Real-time onboard fuel classification in internal combustion engines

Abstract

The present invention provides a method and system for detecting the type of fuel used in an internal combustion engine. [Solution] A method for identifying an unknown fuel used by the internal combustion engine of a vehicle, comprising preparing a set of sensor models, each representing onboard sensor measurements for different known fuels. The sensor models correspond to various sensors mounted on the vehicle, such as a crank sensor, NOx sensor, soot sensor, and exhaust temperature sensor. While the vehicle is operating, measurement data is acquired from two or more sensors in the sensor group and supplied to an onboard fuel identification process. This fuel identification process applies statistical analysis to identify the fuel as one of the known fuels.

Description

Technical Field

[0001] ·Cross-reference to Related Applications This patent application claims priority as of the filing date of U.S. Provisional Patent Application No. 63 / 694,170, filed on September 12, 2024, entitled "Real Tuning - Real Time On-Vehicle Fuel Classification in a Reciprocating Internal Combustion Engine."

Background Art

[0002] Internal combustion engines have been using fossil fuel-derived fuels for decades. Conventionally, the main fuels were diesel and gasoline. However, energy companies are developing new fuels to replace conventional fuels for the purpose of improving fuel efficiency and achieving more sustainable production. Fuels such as renewable diesel and biodiesel are examples of alternative fuels. When including these fuels and their mixtures, as well as other sustainable fuels expected to be utilized in the future (e.g., methanol, ethanol, ammonia, etc.), the fuel options can become complex due to factors such as region, cost, and engine compatibility. Identifying the fuel is essential for determining the compatibility between the fuel and the engine.

Brief Description of the Drawings

[0003] [Figure 1] Shows multiple fuel characteristics and how they vary among four different fuels. [Figure 2] Shows four types of sensors that can be used to identify (classify) the fuel. [Figure 3] Shows four types of sensors that can be used to identify (classify) the fuel. [Figure 4] Shows four types of sensors that can be used to identify (classify) the fuel. [Figure 5] Shows four types of sensors that can be used to identify (classify) the fuel. [Figure 6]This document demonstrates how to use a fuel matrix, which is used to represent fuels and fuel characteristics in statistical models. [Figure 7] This document describes a method for identifying fuel using two sensor parameters (ignition sensor and fuel sensor). [Figure 8] This shows a vehicle equipped with the on-board fuel identification system according to the present invention. [Modes for carrying out the invention]

[0004] The following description concerns a method and system for detecting the type of fuel being used in an operating internal combustion engine. When an unknown fuel is supplied, this method collects measurement data from onboard sensors. These sensors are installed around the engine or within the fuel or exhaust flow and are used to detect the characteristics of the fuel. Once multiple characteristics of the fuel or the effects of its combustion are detected and recorded, this data is combined with the ECU input settings and compared against the expected output obtained from different fuels. Based on the degree of agreement with the expected output, the fuel can be classified into a known category. Identifying the fuel provides a significant opportunity to readjust the engine for optimal performance, reduce emissions and fuel consumption, and thereby improve the environmental and economic aspects of vehicle operation.

[0005] Fuel is identified, or "detected," by one or more sensors placed around the engine or within the fuel flow. These sensors play a role in detecting some of the fuel's characteristics. For example, flex fuel sensors, originally developed for ethanol / gasoline applications, have been shown to exhibit good sensitivity to changes in biodiesel content. Other examples include production accelerometers (knock sensors) and crankshaft sensors, which can provide analysis and input data regarding combustion performance, such as ignition onset. These sensors, which reveal fuel characteristics such as oxygen content, cetane number, and calorific value, form the basis for methods of classifying fuels. A combination of these detected fuel characteristics is used in the classification process to classify the fuel. In short, fuels can be classified using multiple sensors (multi-sensor system).

[0006] Figure 1 shows the characteristics of several fuels and how they change among four different fuels. These four fuels are diesel fuel, pure biodiesel (B100), pure renewable diesel (RD), and Fuel X. Fuel X is an unknown fuel. Diesel fuel is a conventional fuel, while the other fuels can be described as "alternative fuels."

[0007] Various fuel properties include density, cetane number (CN), ignition delay, degree of saturation, hydrogen-to-carbon ratio (H / C), oxygen content (O / C), stoichiometric air-fuel ratio (AFR-stoic), proportion of benzene ring-containing compounds, such as benzene, toluene, and xylene (aromatic compounds), fatty acid methyl ester (FAME) content, distillation rate (Distl-0, Distl-50, Distl-100), and lower heating value (LHV). Figure 1 shows that when fuel properties are measured directly or indirectly for a particular fuel, their variations are useful in identifying the fuel.

