Description:
Complete Specification

The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed.
Field of the invention
[0001] This invention relates to an AI assisted air-mass flow algorithm, and more specifically to the AI assisted air-mass flow algorithm for naturally aspirated internal combustion engines.

Background of the invention
[0002] US 2007073467 AA describes a method for controlling combustion in an internal combustion engine and predicting performance and emissions. This disclosure teaches a method of controlling a direct injection internal combustion engine and predicting the behavior of a direct injection internal combustion engine. An estimation of initial cylinder pressure, air flow and EGR flow (if applicable) is used to establish a system that provides engine behavior by integrating an injection module, combustion module and engine control module to provide data indicative of engine behavior such as brake torque and power, air flow, EGR flow, cylinder pressure, brake specific fuel consumption, start of combustion, heat release rate, turbo-charger speed and other variables. These values can then be used to adjust commanded variables such as start of injection, commanded pulse width, rail pressure to meet operator demand. Also the output data can be used as a tool to determine how a conceptualized engine design will behave. This is particularly useful for gaseous-fueled internal combustion engines where cylinder pressure influences behavior of injected gases in light of the fact that rail pressure and cylinder pressure are, generally of a similar magnitude.

Brief description of the accompanying drawing
[0003] Figure 1 illustrates an AI assisted air-mass flow algorithm for a naturally aspirated internal combustion engine in one embodiment of the invention.

Detailed description of the embodiments
[0004] Figure 1 illustrates an algorithm 100 for intake air-mass flow estimation of a naturally aspirated internal combustion engine. The algorithm 100 comprising the steps of inputting 110 a first set of data values comprising a rail pressure from a rail pressure sensor, an intake manifold air pressure value from an intake manifold air pressure sensor, an engine speed from an engine speed sensor, a high-pressure exhaust gas recirculation position feedback value from an exhaust gas recirculation position feedback sensor, a modelled lambda value from an oxygen sensor in an exhaust path, a modelled exhaust manifold upstream pressure using an oxygen sensor in an exhaust path. The algorithm further comprises inputting 120 a second set of data values comprising a ratio of fuel and oxygen data from a lambda sensor, and a pressure value from one of an exhaust manifold pressure sensor and an exhaust manifold pressure model, and inputting 130 a third set of data values comprising an actual air mass flow value from an intake air mass flow sensor. The algorithm further comprises determining 140 an intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values, inputting 150 the intake air mass flow values into a first neural network based on an actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor, and selecting 160 one intake air mass flow value from the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by means of the first neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value. In addition, the algorithm comprises inputting 170 the selected intake air mass flow value into a second neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor, determining 180 an error value of the intake air mass flow from the selected intake air mass flow value by means of the second neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value and the known value of error of the intake air mass flow, and algebraically summing 190 the selected intake air mass flow value to the error value of the intake air mass flow to obtain the corrected intake air mass flow value.

[0005] Figure 1 illustrates an algorithm 100 for intake air-mass flow estimation of a naturally aspirated internal combustion engine. The algorithm 100 comprises inputting 110 a first set of data values comprising a rail pressure from a rail pressure sensor to an electronic control unit. Therein, an intake manifold air pressure value from an intake manifold air pressure sensor, an engine speed from an engine speed sensor, a high-pressure exhaust gas recirculation position feedback value from an exhaust gas recirculation position feedback sensor, a modelled lambda value from an oxygen sensor in an exhaust path, a modelled exhaust manifold upstream pressure using an oxygen sensor in an exhaust path that is inputted into the electronic control unit. The algorithm further comprises inputting 120 a second set of data values comprising a ratio of fuel and oxygen data from a lambda sensor, and a pressure value from one of an exhaust manifold pressure sensor and an exhaust manifold pressure model to the engine control unit. Therein, a third set of data values comprising an actual air mass flow value from an intake air mass flow sensor is inputted 130 to the electronic control unit.

