Makoto Nagaoka

A novel total hydrocarbon (THC) emission concentration estimation model is proposed for reduction of engine development cost as well as simplification of measurement system. The model is based on machine learning algorithm including the least absolute shrinkage and selection operator (LASSO) regression and bagging techniques. Major features of the proposal model are taking the absorbance spectrum of Fourier transform infrared (FTIR) spectrometer as input and incorporating not only spectra of the engine exhaust gas but also those of individual hydrocarbon and inorganic gas components as training data set. This method was validated on the exhaust gas before the catalyst of a gasoline engine. The results show an error of less than 5% in both steady and transient operating conditions, outperforming the 20 % error of conventional regression model using only the reference hydrocarbon concentrations. We also evaluate the contribution to performance improvements in THC estimation of employing FTIR spectrum and incorporating spectrum information of gas components, respectively.

This paper deals with the machine learning based controller design for realizing nonlinear model predictive control (MPC) with low computational cost, and the application for a diesel engine air path system is shown. Since the solution of an optimal control problem considered in MPC depends on several variables at each time, the relationship between the variables and the solution, that is, the control law could be approximated by a neural network. In the case of high dimensional inputs, however, some efficient sampling methods for the approximation are needed to avoid a combinatorial explosion. To reduce the sampling dimension, we newly applied an efficient sampling method which is combined with a design of experiment. Using the method, the neural network structured optimal controller that operates the valves in the air path system was designed, and its tracking capability to the reference value was demonstrated in the simulation. The computational time of the controller was approximately 0.022ms at each cycle, indicating that fast computation of MPC was achieved.

This paper considers machine learning based virtual design process of engine control system, and the demonstration in a diesel engine air path control is shown. This process contains two steps of machine learning. In the first step, a control-oriented forward model that predicts the transient behavior of the engine is learned from detailed engine model by using recurrent neural network (RNN). In the second step, an inverse model that determines the optimal control inputs to follow the references is learned from the numerical computation results of the offline model predictive control (MPC). The forward and inverse models could be used as a state observer and a controller, respectively, in a control system. An experiment of a diesel air path control system designed by the process was conducted using rapid control prototyping (RCP), and its following capability to the reference was demonstrated.

The submodels of spark channel short circuit and blow-out, which were described in the 1st report, were implemented into a spark ignition model. The short circuit, whose major factor is electric potential differences between two arbitrary locations, occurs in the early phase of the discharge period and the blow-out, whose major factor is discharge current, occurs in the latter phase of the discharge period respectively. The behavior of spark channel, current and voltage of the secondary circuit and the ignition limit by increase of EGR rate agreed with a measurement data of a spark ignition process in a chamber.

A topology optimization method is proposed for the design of shallow-flow channels based on quasi-three-dimensional flow models of laminar and turbulent flows. The models for laminar flow and turbulent flow are derived from the Navier-Stokes equations and the Reynolds-Averaged Navier-Stokes (RANS) equations, respectively, by integrating along the direction of channel thickness. The thickness is employed as the design variable in the topology optimization. The design variables are updated using a time-dependent diffusion equation with a design sensitivity which is calculated by a discrete adjoint approach. Numerical examples for minimizing dissipation energy or variance of flow velocity magnitude using the topology optimization demonstrates that the proposed method is capable of finding optimal solutions that satisfy the KKT conditions. In the former example, the design domain was clearly divided into domains where the thickness was either near the upper limit or near the lower limit. However, in the latter example, the thickness was at an intermediate level in almost the whole the design domain. The distribution of the thickness varied depending on the Reynolds number in both examples.

A direct numerical simulation of compressible turbulent channel flow at low Mach number(Ma = 0.3) subject to strong temperature gradient is conducted. The wall temperature ratios which are defined by a temperature on the upper wall divided by that on the lower one, are set to 2 and 3. It is shown that the flow of high temperature wall side is laminarized as increasing the wall temperature ratio. Moreover, compressible and dilatational motions are induced on the low and high temperature wall sides, respectively. The identity of friction coefficient originally derived by Fukagata et al.(2002) (so-called ‘FIK identity’) is extended to those of friction coefficient and Nusselt number on the compressible channel flow. It is revealed that the viscous variation component attains to 16 % of the friction coefficient on the case of high temperature ratio. Furthermore, the pressure work component occupies about 20 % of the overall Nusselt number for the all cases. The pressure work has a redistributive effect between mean kinetic energy and internal energy. The energy transfer from mean kinetic energy to internal energy appears on the low temperature side whereas that toward the opposite direction on the high temperature side at the present low Mach number flow.

