Patent Application
20090276143
The present invention relates to a method of estimating individual average cylinder torque values of internal combustion engines; more particularly, to methods for optimizing operating parameters such as combustion mixtures and spark timing in such engines; and most particularly, to an improved method for inferentially determining Indicated Mean Effective Pressure (IMEP) for individual cylinders by calculation from instantaneous changes in crankshaft acceleration, and to a method for engine control employing improved IMEP calculation.
Knowledge of individual cylinder values of Indicated Mean Effective Pressure (IMEP) is known in the prior art as a powerful tool for evaluating and correcting poor combustion in an internal combustion engine. By definition, IMEP, in kiloPascals, is defined as the ratio of the indicated work in Newton meters (W) divided by the swept volume per cylinder (V) in liters:
IMEP=W/V (Equation 0)
IMEP is an accepted standard method for measuring combustion in internal combustion engines. The information is valuable in indicating combustion quality and is used extensively in the prior engine arts in engine dynamometer work to characterize and quantify acceptable and unacceptable combustion performance. IMEP is known to be used to determine the limits of engine dilution (e.g., exhaust gas recirculation, camshaft phasing), spark advance angle, and rich/lean limits to engine fueling.
Although IMEP is a valuable parameter for combustion development, its use in real time engine controls has been limited in the prior art in general because its determination has required expensive and non-durable combustion analysis equipment, and because the prior art methods of measurement have been engine-intrusive (e.g., combustion pressure sensors in the engine heads or spark plugs). Other known methods of combustion quality measurement, such as Ion Sense technology, require expensive hardware upgrades and have not been generally available. Off-board rack-type analysis equipment is bulky, expensive, and non-portable. Thus, engine control using IMEP has been largely a laboratory phenomenon rather than being useful day-to-day in an operating vehicle.
Less than ideal combustion performance can arise from a variety of sources including: engine component design limitations; variations in fuel properties in the field; aged engine components; and manufacturing tolerances of engine subassemblies and components. Manufacturing tolerances for valve train intake and exhaust ports and valves; fouled plugs, ports or injectors; and/or design trade offs affecting fuel, purge, PCV and EGR distribution, can all contribute to degraded combustion quality. Contributors to degraded combustion can affect performance of individual cylinders, or of the engine as a whole.
An individual cylinder torque estimator, used in conjunction with an appropriate engine algorithm in real time control, can mitigate the sources of cylinder-to-cylinder combustion variability, ultimately improving, for example, idle quality, NVH due to torque imbalance, peak power, and cold start emissions.
Prior art methods which attempt to estimate individual cylinder torque values focus on assessing combustion performance based upon a single cycle or single-cylinder event. When attempting to evaluate combustion quality, quantifying only single-cylinder events can be misleading due to cyclic variability of fuel transients in the ports, or to unburned fuel residuals which remain after partial burns or misfires. Incomplete mixing and burn due to in-cylinder turbulence which is unrepresentative of overall combustion behavior may also result in poor combustion on a single cylinder event basis.
Using a statistical evaluation of IMEP as a metric to gauge combustion quality is therefore advantageous and superior. The Coefficient of Variance of IMEP (COVIMEP) is a statistical evaluation of combustion quality. COVIMEP is a way of characterizing engine combustion that is well accepted across the automotive industry. As such, it provides an objective and standard means for quantifying combustion performance. Because of its ready availability, correlation to other engine performance characteristics, for example, brake-specific emissions values, is also possible.
In addition to the lack of a good metric for evaluating combustion quality that can be used in real time control, prior art methods have also required additional development effort to calibrate their models. While such development effort is of value for improving the model's accuracy, it provides limited additional benefit beyond the express purpose of individual cylinder torque estimation.
Further, depending on complexity, prior art methods can be computationally expensive which limits their use, especially at high engine speeds when the chronometric impact of calculations which must be performed in the period between cylinder firing events, i.e. calculations for individual cylinder torque estimation, is greatest.
What is needed in the art is a method for providing cylinder IMEP information, and an associated control metric, that does not require additional engine hardware or significant development effort and computational expense, while at the same time providing good utility for real time engine control.
