Prechamber spark control with two ignition sources

By using feedforward artificial neural network to control the pre-combustion chamber and auxiliary spark ignition timing in TJI engines, the problem of poor combustion quality in the prior art is solved, and higher fuel economy and engine stability are achieved.

CN119096043APending Publication Date: 2024-12-06FCA US LLC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380036575.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-04-26
Filing Date
2023-04-24
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art is difficult to effectively control the pre-chamber spark and auxiliary spark ignition timing in turbulent jet ignition (TJI) engines, resulting in poor combustion quality, reduced fuel economy and engine knocking.

Method used

Using a trained feedforward artificial neural network (ANN) to control the pre-combustion chamber and auxiliary spark ignition timing, the target spark timing is generated by accessing operating parameters from the sensor group, and the target spark timing is determined to determine the maximum braking torque (MBT) spark timing. The difference in spark interlacing can be achieved.

Benefits of technology

Improves the accuracy of spark control, reduces the calibration work of the spark model, significantly improves combustion stability and fuel economy, and optimizes engine torque efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119096043A_ABST
    Figure CN119096043A_ABST
Patent Text Reader

Abstract

A combustion control system and method for a turbulent jet ignition engine are presented. The controller is configured to access a trained feedforward artificial neural network (ANN) configured to model a first spark and a maximum braking torque (MBT) from the first ignition source based on measured operating parameters, generate the first spark and the MBT using the ANN, generate a second spark of the second ignition device, and determine a target spark timing. The ANN may be further configured to receive an input related to spark interleaving.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Non-Provisional Application No. 17 / 729,709, filed on April 26, 2022, the entire contents of which are hereby incorporated by reference. Technical Field

[0003] The present application relates generally to spark ignition (SI) engines and, more particularly, to techniques for determining pre-chamber spark ignition timing and secondary spark ignition timing for engines having turbulent jet ignition (TJI). Background Art

[0004] In today's internal combustion engines, the spark timing with respect to the corresponding piston stroke affects the quality of combustion. Therefore, it is desirable to accurately determine and control the spark timing, because poor combustion quality due to inaccurate spark timing can lead to reduced fuel economy and / or engine knock. An internal combustion engine using turbulent jet ignition (TJI) typically includes a pre-combustion chamber with a first spark plug installed inside. The second spark plug is typically installed at an offset position in the main combustion chamber. The flame is started in the pre-combustion chamber and sprayed into the main combustion chamber to ignite a large amount of (bulk) fuel-air mixture. In conventional internal combustion engines with a single spark plug (or multiple spark plugs ignited simultaneously) in each cylinder, the torque can be estimated by multiplying the maximum potential torque of the engine by the torque efficiency. However, for a turbulent jet ignition (TJI) engine, there are both pre-combustion chamber ignition from the first spark plug and auxiliary ignition from the second spark plug. In this regard, both the pre-combustion chamber ignition timing and the auxiliary ignition timing have an impact on the torque efficiency. Therefore, there remains a need for improvements in the art to properly account for the two ignition timings associated with a TJI engine when determining torque. Summary of the invention

[0005] According to an exemplary aspect of the present invention, a combustion control system for a TJI engine is proposed. In an exemplary embodiment, the TJI engine has a first ignition device arranged in a pre-combustion chamber at the top of the cylinder and a second ignition device arranged to communicate with the main combustion chamber. In some embodiments, the controller is configured to: access a trained feedforward artificial neural network (ANN), which is configured to model a first spark from the first ignition device and a maximum brake torque (MBT) spark based on operating parameters; use the ANN to generate a first spark and an MBT spark for the engine; generate a second spark for the second ignition device; determine the target spark timing of the engine based on at least one of the first spark, the second spark, and the MBT spark.

[0006] According to some embodiments, the second spark is modeled according to an ANN. In some embodiments, the ANN is defined by four layers, including an input layer, two hidden layers with twelve neurons in each layer, and an output layer.