A poppet type nozzle which promotes atomization by utilizing cavitation phenomena was designed to clarify the atomization mechanism. The configuration of the nozzle which has two orifices and an intermediate cavity between the orifices was decided by numerical analysis on the cavitating flow in the nozzle. In this report, atomization mechanism and spray characteristics of the nozzle are described. The following results are obtained: (1) The net-like fuel liquid sheet is observed closed to orifice. The net-like structure with cavitation bubbles is collapsed to droplets via ligaments. (2) The present nozzle has the features of generating smaller drop size and lower spray tip penetration than the conventional poppet type nozzle.

A liquid fuel injection nozzle with a novel fuel atomization strategy which actively utilizes the effect of collapse of cavitation bubbles in a liquid jet is proposed. The nozzle comprises two orifices and an intermediate cavity between the orifices. The role of the upstream first orifice is to generate the cavitation bubbles at both inlet and exit of the orifice. The intermediate cavity is placed to hold the cavitation bubble within it. The role of the downstream second orifice is to mix the cavitation bubbles and liquid fuel, and to inject the mixture into atmosphere. Each size and relative positions of the first orifice, the intermediate cavity and the second orifice were studied using multi-phase computational fluid dynamics. Finally, a poppet type nozzle whose void fraction at the nozzle exit is 0.5 was designed.

Large eddy simulation (LES) using a mixed-time-scale (MTS) SGS model is applied to the intake flows in simplified internal combustion engine geometry. A modified colocated grid system is employed so as to perform calculations stably with a central difference scheme for convection terms. The results are compared with corresponding experimental data and the computational results obtained by using the low-Reynolds-number linear κ-ε and cubic nonlinear κ-ε-A_2 turbulence models. The LES results show the best agreement with experimental data in three computational cases, not only in the mean velocity profiles but also in the profiles of turbulent energy. These results suggest that LES using the MTS SGS model is an effective method for accurately predicting the performance of a combustion engine involving turbulent diffusion of spray and combustion flame. In addition, the results are compared with those obtained by a quasi-direct simulation employing the QUICK scheme for the convection term and no turbulence model. It is clarified that the numerical viscosity from the QUICK scheme has a great influence on the computational results and decreases the prediction accuracy.

This paper presents discussions on predicting turbulent flow in a simplified internal combustion engine geometry consisting of an intake port, valve and cylinder. The low-Reynolds-number linear k-ε, cubic nonlinear k-ε and k-ε-A₂ turbulence models are applied to the flow field computation and evaluated against experimental data. The results suggest that the cubic nonlinear eddy viscosity modelling of turbulence is very successful for predicting mass flow rates and in-cylinder velocity distributions.

This paper presents discussions on predicting turbulence and heat transfer in square sectioned 180° U-bend ducts whose curvature ratios are Rc/D=0.65 and 3.357. The low-Reynolds-number cubic nonlinear κ-ε and κ-ε-A_2 turbulence models are applied for the flow field computation. In the thermal field computations, the higher order gradient diffusion heat flux model is applied in comparison with the standard models. The results suggest that the cubic nonlinear eddy viscosity and the higher order gradient diffusion heat flux models are very successful for predicting such three dimensional turbulence and heat transfer fields while some more modification in the models may be needed to capture heat transfer in the strongly curved case.

A theoretical model for a deforming single droplet with aerodynamic force is presented. The droplet shape is assumed to be spheroidal. A nonlinear oscillating equation is derived from energy conservation in a droplet. The present (OSD) model is compared with other droplet deformation models, Taylor Analogy Breakup (TAB), Droplet Deformation Breakup (DDB) and Synthesized Spheroidal Particle (SSP) models. It is found that the linearized OSD model is reduced to the TAB model. Numerical studies show that only present model and TAB model are compatible with Lamb's fundamental oscillating mode for a liquid droplet. On a droplet deformation at the initial stage of aerodynamic breakup, the calculation by the OSD model and TAB model agree well with the measurement.

Preconditioned iterative methods in an implicit unstructured grid solver of compressible flows are compared. The three-dimensional Euler and Navier-Stokes equations are discretized by the edge-based finite volume method. The artificial dissipation scheme for the inviscid numerical fluxes is developed for the unstructured grid from Jameson's scheme for the structured grid. The following methods are compared to the supersonic and transonic inviscid flows and subsonic viscous laminar and turbulent flows : the central difference methods with the scalar and matrix dissipation scheme and Roe's upwind method with the MUSCL sheme for the inviscid fluxes ; the scalar and matrix dissipation method for the left-hand side operator in the implicit scheme ; the ILU, DILU and SGS factorizations for the preconditioner ; and GMRES and Bi-CGSTAB methods for the linear system solver. The results show that a combination of the matrix dissipation in the inviscid fluxes and the scalar dissipation in the left-hand side works better than the other shemes. In the preconditioned iterative methods, a combination of DILU or SGS and the Bi-CGSTAB is recommended for the aspect of required memory.