It is a principal object of the present invention to provide realtime IMEP and COVIMEP for each cylinder of a multi-cylinder engine, and the engine as a whole, from calculated crankshaft velocities and accelerations, and from a pre-existing algorithm which requires little or no additional calibration effort for the present purpose. Additionally, the present invention includes calculations which are simplified and optimized for computational efficiency and speed.
Briefly described, the current invention decouples the calculation of the transient, inter-cycle component of indicated torque from its quasi-steady, multi-cycle component. A torque balance, or conservation of kinetic energy of rotating and reciprocating engine components, is used to estimate the transient component, and a pre-existing, cycle-averaged engine indicated torque algorithm is used to calculate the quasi-steady component.
Of importance to the present discussion and method are terms in common use in the art for the tracking and timing of cylinder events within a cycle of a given multi-cylinder internal combustion engine. These terms are crankshaft time stamp, and cylinder reference event period.
A crankshaft time stamp is the time at which a specific crankshaft position is sensed on a toothed wheel attached to the crankshaft and is typically accomplished through a microprocessor connected to a variable reluctance device (VRD). The VRD senses a voltage change associated with a specific tooth passing the VRD's fixed crank angle location. As the engine rotates, the voltage change is marked in time (stamped), via the microprocessor's internal clock.
In general, the microprocessor acquires crank shaft time stamps for specific teeth located at predetermined crank angle locations. Knowing the crank angle location of the teeth, and the period of time between any two teeth (the difference in the time stamps), allows for the calculation of the average engine velocity between the teeth. These time periods typically are also corrected for tooth errors that result from manufacturing tolerances of the high data rate wheel. Tooth error correction is performed via an algorithm learning process that takes place during fuel cut-off overrun engine condition(s).
When the two teeth of interest are located equidistant in crank angle from each of two consecutive cylinders' top dead center locations, the period of time is referred to as a cylinder reference event period. The ratio of the difference in crank angle between two consecutive teeth divided by the difference in their time stamps approaches an instantaneous value of engine speed for wheels with a large number of teeth (i.e. as in a high data rate wheel), for example, 58 teeth.
As previously noted, a total indicated engine torque estimate comprises two components, transient and quasi-steady. In the present invention, the transient component of indicated torque is derived from variations in average crankshaft velocity. The quasi-steady component is determined from a quasi-steady indicated torque model.
Changes in crankshaft velocity from one cylinder to the next are equated to changes in engine kinetic energy within an engine cycle (inter-cycle). Changes in inter-cycle engine kinetic energy are attributed to changes in energy-averaged cylinder torque values. Referring to Equation 1 above, by definition, energy averaged changes in cylinder torque are equivalent to changes in cylinder IMEP value.
Conversely, changes in engine kinetic energy which occur over multiple engine cycles are accounted for through cycle averaged indicated torque as estimated by the quasi-steady model. A state-space approach is used to sum the changes in kinetic energy over time, yielding energy-average cylinder torque or IMEP values.
During initialization, the quasi-steady component of cylinder indicated torque is used to “seed” the total indicated torque estimate for the first engine cycle. After initialization, the quasi-steady indicated engine torque is used to continuously re-center the total indicated engine torque values. The term “seed” is used to denote each initialization of the algorithm as described in detail below.
The quasi-steady indicated engine torque estimate represents a cycle-averaged torque value. Knowing the average torque for the first engine cycle and the estimated torque changes for each cylinder in the cycle allows for the determination (or initialization) of each cylinder's torque for the first engine cycle. In a similar way, for all subsequent engine cycles after the first, the quasi-steady indicated torque value is used to re-center the average engine torque calculated by the model by adding or subtracting a percentage of the difference between the model's estimated engine torque and the quasi-steady value.
A detailed description of the quasi-steady component of cylinder torque is provided below in an exemplary illustration of a method commonly employed in the prior art as part of an automotive engine control scheme used in the estimation of indicated engine torque. This calculation is re-used in the present invention to supplement the calculations of individual cylinder torque values. In and of itself, this quasi-steady torque estimation is in common prior art use in microprocessor-based engine control and as such does not represent anything novel; however, prior art methods which utilize a time-based approach to calculate transient torque do not avail themselves of this historically well-tested, parameterized, and accurate means for estimating the quasi-steady torque component of an individual cylinder torque model.