[0007] According to other embodiments, the controller is further configured to determine a spark stagger corresponding to the difference between the first spark timing and the second spark timing. In some embodiments, the spark stagger is based on a two-dimensional surface with engine speed and charge mass as input. In an embodiment, the controller is also configured to access an ANN, which is further configured to model the spark stagger. In an embodiment, the controller is further configured to use the ANN to generate a first spark and an MBT spark for the engine with operating parameters and spark stagger from a sensor group as input. In an embodiment, a second spark is generated based on the difference between the spark stagger and the first spark timing. In an embodiment, the ANN is defined by four layers, including an input layer, two hidden layers with thirteen neurons in each layer, and an output layer. In an embodiment, the operating parameters include at least six parameters, including charge mass and temperature, intake camshaft position and exhaust camshaft position, engine speed, and engine coolant temperature.

[0008] According to another exemplary aspect of the present invention, a combustion control method for a TJI engine is proposed. In an exemplary embodiment, the TJI engine has a first ignition device arranged in a pre-combustion chamber at the top of the cylinder and a second ignition device arranged to communicate with the main combustion chamber. In an exemplary embodiment, the method includes: receiving measured operating parameters from a sensor group by a controller of the vehicle; accessing an ANN by the controller, the ANN being configured to model a first spark from the first ignition device and a maximum brake torque (MBT) spark based on the measured operating parameters, by the controller and from the sensor group; using the ANN by the controller to generate a first spark and an MBT spark for the engine with the measured parameters from the sensor group as input; generating a second spark for the second ignition device by the controller; determining the target spark timing of the engine by the controller based on at least one of the first spark, the second spark, and the MBT spark.

[0009] According to some embodiments, the method further includes: generating a second spark using an ANN by a controller. In an embodiment, the ANN is defined by four layers, including an input layer, two hidden layers with twelve neurons in each layer, and an output layer. In other embodiments, the method further includes: determining, by a controller, a spark interleaving corresponding to a difference between the first spark timing and the second spark timing. In an embodiment, the spark interleaving is based on a two-dimensional surface with engine speed and charge mass as inputs. In an embodiment, the method further includes: accessing the ANN by a controller, the ANN being further configured to model the spark interleaving. In an embodiment, the method further includes: generating a first spark and an MBT spark for the engine using the ANN by a controller with six measured parameters and the spark interleaving as inputs. In an embodiment, the second spark is generated based on the difference between the spark interleaving and the first spark timing. In an embodiment, the ANN is defined by four layers, including an input layer, two hidden layers with thirteen neurons in each layer, and an output layer.

[0010] Other areas of application of the teachings of the present disclosure will become apparent from the detailed description, claims, and drawings provided below, in which the same reference numerals refer to the same features in the several views of the drawings. It should be understood that the detailed description, including the disclosed embodiments and the drawings cited therein, is merely exemplary in nature and is intended for illustration purposes only and is not intended to limit the scope of the present disclosure, its application, or use. Therefore, changes that do not depart from the spirit of the present disclosure are intended to be within the scope of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a front view of a cylinder top of an exemplary TJI engine according to the principles of the present disclosure;

[0012] Figure 2 It is along Figure 1 A cross-sectional view taken along line 2-2;

[0013] Figure 3 is a schematic diagram of an engine system including a controller implementing various control methods according to the present disclosure;

[0014] Figure 4 is a flow chart of an example spark estimation diagram according to one example of the present disclosure;

[0015] Figure 5 is a flow chart of an example spark estimation map according to another example of the present disclosure; and

[0016] Figure 6 is a flow chart of an example torque estimation control method according to the principles of the present disclosure. DETAILED DESCRIPTION

[0017] As previously discussed, there is still a need for improvements in the relevant technology to properly consider the pre-chamber and auxiliary spark ignition timing associated with the TJI engine when determining the engine torque. The methods and techniques discussed in this article solve the problem of engine spark estimation for a TJI engine with two independent ignition timings. As will be described in detail herein, the present method provides a solution to control the pre-chamber and auxiliary spark ignition timings by using a trained feedforward artificial neural network (ANN). The first ANN models the pre-chamber spark, auxiliary (side) spark, and maximum brake torque (MBT) spark timing based on operating parameters. In a non-limiting example, the operating parameters include six inputs: intake camshaft position and exhaust camshaft position, the mass and temperature of the charge provided to each cylinder of the SI engine, the engine speed, and the engine coolant temperature (ECT). MBT is the maximum torque achieved at the optimal engine speed and charge. It will be understood that if the engine is running at MBT, the torque efficiency will be 100%. As will be discussed in greater detail herein, a second ANN model based on seven inputs is also disclosed, including the six inputs used for the first ANN and additional interleaved inputs.

