. Construct a hydraulic free piston engine simulation model, including thermodynamic model, hydraulic model and dynamic model; S2. Use a hydraulic free piston engine simulation model to simulate and analyze the hydraulic free piston engine, obtain the corresponding relationships between the output power of the hydraulic free piston engine and the return pressure, output pressure and combustion duration, and draw a MAP diagram; S3, taking the output power, the return pressure, the output pressure and the combustion duration as input features and the injection quantity as the output parameter, constructing an injection quantity prediction model based on the BP neural network; S4. The return pressure, output pressure, combustion duration and the injection amount obtained based on the injection amount prediction model are input as input parameters into the hydraulic free piston engine simulation model. By adjusting the parameters, stable control of the target power and the working condition switching command is achieved.

2. The free piston engine stability control method based on machine learning according to claim 1, characterized in that: In S1, the method for constructing the thermodynamic model is specifically as follows: Based on the first law of thermodynamics and the law of cylinder volume change driven by unbalanced pressure, the corresponding relationship between the cylinder pressure of the engine and the heat change of the working fluid in the cylinder is obtained; Determine the relationship between the heat change of the working fluid in the cylinder, the heat generated by the fuel combustion, and the heat lost by the heat exchange between the working fluid in the combustion chamber and the environment.

3. The free piston engine stability control method based on machine learning according to claim 2, characterized in that: Based on the first law of thermodynamics and the law of cylinder volume change driven by unbalanced pressure, the corresponding relationship between the cylinder pressure of the engine and the heat change of the working fluid in the cylinder is obtained, which is specifically: Based on the first law of thermodynamics and the law of cylinder volume change driven by unbalanced pressure, we get: δQ=dU+δW (1) P C V C =m a RT (2) In the formula, δQ represents the heat absorbed by the thermodynamic system from the outside, dU is the increased internal energy of the system, and δW represents the work done by the system to the outside; P C is the gas pressure acting on the piston, V C Represents the cylinder volume driven by unbalanced pressure, m a represents the mass of the ideal gas, R represents the air gas constant, and T represents the temperature of the gas in the cylinder; Further deduction and solution of various parameters in formula (1) yield: δW=P C ×dV C (3) dU=m a C V dT (4) In the above formula, C V represents the specific heat at constant volume; The above formulas (1)-(5) are calculated together to obtain the corresponding relationship between the cylinder pressure and the change of the thermal energy of the working fluid in the cylinder: The relationship between the change in the heat of the working medium in the cylinder and the heat generated by the combustion of the fuel and the heat lost by the heat exchange between the working medium in the combustion chamber and the environment is determined as follows: The heat change of the working fluid in the cylinder is affected by the heat generated by the combustion of the fuel and the heat lost by the heat exchange between the working fluid and the environment in the fuel cavity. The heat generated by the combustion is solved by the Weibull function, while the heat lost by the heat exchange between the working fluid and the environment in the combustion cavity is related to the working fluid temperature and the ambient temperature. The expression is as follows: Q=Q C +Q W (7) In the formula, Q represents the heat conversion of the working fluid in the cylinder, Q C Indicates the heat generated by fuel combustion, Q W It represents the heat loss caused by the heat exchange between the working fluid in the combustion chamber and the environment, H u Represents diesel lower calorific value, g f is the circulating oil supply, η u Represents the combustion efficiency, m c Indicates the combustion quality index, t c represents the combustion duration, α is the instantaneous heat transfer coefficient, T and T amb Represent the working fluid temperature and the ambient temperature respectively.

4. The free piston engine stability control method based on machine learning according to claim 1, characterized in that: In S1, the method for constructing the hydraulic model is specifically as follows: Oil inlet holes and oil outlet holes are respectively set at the three pump chamber positions of the hydraulic free piston engine simulation model; and the fluid continuity equation is used to obtain the actual oil inlet and outlet of the engine as well as the leakage loss flow.