The present invention is useful in control of spark-ignited engines and combustion-ignited engines.
Novelties of the present invention include:
1. accurate, by-cylinder/engine IMEP, and by-cylinder/engine COVIMEP calculations using only tooth error-corrected engine speed and quasi-steady engine indicated torque algorithm to both “seed” and re-center the total cylinder torque estimate, and a commercially available engine control unit which eliminates the expense and intrusiveness of direct IMEP measurements with pressure sensors;
2. state space analytical technique which significantly reduces the level of calibration effort needed to parameterize the model. The current invention utilizes readily-available steady state engine dynamometer (“mapping”) data for the determination of the quasi-steady component of cylinder torque. For the most part, the calibration parameters are reduced to physical constants of the engine and readily available steady state engine dynamometer (“mapping”) data. Calibration effort expended to refine torque estimates can benefit other users of the torque data. Since this quasi-steady indicated torque estimate is generally available and in use in prior art engine controls, the present method requires no additional calibration or engine parameterization for this part of the solution;
3. coarse discretization of the instantaneous engine speeds in order to reduce the computational requirements to a level acceptable for real time processing in commercially available microcontrollers. Care must be taken in the choice of crank locations and difference equations used to ensure they are accurate enough for the intended application. Simple finite difference formulas, plus existing hardware and designs currently in use for misfire detection, are leveraged for this purpose. Unlike prior art methods requiring the acquisition of multiple crankshaft time stamps for each cylinder reference event period (refer to U.S. Pat. No. 6,029,109, which specifies the use of four such periods) plus associated calculations, the current invention requires only one time stamp per cylinder and simple numerical difference formulas to represent changes in indicated torque values;
4. use of the Coefficient of Variance (COVIMEP) metric, and COVIMEP calculations which are optimized to minimize chronometric impacts and have good utility for real time engine control; and
5. “seeding” of the indicated cylinder torque estimate and re-centering around a quasi-steady engine torque value.
The present invention will now be described, by way of example, with reference to the accompanying drawings, in which:
The exemplification set out herein illustrates a presently-preferred embodiment of the invention, and such exemplification is not to be construed as limiting the scope of the invention in any manner.
The transient inter-cycle indicated torque component may be determined in two ways: either indirectly, through calculation of engine kinetic energy change via the difference in average torque from one cylinder event to the next multiplied by the crank angle over which average torque difference acts, or directly, through changes in measured instantaneous crank shaft velocities from one cylinder event to the next. For illustration purposes, the development of average torque changes (indirect method) will be described in detail here.
Referring to
The resulting torque balance for the transient component of engine torque is mathematically shown in Equation (1). The difference between the indicated torque and the sum of friction and load torque is what's available to accelerate the engine (IEωE). As shown in Equation (2), torques are divided into transient or alternating (T(t)) and cycle averaged values (
Substituting Equations (2) and (3) into Equation (1) and discretizing over the current and previous cylinder events (j and j−1) yields Equation (4). Equation (4) shows that the change in average indicated torque from the previous to current cylinder event is equal to the difference in engine acceleration multiplied by the average engine inertia. Equation (4) can be written in the form of a change in kinetic energy (ΔK.E.) by multiplying by the crank range (ΔΘ) over which the torque difference is assumed to act (Equation 5). Since the change in kinetic energy is assumed to result from gas torque above or below the cycle averaged level, from the definition of IMEP the change in kinetic energy is also represented by the difference in IMEP times cylinder displacement.
Quasi-steady indicated engine torque 16 is determined from measured engine air and fuel flow [1]. This is typically done using a speed density algorithm utilizing sensed manifold absolute pressure or mass air flow meter, for measuring air flow 18, plus characterizing injector flow and monitoring injector pulse width for estimating fuel flow.