[0018] First reference Figure 1 and Figure 2 , an exemplary cylinder head is shown and generally identified by reference numeral 10. The cylinder head 10 is included in an internal combustion engine 20 having an engine block 22 including one or more cylinders 26. A piston 30 is supported to reciprocate within the cylinder 26 defined in the engine block 22. The cylinder head 10, the cylinder 26 and the piston 30 cooperate to define a combustion chamber 32. The exemplary internal combustion engine 20 includes two intake ports 40 and two exhaust ports 42. As is well known, the intake ports 40 and the exhaust ports 42 are opened and closed via valves to provide fluid communication between the cylinders and the intake manifold and the exhaust manifold (not specifically shown). It should be understood that although two intake ports 40 and two exhaust ports 42 are shown, the internal combustion engine 20 may include any number of intake valves and / or exhaust valves. For example only, the engine block 22 may be configured to have four cylinders. It should be understood that the methods and control strategies discussed herein may be applicable to TJI engines having different numbers of cylinders.

[0019] The cylinder head 10 includes a pre-chamber 50 having a pre-chamber insert 52 disposed therein. A first ignition device or spark plug 54 is disposed in the pre-chamber 50. The pre-chamber insert 52 defines a plurality of small apertures 56 therein. The apertures 56 provide communication between the pre-chamber 50 and the combustion chamber 32. An injector, not specifically shown, may deliver fuel to the pre-chamber 50. The first spark plug 54 may ignite the fuel in the pre-chamber 50.

[0020] A second ignition device or spark plug 60 is disposed in an adjacent cavity 62 disposed on the cylinder top 10. In some embodiments, the second spark plug 60 may be referred to as a side spark plug. Once ignited, the fuel is forced through the orifice 56 of the pre-chamber insert 52. The flame is initiated inside the pre-chamber 50 and injected into the main combustion chamber 32 to ignite a large amount of the fuel-air mixture.

[0021] Now refer to Figure 3 , a combustion control system 66 for a vehicle 68 is shown having an engine control unit or controller 70 that transmits signals between the engine 20 and various vehicle systems 72. The controller 70 may receive various input signals from the vehicle systems 72 and sensors 74. The vehicle systems may include, but are not limited to, a heating, ventilation, and air conditioning (HVAC) system, an accelerator pedal 80, and an alternator. For example, a driver of the vehicle 68 may provide input indicating a torque request via the accelerator pedal 80. The sensors 74 may include an air charge sensor 90, an intake camshaft position sensor 92, an exhaust camshaft position sensor 94, an engine speed sensor 96, and an engine coolant temperature (ECT) sensor 98.

[0022] The engine 20 draws air into the intake manifold through an intake system regulated by a throttle. The air flow (MAF) and / or manifold absolute pressure (MAP) sensors and the intake air temperature (IAT) together constitute the charge sensor 90. The charge sensor 90 measures the air flow / pressure indicating the mass and temperature of the charge provided to each of the multiple cylinders 26 in the engine 20. Although shown and described as a single charge sensor 90, it should be understood that there may be multiple sensors (for example, in addition to the IAT sensor, there may also be MAF and / or MAP sensors). For example, the air mass may be determined or estimated using a charge model that models the mass based on the intake air flow and / or air pressure and potentially other parameters. The intake camshaft position sensor 92 measures the position of the intake camshaft (not shown) of the engine 20. The exhaust camshaft position sensor 94 measures the position of the exhaust camshaft (not shown) of the engine 20. The engine speed sensor 96 measures the rotational speed of the crankshaft of the engine 20. The ECT sensor 98 measures the temperature of the coolant (water, oil, etc.) flowing through and thereby cooling the engine 20 .