[0008] Figures 2-5 show four types of sensors that can be used to identify (classify) fuels. Additional sensors can also be used to measure or estimate additional fuel characteristics, as described below. As shown in Figure 1, each fuel has different characteristics, so a particular sensor will output different measurements depending on the type of fuel.

[0009] In Figures 2-5, the sensors are a fuel sensor (Figure 2), a combustion sensor (Figure 3), a soot sensor (Figure 4), and a NOx sensor (Figure 5). The combustion sensor can be an acceleration-type knock sensor or a crank sensor, or a combination of both. Both sensors have been demonstrated to have the ability to describe the characteristics of combustion events. Some of these sensors are commonly installed in vehicles using internal combustion engines. The fuel sensor is used with flex fuels and measures the conductivity of the fuel. Each sensor provides a different measurement output depending on the fuel used for combustion.

[0010] The triangle represents the relative amounts of three types of fuel: renewable diesel, diesel, and biodiesel. For each fuel, its purity is indicated at the vertices of the triangle. As shown in the diagram, the measured values ​​detected by each sensor differ depending on the fuel. For example, in the case of diesel fuel, the fuel sensor reading is lower than that of biodiesel, while the soot sensor reading is higher compared to the other fuels. Renewable diesel can be distinguished from biodiesel and diesel by the soot and NOx sensors.

[0011] Figure 6 shows how a fuel matrix can be used to generate boundaries for fuel properties for statistical modeling. The three fuels (X1, X2, and X3) represent pure fuels (diesel, biodiesel, and renewable diesel), respectively. The midpoints represent mixtures of these fuels. This six-point grid design represents a model with three fuels and two mixture components. Other models can also be used for additional fuels and components.

[0012] These models can be used as the basis for statistical analysis performed by applying sensor measurements to the model. For example, this model functions as a matrix in a neural network structure, with each sensor becoming an input channel to the neural network. In a neural network, the matrix is ​​a grid of numbers (especially weights), which represent the strength of connections between neurons and are used to perform computations that process information, transforming input data through layers to produce outputs. The matrix organizes the network's parameters and data into a structured form, enabling efficient numerical computation that is fundamental for the neural network to learn and predict. Other machine learning processes, such as support vector machines, are similarly applicable. Generally, if each data point belongs to a certain class, the goal is to determine which class a new data point belongs to.

[0013] Figure 7 shows the output of sensor measurements applied to a support vector process, illustrating how two sensor parameters (ignition and fuel sensors) can be used for fuel identification. Regarding ignition detection, ID HRR Zero refers to the point where the ignition position is defined by the rate of heat recovery (HRR) being above zero. This sensor is used to indicate the start of combustion. Nine different fuels were detected using a support vector machine process under different speed-load operating conditions. The confidence level of the two-sensor classification can be improved by reconstructing the data analysis and selecting specific data from specific engine operating conditions.

[0014] To identify an unknown fuel being consumed by a particular vehicle, this method begins with selecting available parameters from sensors on board the vehicle. The most suitable parameters are expected to be measurements from fuel sensors, crank sensors, NOx sensors, soot sensors, and exhaust temperature sensors. However, this method is not limited to these sensors, and other sensors such as rail pressure sensors and injection timing sensors may be used. In some embodiments, the vehicle's engine control unit is already configured to collect measurement data, which can then be fed into the classification process.

[0015] The fuel identification method attempts to "detect" the fuel using available onboard sensor measurements. These measurement types are established or calibrated using previous fuel and engine measurements and operating conditions. Based on the sensor measurement data, specific fuels can be classified into their fuel type by using the statistical analyses described here. Various statistical analysis techniques can be used to apply the sensor measurement data to predetermined models of known fuel characteristics and combine the results.

[0016] This method can be combined with machine learning to support the fuel classification process and reduce the number of channels to lessen the data processing load. For example, machine learning can more efficiently identify any available channels (sensor or engine control unit output).

[0017] One of the features of this method is the selection of a specific technique for analyzing the measurement data. This selection can be made by the engine driver or automatically based on the real-time engine status. The three available techniques are as follows:

[0018] (1) Look-up calibration method: A method that uses fixed or referenced engine operating conditions.