[0006] The algorithm 100 further comprises determining 140 the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by the electronic control unit. Once the intake air mass flow value is determined, the intake air mass flow value is inputted 150 into a first neural network based on an actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor. The electronic control unit therein selects 160 one intake air mass flow value from the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by means of the first neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value.

[0007] In addition, the algorithm 100 comprises inputting 170 the selected intake air mass flow value into a second neural network that runs in the electronic control unit based on the actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor. Therein, an error value of the intake air mass flow from the selected intake air mass flow value is determined 180 by the electronic control unit by means of the second neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value and the known value of error of the intake air mass flow. The algorithm for intake air-mass flow estimation of a naturally aspirated internal combustion engine further comprises algebraically summing 190 the selected intake air mass flow value to the error value of the intake air mass flow to obtain the corrected intake air mass flow value.

[0008] The algorithm 100 for intake air-mass flow estimation of the naturally aspirated internal combustion engine further comprises evaluating 200 the corrected intake air mass flow value against a recorded corrected intake air mass flow value that was obtained in the previous step by the electronic control unit. Therein, the algorithm 100 further comprises determining 210 correction co-effecients for the first neural network and the second neural network based on the evaluation of the corrected intake air mass flow value against the recorded corrected intake air mass flow value, and obtaining 220 the correction co-efficients iteratively until the corrected intake air mass flow value become equal to the recorded corrected intake air mass flow value. The algorithm 100 further comprises transmitting 230 the correction co-efficients to an electronic control unit consisting of the first neural network and the second neural network, and inputting 110 the first set of data values comprising the rail pressure from the rail pressure sensor, the intake manifold air pressure value from the intake manifold air pressure sensor, the engine speed from the engine speed sensor, a high-pressure exhaust gas recirculation position feedback value from an exhaust gas recirculation position feedback sensor, a modelled lambda value from an oxygen sensor in the exhaust path, a modelled exhaust manifold upstream pressure using an oxygen sensor in an exhaust path to the electronic control unit..

[0009] The algorithm 100 for intake air-mass flow estimation of the naturally aspirated internal combustion engine further comprises inputting 120 the second set of data values comprising the ratio of fuel and oxygen data from the lambda sensor, and the pressure value from one of the exhaust manifold pressure sensor and the exhaust manifold pressure model, and inputting 130 a third set of data values comprising an actual air mass flow value from an intake air mass flow sensor to the electronic control unit. Based on the first set of data values, the second set of data values, and the third set of data values, the electronic control unit determines 140 the intake air mass flow value for each of the three set of data values. The algorithm 100 therein comprises inputting 150 the intake air mass flow values into the first neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and the temperature value of an exhaust manifold temperature sensor into the electronic control unit. The electronic control unit 160 selects one intake air mass flow value from the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by means of the first neural network.

[0010] The algorithm 100 for intake air-mass flow estimation of the naturally aspirated internal combustion engine further comprises inputting 170 the selected intake air mass flow value into the second neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and the temperature value of the exhaust manifold temperature sensor into the electronic control unit. The algorithm 100 further comprises determining an error value of the intake air mass flow from the selected intake air mass flow value by means of the second neural network, and algebraically summing the selected intake air mass flow value to the error value of the intake air mass flow to obtain the corrected intake air mass flow value from the electronic control unit.

[0011] The algorithm 100 for intake air-mass flow estimation of the naturally aspirated internal combustion engine further comprises the first neural network that depends on inputs from a current engine cycle. In addition, the algorithm 100 comprises the second neural network that depends on inputs from the current engine cycle and the previous engine cycle.

[0012] It must be understood that the embodiments explained above are only illustrative and do not limit the scope of the disclosure. Many modifications in the embodiments with regard to dimensions of various components are envisaged and form a part of this invention. The scope of the invention is only limited by the scope of the claims.