[0023] The controller 70 controls the operation of the engine 20. The controller 70 is configured to receive inputs from the vehicle system 72 and the sensor 74. The controller 70 then controls the engine 20 (e.g., airflow, fuel, and spark) to achieve the torque request. Typically, the controller 70 is configured to determine a spark model that calculates an optimal ignition set point based on fuel economy, emissions, knock, and noise, vibration, and harshness (NVH). The controller can then use the spark model as an input to the torque model. The controller 70 can determine a torque model or torque request for the engine 20 based on various conditions (such as indicated from the vehicle system 72) to achieve the desired engine torque requested by the driver. The torque model of a conventional engine with a single spark plug (or multiple spark plugs ignited simultaneously) in each cylinder uses a single incremental spark to estimate the engine torque. The incremental spark is the difference between the maximum brake torque (MBT) spark timing and the actual spark timing. The MBT spark timing is output by the spark model. This conventional spark model is unsatisfactory for a TJI engine with two independent ignition timings.

[0024] As explained above, the prior art methods of determining spark are not ideal for a TJI engine because a TJI engine has pre-chamber ignition from the first spark plug 54 and auxiliary ignition from the second spark plug 62. As will be described in greater detail herein, the present systems, methods, and techniques allow for independent control of pre-chamber spark timing using the first spark plug 54 and independent control of auxiliary spark timing using the second spark plug 60. Thus, the methods disclosed in the present disclosure take into account both the pre-chamber spark timing from the first spark plug 54 and the auxiliary spark timing using the second spark plug 60.

[0025] According to the present disclosure, MBT spark timing is defined as the start spark timing to reach MBT. For example, when the first spark plug 54 of the pre-chamber 50 starts combustion earlier than the auxiliary spark from the second spark plug 60, then the MBT spark timing is the pre-chamber spark timing to reach MBT. Similarly, when the auxiliary spark from the second spark plug 60 starts combustion earlier than the first spark plug 54 in the pre-chamber 50, then the MBT spark timing is the auxiliary spark timing to reach MBT. For example, if the pre-chamber 50 starts combustion earlier than the auxiliary spark, the incremental spark is calculated as the difference between the MBT and the pre-chamber 50 spark timing.

[0026] The timing of the pre-chamber spark ignition (ignition of the first spark plug 56) and the auxiliary spark ignition (ignition of the second spark plug 60) is independently controlled by the controller 70. As used herein, the term "spark interleaving" is used to refer to the pre-chamber ignition timing minus the auxiliary spark plug ignition timing. If the spark interleaving is positive, the controller 70 determines that the pre-chamber 50 has initiated combustion and the auxiliary spark fires thereafter. Similarly, if the spark interleaving is negative, the controller 70 determines that the auxiliary spark initiated combustion. The interleaving between the pre-chamber 50 spark ignition timing and the auxiliary spark ignition timing will be determined taking into account the combustion rate, fuel economy, combustion stability and other factors under each engine operating condition.

[0027] The remote calibration system, which is not part of the vehicle 68, includes a computer system that interacts with a dynamometer (not shown), which may be part of the calibration system or separate from the calibration system, for obtaining dynamometer data of the engine 20, which is used to generate ANN calibrations that are provided as input to the controller 70 and then used for spark and torque control. The term "obtain" as used herein refers to transmitting all generated ANN calibrations to the controller 70 for storage in its memory and subsequent retrieval for use.

[0028] The ANN is designed to enable it to accurately control both the pre-chamber spark and the auxiliary spark in the TJI engine. The design process involves the selection of various parameters, such as but not limited to the type and number of input / output, the number of hidden layers, the number of neurons per layer, and the training / transfer function. In an exemplary embodiment, the optimal ANN design is six inputs (charge mass / temperature, engine speed, intake / exhaust camshaft position, and ECT), two outputs (basic or pre-chamber and MBT spark timing), two hidden layers (twelve neurons per layer), Bayesian regularized back propagation training function (also referred to as "trainbr") and hyperbolic tangent sigmoid transfer function (also referred to as "tansig"). However, it should be understood that this is only an exemplary ANN design and these parameters can vary according to specific vehicle / engine applications. As an example only, another suitable training function, such as Levenberg-Marquardt back propagation training function (also referred to as "trainlm"), and / or another suitable transfer function, such as rectified linear unit (also referred to as ReLu) can be used. Other suitable types and / or numbers of inputs and / or outputs can also be used. As previously described, in one exemplary embodiment, a PQ activation function may be implemented to further reduce the throughput of the processor. In another exemplary embodiment, as described herein, the optimal ANN design is seven inputs, which adds an additional interleaved mapping input. In a second exemplary embodiment, as described below, spark interleaving is added as the seventh input.