[0019] (2) Relative calibration method: A method that uses a series of engine operating conditions by means of real-time measurements of intentionally controlled or actual engine / vehicle operation.

[0020] (3) ECM-based method: Similar to the lookup method and the relative method, but limited to channels from the engine control management (ECM) of the vehicle. By this method, vehicle manufacturers can implement fuel identification methods using their specific on-board sensors.

[0021] For learning purposes, if the identification process cannot identify the fuel under a series of engine operating conditions and available in-vehicle parameters, report the uncertainty and register the fuel during operation as a new fuel. Eventually, this process will learn to identify specific operating conditions and in-vehicle parameters that improve the identification results.

[0022] Once the fuel is identified, this information can be further used for other purposes. Examples include engine and emission control, and durability improvement. Engine and emission control aims to optimize engine and emission performance after the engine has learned sufficiently about the fuel. Regarding durability, the vehicle will have a history of the fuel types used over a certain period. Different fuels can affect various engine components such as injectors, pistons, and aftertreatment devices. By creating records that build a time history in combination with the known sensitivities of engine components, the remaining life of the components or the time until maintenance is required can be estimated.

[0023] Figure 8 shows a vehicle 90 equipped with an on-board fuel identification system according to the present invention. The vehicle 90 includes a plurality of on-board sensors described herein. These may include a crank sensor 92, a NOx sensor 93, a soot sensor 94, and an exhaust temperature sensor 95, and may also include other sensors not explicitly shown in Figure 8. Certain fuel identification systems according to the present invention can use some combination of these sensors. Usually, at least three sensors are used.

[0024] Measurement data from sensors 92, 93, 94, and 95 is sent to an on-board fuel identification process 97, which, as described above, analyzes and correlates the measurement data to determine the fuel that most closely matches the fuel characteristics indicated by the measurement data. The identification process is executed entirely on-board and does not require external input or communication.

[0025] The identification system can include an operator interface 99 to which the identified fuel type is reported. The fuel type may be reported to other control systems, such as an emissions control system, and may be used as an input to optimize emissions control for that fuel.

Claims

1. A method for identifying an unknown fuel used in the internal combustion engine of a vehicle, The vehicle has multiple onboard sensors, A step of creating a sensor model set consisting of models representing onboard sensor measurements for different known fuels, wherein the sensor model corresponds to a sensor selected from a group of sensors including a crank sensor, a NOx sensor, a soot sensor, and an exhaust temperature sensor. A step of acquiring measurement data from two or more sensors of the sensor group while the vehicle is in operation, A step of supplying the measurement data to an onboard fuel identification process that applies statistical analysis to the measurement data, A step of using the fuel identification process to identify the unknown fuel as one of the known different fuels, A method that includes this.

2. The vehicle has an engine control unit, The method according to claim 1, further comprising the step of supplying data from the engine control unit to the fuel identification process for use in fuel identification.

3. The method according to claim 1, wherein the statistical analysis is performed using a neural network.

4. The method according to claim 1, wherein the statistical analysis is performed using support vector processing.

5. The method according to claim 1, further comprising the step of supplying the results of the fuel identification process to the vehicle's emission control system.

6. The method according to claim 1, further comprising the step of determining engine operating conditions under which the step of acquiring the measurement data is performed.

7. The method according to claim 1, further comprising the step of using a lookup table to determine the engine operating conditions under which the step of acquiring the measurement data is performed.

8. An onboard system for identifying an unknown fuel used by the internal combustion engine of a vehicle, A set of two or more onboard sensors selected from a group of sensors including a crank sensor, NOx sensor, soot sensor, and exhaust temperature sensor, A sensor model set consisting of each model representing onboard sensor measurements for different known fuels, An onboard fuel identification process that acquires measurement data from two or more sensors while the vehicle is in operation, applies statistical analysis to the measurement data, and identifies the unknown fuel as one of the known different fuels, A system equipped with these features.

9. The vehicle has an engine control unit, The fuel identification process receives data used for fuel identification from the engine control unit. The system according to claim 8.

10. The system according to claim 1, wherein the fuel identification process is implemented by a neural network.

11. The system according to claim 1, wherein the fuel identification process is implemented by support vector processing.

12. The system according to claim 1, further comprising a stored lookup table for determining the engine operating conditions under which the step of acquiring the measurement data is performed.

Images