5. The free piston engine stability control method based on machine learning according to claim 4, characterized in that: The oil inlet and outlet holes are respectively set at the three pump chamber positions of the hydraulic free piston engine simulation model; and the actual oil inlet and outlet of the engine and the leakage loss flow are obtained by using the fluid continuity equation, which is specifically: Oil holes are set at the three pump chambers of the hydraulic free piston engine to generate pressure. The fluid continuity equation is used to derive the fluid equations of the three pump chambers in the working state, which are as follows: In the above formula, P h represents the pump chamber output pressure, K represents the elastic modulus, Q in With Q out Respectively represent the actual oil inlet and outlet of the oil inlet, A p represents the piston contact area, △q represents the leakage loss flow; Expanding the flow parameters in formula (10) yields: △q=P h ×λ (11) In the above formula, λ represents the leakage coefficient, c0 represents the flow coefficient, A0 represents the flow area, ρ represents the hydraulic oil density, P c Represents the pressure constant.

6. The free piston engine stability control method based on machine learning according to claim 1, characterized in that: In S1, the method for constructing the kinetic model is specifically as follows: Based on Newton's second law, the kinematic law of the piston of the hydraulic free piston engine under specific working conditions is analyzed, and the variation law of the displacement, velocity and acceleration of the piston is obtained.

7. The free piston engine stability control method based on machine learning according to claim 6, characterized in that: Based on Newton's second law, the kinematic law of the piston of the hydraulic free piston engine under specific working conditions is analyzed, and the change law of the displacement, velocity and acceleration of the piston is obtained, which is specifically: The power of the hydraulic free piston engine includes: cylinder pressure, hydraulic pressure and friction; the left side of the hydraulic free piston engine generates cylinder pressure, and the right side of the hydraulic free piston engine generates hydraulic pressure; Based on Newton's second law, the hydraulic free piston moves as follows: In the formula, and They are the displacement, velocity and acceleration of the piston, M represents the piston mass; P represents the pressure in the cylinder, P1 represents the recovery pressure, P2 represents the periodic pressure change, P3 represents the output pressure, S, S1, S2, S3 represent the contact area of ​​each piston, Indicates that the friction force on the piston is opposite to the direction of motion, F f Represents the reciprocating friction force.

8. The free piston engine stability control method based on machine learning according to claim 1, characterized in that: In S3, the output power, the return pressure, the output pressure and the combustion duration are used as input features, and the injection amount is used as an output parameter to construct a fuel injection amount prediction model based on a BP neural network, specifically: Based on the hydraulic free piston engine simulation model, a data set constructed by output power, recovery pressure, output pressure, combustion duration, and injection amount was obtained, and the data set was divided into a training set and a test set at a ratio of 500:100; The training set is used to train the network parameters of the fuel injection quantity prediction model; The test set is used to evaluate the generalization ability of the injection quantity prediction model.

9. The free piston engine stability control method based on machine learning according to claim 1, characterized in that: In S4, the return pressure, output pressure, combustion duration and the injection amount obtained based on the injection amount prediction model are input as input parameters to the hydraulic free piston engine simulation model, and the stable control of the target power and the working condition switching instruction is achieved through parameter adjustment, specifically: Determine the optimal switching time based on the movement displacement and movement speed of the piston; The piston movement state of the hydraulic free piston engine is collected in real time by the control end. After receiving the required power change instruction, the optimal switching point that meets the switching conditions after that moment is obtained according to the piston movement state to achieve smooth switching of the hydraulic free piston engine.

10. The free piston engine stability control method based on machine learning according to claim 9, characterized in that: After receiving the required power change instruction, the optimal switching point that first meets the switching condition after that moment is obtained according to the piston movement state, so as to realize the smooth switching of the hydraulic free piston engine, wherein the required power is determined specifically as follows: The required power needs to compare the actual power at this moment and the difference between the required power at the previous moment. When the difference between the two is less than the constraint range, the target power at this moment is directly determined; when the difference between the two is greater than the constraint range, the required power at this moment is the sum of the required power at the previous moment and the maximum value of the constraint. The small part of the power that is not met will continue to be compared in the next cycle. Finally, the hydraulic free piston engine completes the required power capacity output within a few cycles.

Images