[0029] For example, the ANN may be trained by an external calibration system using dynamometer data of a SI engine that is collected by the external calibration system during a vehicle calibration period. In one exemplary embodiment, the external calibration system is configured to collect all necessary nominal and non-nominal calibration data during an initial first calibration sub-period, thereby eliminating the need to collect calibration data during a second calibration sub-period that is typically reserved for non-nominal calibration data collection until vehicle startup.

[0030] The remote calibration system, which is not part of the vehicle 68, includes a computer system that interacts with a dynamometer (e.g., a dynamometer sensor), which may be part of the calibration system or separate from the calibration system, to obtain dynamometer data for the engine 20, which is used to generate an ANN calibration, which is provided as an input to the controller 70 and then used for combustion phasing control. The term "obtain" as used herein refers to transmitting all generated ANN calibrations to the controller 70 for storage in its memory and subsequent retrieval for use. The ANN is designed so that it can accurately estimate spark for subsequent combustion control in the SI engine. The design process involves the selection of various parameters, such as, but not limited to, the type and number of inputs / outputs, the number of hidden layers, the number of neurons per layer, and the activation / transfer functions. In an exemplary embodiment, the trained ANN defines 12x12 neurons per layer ( Figure 4 ), but it is understood that a slightly larger trained ANN (e.g., 13x13 neurons per layer ( Figure 5 )).

[0031] Reference Figure 4, shows a first spark control or model implemented by the controller 70 and generally identified by reference numeral 100. Generally, the spark model 100 uses an ANN 102 to estimate the timing of a pre-chamber spark 110, an auxiliary spark 112, and a maximum brake torque (MBT) spark 114. The ANN 102 uses an input layer 118 having six inputs, including engine speed 120, charge per cylinder per cycle 122, intake cam position 124, exhaust cam position 126, engine coolant temperature 128, and intake charge temperature 130. The ANN 102 determines the optimal spark values ​​for two ignition set points based on fuel economy, emissions, knock, and noise, vibration, and harshness (NVH). In an example, the engine speed 120 may be provided by the engine speed sensor 96, the charge mass 122 and the charge temperature 130 may be provided by the charge sensor 90, the intake cam position 124 may be provided by the intake camshaft position sensor 92, the exhaust cam position 126 may be provided by the exhaust camshaft position sensor 94, the final speed 120 may be provided by the engine speed sensor 96, and the engine coolant temperature 128 may be provided by the ECT sensor 98. The ANN 102 is defined by four layers, including an input layer 118, two hidden layers 132 having twelve neurons each, and an output layer 134, the output layer 134 including the pre-chamber spark 110, the auxiliary spark 112, and the maximum brake torque (MBT) spark 114.

[0032] Reference Figure 5 , a second spark control or model implemented by the controller 70 is shown and is generally identified by reference numeral 160. Generally, the spark model 160 uses the ANN 142 to estimate the timing of the pre-chamber spark 110 and the maximum brake torque (MBT) spark 114. The second spark control 160 uses an interleaving map 170 as a pre-calibrated training data set. The interleaving map 170 can output an interleaving 176 based on the engine speed 172 and the charge 174. The model can be retrained for any interleaving changes in the engine mapping. The second spark control 160 improves the robustness of interleaving changes by adding the interleaving as an additional input 176 to the input layer 177 of the neural network 142. In this regard, only the pre-chamber spark 110 and the spark MBT 114 are retained as outputs in the output layer 178 of the neural network 142.