, Claims:CLAIMS

We Claim

1. An algorithm (100) for intake air-mass flow estimation of a naturally aspirated internal combustion engine, the algorithm (100) comprising the steps of:
inputting (110) a first set of data values comprising a rail pressure from a rail pressure sensor, an intake manifold air pressure value from an intake manifold air pressure sensor, an engine speed from an engine speed sensor, a high-pressure exhaust gas recirculation position feedback value from an exhaust gas recirculation position feedback sensor, a modelled lambda value from an oxygen sensor in an exhaust path, a modelled exhaust manifold upstream pressure using an oxygen sensor in an exhaust path;
inputting (120) a second set of data values comprising a ratio of fuel and oxygen data from a lambda sensor, and a pressure value from one of an exhaust manifold pressure sensor and an exhaust manifold pressure model;
inputting (130) a third set of data values comprising an actual air mass flow value from an intake air mass flow sensor;
determining (140) an intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values;
inputting (150) the intake air mass flow values into a first neural network based on an actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor;
selecting (160) one intake air mass flow value from the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by means of the first neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value;
inputting (170) the selected intake air mass flow value into a second neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and a temperature value of an exhaust manifold temperature sensor;
determining (180) an error value of the intake air mass flow from the selected intake air mass flow value by means of the second neural network that has been previously trained with known first set of data values, the second set of data values, and the third set of data values and the corresponding intake air mass flow value and the known value of error of the intake air mass flow; and
algebraically summing (190) the selected intake air mass flow value to the error value of the intake air mass flow to obtain the corrected intake air mass flow value.

2. The algorithm (100) for intake air-mass flow estimation of the naturally aspirated internal combustion engine in accordance with Claim 1, further comprising the steps of:
evaluating (200) the corrected intake air mass flow value against a recorded corrected intake air mass flow value;
determining (210) correction co-effecients for the first neural network and the second neural network; and
obtaining (220) the correction co-efficients iteratively until the corrected intake air mass flow value become equal to the recorded corrected intake air mass flow value.

3. The algorithm (100) for intake air-mass flow estimation of the naturally aspirated internal combustion engine in accordance with Claim 2, further comprising the steps of:
transmitting (230) the correction co-efficients to an electronic control unit consisting of the first neural network and the second neural network;
inputting 110 the first set of data values comprising the rail pressure from the rail pressure sensor, the intake manifold air pressure value from the intake manifold air pressure sensor, the engine speed from an engine speed sensor, a high-pressure exhaust gas recirculation position feedback value from an exhaust gas recirculation position feedback sensor, a modelled lambda value from an oxygen sensor in the exhaust path, a modelled exhaust manifold upstream pressure using an oxygen sensor in an exhaust path;
inputting (120) the second set of data values comprising the ratio of fuel and oxygen data from the lambda sensor, and the pressure value from one of the exhaust manifold pressure sensor and the exhaust manifold pressure model;
inputting (130) the third set of data values comprising the actual air mass flow value from the intake air mass flow sensor;
determining (140) the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values;
inputting (150) the intake air mass flow values into the first neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and the temperature value of the exhaust manifold temperature sensor;
selecting (160) one intake air mass flow value from the intake air mass flow value from each of the first set of data values, the second set of data values, and the third set of data values by means of the first neural network;
inputting (170) the selected intake air mass flow value into the second neural network based on the actual value of engine speed, engine torque, fuel injection quantity, and the temperature value of the exhaust manifold temperature sensor;
determining (180) the error value of the intake air mass flow from the selected intake air mass flow value by means of the second neural network;
algebraically summing (190) the selected intake air mass flow value to the error value of the intake air mass flow to obtain the corrected intake air mass flow value.

4. The algorithm (100) for intake air-mass flow estimation of the naturally aspirated internal combustion engine in accordance with Claim 3, further comprises of the first neural network that depends on inputs from a current engine cycle.

5. The algorithm (100) for mass-flow estimation of the naturally aspirated internal combustion engine in accordance with Claim 3, further comprises of the second neural network that depends on inputs from the current engine cycle and the previous engine cycle.