Engine air fuel ratio 24, is determined from the ratio of these two values. Total delivered spark advance is also monitored 20. Engine speed 22, EGR, and operating temperatures and steady state engine performance maps describing either brake or indicated engine torque 29 are also used as input to the quasi-steady engine torque model. Engine or component performance maps may also be used to describe mechanical friction 28 and pumping 30 losses as well as accessory torque requirements (not shown in the figure). It is an important advantage of the present invention that all of these data inputs are already present in modern automotive engine control; thus, no additional parameterization or apparatus is required to obtain the quasi-steady indicated engine torque estimate 16.
The quasi-steady indicated engine torque 16 is used to both “seed” and continuously re-center [6] the cylinder IMEP estimator around the current cycle averaged value 34.
Instantaneous or average values of engine speed are determined from a high data rate crankshaft target wheel 36 and variable reluctance sensor 38 in known fashion [2]. The delta time values are corrected for tooth errors 40 [3]. These tooth errors result from manufacturing tolerances of target wheel 36. Instantaneous or average engine speed values 42 are used in a numerical difference formula to estimate engine angular acceleration 42 [4]. Changes in engine angular acceleration are then used to calculate changes in engine torque (and kinetic energy) from one cylinder/ref event to the next 44 [5]. Using the seed value 16 of estimated engine torque from [1], subsequent levels of torque needed to accelerate the engine at each ref event are evaluated 46.
From cylinder IMEP levels 14, corresponding values of the Coefficient of Variance of IMEP (COVIMEP} are determined for each cylinder and for the engine as a whole. After individual cylinder torque and IMEP values are determined, a numerically optimized technique is used to evaluate COVIMEP. The present method utilizes a buffer of previously calculated cylinder IMEP values and a calculation which tracks the sum and the sum of squares of the buffer. The optimization reduces the computational requirements of calculating COVIMEP at each cylinder event through a reformulation of the coefficient of variance (COV) equation. This reformulation results in a computational savings of N−1 additions and subtractions (where “N” is the COV sample size), when compared to the traditional method of COV calculation.
A computationally efficient calculation for the Coefficient of Variance in accordance with the present invention is a follows:
The Coefficient of Variance is equal to the standard deviation (σ) over the mean (
COV=σ/
The standard deviation is equal to the square root of the sum of the square of the difference between the mean and the individual values divided by the number of samples minus one:
Expanding the series and substituting for the mean
yields a more computationally efficient form of the equation for COV:
By storing the sequential individual sample values in a buffer and tracking the sum of the square and square of the average of the buffered values, the COV may be calculated in an efficient manner with no loss of accuracy. This method requires only one addition and one subtraction for each new value in the sample (adding and subtracting the newest and oldest values in the buffer, respectively, to and from their sums), compared to the prior art method of N additions and N subtractions in the traditional COV calculation. This results in a savings of (N−1) additions and subtractions.
Using a torque balance or kinetic energy formulation for cylinder torque is disclosed in the prior art in a number of patents (see, for example, U.S. Pat. Nos. 6,029,109 and 6,302,083). In these other patents, however, the same formulation is used as the primary means of calculating both the quasi steady and alternating components of cylinder torque. The method of the present invention is novel in that it employs the torque balance/kinetic energy to calculate only the alternating torque component. The quasi steady component is determined from the various measured engine quantities shown in
The use of steady-state engine mapping data in determining the quasi steady component of engine torque also has been disclosed in the prior art in at least one other patent (U.S. Pat. No. 6,223,120) and in SAE paper 2001-01-0990, but this prior art method solves for torque in the frequency domain, not the time domain. Thus, knowledge of the time the events occurred is lost and the result cannot be easily combined as a metric for use in control or to assess accuracy in real time. The level of computational effort required for this prior art method may also not yet facilitate real time computation in today's microprocessors. SAE paper 2001-01-0990 indicates only a “near real time implementation”.
Accuracy of estimating IMEP and COVIMEP in accordance with the present invention is shown in
Referring to
Referring to
While the invention has been described by reference to various specific embodiments, it should be understood that numerous changes may be made within the spirit and scope of the inventive concepts described. Accordingly, it is intended that the invention not be limited to the described embodiments, but will have full scope defined by the language of the following claims.