[0033] The secondary spark 182 is calculated by subtracting the stagger 176 from the pre-chamber spark timing 110. The secondary spark control 160 provides a stagger map 170 (as an input) upstream of the ANN 142 to calibrate the stagger on the engine map. The seven-input ANN spark model will protect the upstream stagger changes when the ANN 142 is trained with DOE data sets that vary stagger, engine speed, charge, intake and exhaust cam positions, coolant temperature, and charge temperature. The trained feed-forward artificial neural network 142 is defined by four layers, including an input layer 177, two hidden layers 192 with thirteen neurons each, and an output layer 178 including the pre-chamber spark 110 and the maximum brake torque (MBT) spark 114.

[0034] The ANN 142 uses seven inputs, including engine speed 120, charge volume per cylinder per cycle 122, intake cam position 124, exhaust cam position 126, engine coolant temperature 128, intake charge temperature 130, and stagger 176. The ANN 142 determines the optimal spark value for two ignition set points based on fuel economy, emissions, knock, and noise, vibration, and harshness (NVH). In an example, the engine speed 120 can be provided by the engine speed sensor 96, the charge volume 122 and the charge temperature 130 can be provided by the charge sensor 90, the intake cam position 124 can be provided by the intake camshaft position sensor 92, the exhaust cam position 126 can be provided by the exhaust camshaft position sensor 94, the final speed 120 can be provided by the engine speed sensor 96, and the engine coolant temperature 128 can be provided by the ECT sensor 98.

[0035] Both ANN 102 and ANN 142 capture the nonlinear relationship between the input factors. The results show that the spark control accuracy is improved, especially at non-nominal cam positions and non-nominal temperatures. By controlling two spark timings (pre-chamber spark and side spark in the TJI engine), ANN control model 100 and ANN control model 160 significantly reduce the calibration work required in the spark model.

[0036] Referring now to FIG. 7 , an example flow chart of an example torque estimation control method 200 according to the principles of the present disclosure is shown. Although the method 200 is described with respect to the exemplary TJI engine 20 , it should be understood that the method may be applicable to any TJI engine. Control begins at 212 . At 214 , control determines whether the vehicle system 72 has received a torque command. In this example, at 214 , control determines whether the accelerator pedal 80 has received a torque command. If a torque command has not been received, control loops to 214 . If a torque command has been received, control starts the torque model at 220 . At 222 , the controller 70 determines whether the engine 20 is a TJI engine. If the engine is not a TJI engine, the method returns at 224 . If control determines that the engine 20 is a TJI engine, the controller 70 commands an MBT spark from the TJI spark model at 230 . When using spark control 100 ( Figure 3 ), control proceeds to 234 and 236. Alternatively, when spark control 160 ( Figure 4 ), control proceeds to 244 and 246.

[0037] At 234 and 244, control receives the intake cam position 124 and the exhaust cam position 126; the air volume 122 and the air temperature 130; the engine speed 120 and the engine coolant temperature 128. At 236, control calculates the pre-chamber spark 110, the side spark 112, and the spark MBT 114 using the above-mentioned six-factor ANN 102. Alternatively, at 246, control calculates the pre-chamber spark 110, the side spark 112, and the spark MBT 114 using the seven-factor ANN 142 (which uses the predetermined spark stagger as an additional input as described above). Control estimates the spark of the pre-chamber spark and the side spark at 250. Control estimates the actual torque at 252. Then, control determines whether the torque request has been met at 260. If the torque request has not been met, control loops to 230. If the torque request has been met, control ends at 262. It should be understood that, alternatively, control may loop to 214 for continuous control. Figure 6 The described techniques include optional steps related to torque modeling. It should be understood that the present disclosure is specifically directed to determining spark values ​​for two ignition set points. In this regard, torque techniques are merely exemplary in nature and are generally not within the scope of the present disclosure.