A calculation method for steady compressible flows using a solution adaptive unstructured mesh is proposed and it's applicability is demonstrated for two-dimensional and axisymmetric engine intake flows. The mesh is generated by the rule-based Delaunay triangulation method. The governing equations are discretized by the finite volume method. The second order accurate Roe's upwind scheme is used for the inviscid flux calculation. The viscous fluxes are calculated by the transformation to the local general coordinates. The utility of the adaptive unstructured mesh method is shown in the 2D engine intake flow calculations. The present method using a low Reynolds number k-c model is valid for the velocity profile in the turbulent boundary layer. The calculated discharge coefficients with valve lifts in an axisymmetric engine agree well with the measurements.

圧縮性流れの解法において、非構造格子上で有限体積法による離散化と2次精度の空間差分スキームを用いる際、時間解法には前処理付きBi-CGSTAB法が当時一般的なRunge-Kutta法やガウスザイデル法に比べて定常解への収束が早くかつ安定であることを示した。

Numerical prediction of a steady intake airflow was performed with an unstructured grid. Changing surface rougheness and configuration of an intakeport, airflow rate and velocity vectors were calculated. Comparing the calculated results with the experimental ones by LDV measurements, the accuracy was evaluated. It was found that airflow rate can be predicted with the accuracy of 5 percent, and the predicted velocities, especially in the intakeport, agreed with the measured ones. The predicted airjets at the valve annular passage were slower than the experimental ones. The application of the standard k-ε turbulence model to the airjet at the valve passage was found to be an essential factor for the discrepancy.

This paper is devoted to the analyses and optimization of simple and sophisticated cycles, particularly for various gas turbine engines and aero-engines (including the scramjet engine) to achieve maximum performance. The optimization of such criteria as thermal efficiency, specific output, and total performance for gas turbine engines, and overall efficiency, nondimensional thrust, and specific impulse for aero-engines has been performed by the optimization procedure with the multiplier method. Comparison of results with analytical solutions establishes the validity of the optimization procedure.

The unstructured upwind method is proposed to solve compressible flow with complicated boundaries or an arbitrarily shaped mesh. The flux difference splitting (FDS) scheme and flux vector splitting (FVS) scheme, which are extended to the unstructured grid system, are compared. The Euler equation is discretized by the control volume method. All conserved variables are stored at the cell center. The MUSCL approach for the cell-centered unstructured grid is proposed to obtain higher-order accuracy. A three-stage Runge-Kutta method is used for the time stepping scheme. The present calculation methods are applied to two-dimensional front-facing step flows. The results demonstrate that the present MUSCL approach is effective to obtain high resolution, and the triangular mesh solver is comparable to the uniform orthogonal mesh solver. FDS gives slightly better resolution than FVS.

Fuel mixture formation processes were predicted using gasoline droplet data obtained by LDA measurements. In the intake period, air and fuel inlet boundary conditions were given at the valve annular passage. The last calculated result of the intake was used as the initial condition of the compression period. The effects of combustion chamber configuration on the mixture homogeneity were investigated in a lean-burn engine. Four types of piston cavities were investigated: pancake, Heron, off-cylinder-axis Heron type, and complicated configuration chamber. The complicated one improves homogeneity since the flow produces two inclined vortices in the cylinder, which play a great role in promoting convective mixing.

新開発の小型直噴ディーゼルエンジンにおける燃焼室内の流れと燃料噴霧衝突挙動を3次元シミュレーションで解析し、その燃焼室形状が乱れを強め、混合気形成を促進していることがわかった。

Diesel soot formation processes were simulated, using several models previously proposed for soot formation and oxidation. The models were assessed and improved by comparing the results with the experiments. Consequently, the model which represented qualitatively the total mass history, the formation position and the spacial distribution of soot was established. The outline and characteristics of the model are as follows. (1) The soot formation model consists of both the Tesner and Farmer models. The soot oxidation model consists of both the Nagle and Magnussen models. The model giving a smaller rate was selected for both models (formation and oxidation). (2) Some of the coefficients in the Tesner model were corrected depending on the in-cylindr volume change. (3) The Farmer model was used only to cover the deficiency of the Tesner model which must be improved due to tempareture dependence.

Many works on gas turbine engines contain much information concerning the performance of various engine schemes both in simpler and in more complicated forms. Such information enables the design performance to be estimated, but there appears to have been very little works on the optimum performance under non-design conditions. In this paper, we attempt to include a comparison of part load performance characteristics and to optimize them. The multiplier method is adopted as an optimization procedure. As a result of this work, it is possible to indicate the optimization technique by which the comparable part load performance of different gas turbine schemes may be assessed.

This paper is devoted to the analyses and optimization of simple and sophisticated cycles, particularly for various gas turbine engines and aero-engines, to achieve maximum efficiency or maximum specific output. The optimization of such criteria as thermal efficiency, specific output and total performance for gas turbine engines, and overall efficiency and non-dimensional thrust for aero-engines have been performed by the multiplier method. The comparisons of results with analytical solutions establishes the validity of the optimization technique.