[0038] It should be understood that the term "controller" used herein refers to any suitable control device configured to perform at least a portion of the technology disclosed herein. Non-limiting examples include an application specific integrated circuit (ASIC), one or more processors, and a non-transitory memory having instructions stored thereon, which, when executed by one or more processors, causes the controller to perform a set of operations corresponding to at least a portion of the technology disclosed herein. One or more processors may be a single processor or two or more processors running in a parallel or distributed architecture. It should also be understood that the mixing and matching of features, elements, methods, and / or functions between various examples may be explicitly contemplated herein, so that those skilled in the art will understand from this teaching that, unless otherwise stated above, the features, elements, and / or functions of one example may be appropriately incorporated into another example.

Claims

1. A combustion control system for a turbulent jet ignition (TJI) engine having a first ignition device disposed in a pre-combustion chamber at the top of a cylinder and a second ignition device disposed for communication with a main combustion chamber, the system comprising: a sensor set configured to measure an operating parameter; and A controller configured to: accessing a trained feed-forward artificial neural network (ANN), the ANN configured to model a first spark from the first ignition device and a maximum brake torque (MBT) spark based on the operating parameters; generating the first spark and the MBT spark for the engine using the ANN with the measured operating parameters from the sensor group as input; generating a second spark for the second ignition device; A target spark timing of the engine is determined based on at least one of the first spark, the second spark, and the MBT spark. 2 . The system of claim 1 , wherein the controller is further configured to determine a spark stagger corresponding to a difference between the first spark timing and the second spark timing.

3. The system of claim 2 wherein the spark stagger is based on a two-dimensional surface having engine speed and charge mass as inputs. 4 . The system of claim 3 , wherein the controller is further configured to access the ANN, the ANN further configured to perform additional modeling of the spark interleaving. 5 . The system of claim 4 , wherein the controller is further configured to generate the first spark and the MBT spark for the engine using the ANN with measured parameters from the sensor group and the spark stagger as inputs. 6 . The system of claim 5 , wherein the second spark is generated based on a difference between the spark stagger and the first spark timing.

7. The system of claim 2, wherein the ANN is defined by four layers, including an input layer, two hidden layers each having thirteen neurons, and an output layer.

8. The system of claim 1 wherein the operating parameters include at least six parameters including charge mass and temperature, intake and exhaust camshaft positions, engine speed, and engine coolant temperature.

9. The system of claim 1, wherein the second spark is modeled by the ANN.

10. The system of claim 9, wherein the ANN is defined by four layers, including an input layer, two hidden layers with twelve neurons in each layer, and an output layer.

11. A combustion control method for a turbulent jet ignition (TJI) engine having a first ignition device disposed in a pre-combustion chamber at a cylinder top and a second ignition device disposed for communication with a main combustion chamber, the method comprising: receiving, by a controller, measured engine operating parameters from a sensor group; accessing, by the controller, a trained feed-forward artificial neural network (ANN), the ANN configured to model a first spark in the first ignition device and a maximum brake torque (MBT) spark based on operating parameters; generating the first spark and the MBT spark for the engine by the controller using the ANN with measured parameters from the sensor group as input; generating, by the controller, a second spark for the second ignition device; A target spark timing of the engine is determined by the controller based on at least one of the first spark, the second spark, and the MBT spark.

12. The method of claim 11, further comprising: The second spark is generated by the controller using the ANN.

13. The method of claim 12, wherein the ANN is defined by four layers, including an input layer, two hidden layers with twelve neurons in each layer, and an output layer.

14. The method of claim 11, further comprising: A spark stagger corresponding to a difference between the first spark timing and the second spark timing is determined by the controller.

15. The method of claim 14, wherein the spark stagger is based on a two-dimensional surface having engine speed and charge mass as inputs.

16. The method of claim 15, further comprising: The ANN is accessed by the controller and is further configured to perform additional modeling of the spark interleaving.

17. The method of claim 16, further comprising: The first spark and the MBT spark are generated by the controller for the engine using the ANN with the measured parameters and the spark stagger as inputs.

18. The method of claim 17, wherein the operating parameters include at least six parameters including charge mass and temperature, intake and exhaust camshaft positions, engine speed, and engine coolant temperature.

19. The method of claim 17 wherein said second spark is generated based on a difference between said spark stagger and said first spark timing.

20. The method of claim 18, wherein the ANN is defined by four layers, including an input layer, two hidden layers each having thirteen neurons, and an output layer.