Neural network-based virtual sensor for automatic...

Data processing: vehicles – navigation – and relative location – Vehicle control – guidance – operation – or indication – Transmission control

Reexamination Certificate

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

C701S051000, C706S905000

Reexamination Certificate

active

06275761

ABSTRACT:

TECHNICAL FIELD
This invention pertains to the control of an automatic transmission using a programmed microprocessor. More specifically, this invention pertains to a virtual sensor for estimating torque converter slip in the operation of a powertrain module in shift control for an automatic transmission.
BACKGROUND OF THE INVENTION
Powertrain control modules (PCM) are now widely used in automotive vehicles to control operations of the vehicle engine and multi-gear ratio automatic transmission. A PCM includes a microprocessor and suitable associated memory chips, input-output devices and the like and is programmed by a vehicle manufacturer to control engine and transmission functions such as air and fuel intake, spark timing and transmission shift schedules. The PCM receives data concerning engine and transmission operation from many electrical and electromechanical sensors.
In an automobile, the transmission is the component that transfers torque from the engine to the wheels to move the vehicle. The transmission does this by providing several forward gear ratios and one reverse gear ratio which enable the engine to accelerate the vehicle quickly, obtain high speeds and reverse the vehicle. An automatic transmission also allows the vehicle to stop while the engine is running without a manual clutch pedal. A torque converter provides this function by acting as a fluid coupling between the crank shaft and flywheel of the engine and the torque input to the transmission. Thus, in some operating modes of the powertrain, there is slippage in the torque converter and a difference between the speed of the engine and the speed of the output shaft of the transmission which cannot be determined from knowledge of the gear ratios of the transmission. This difference, e.g., in revolutions per minute (rpm), is known as slip.
The PCM requires certain inputs in order to control shifting of the forward gear ratios of the transmission. Such inputs include, for example, vehicle speed, throttle position, engine speed, present gear ratio, and transmission output speed. These inputs are provided by suitable electrical and electromechanical sensors associated with the drive axle, throttle and crankshaft, respectively, and electrically connected to the PCM. The PCM then sends signals to shift solenoids to cause suitable transmission upshifts and downshifts, to a transmission fluid pressure control solenoid to adjust shift feel, and to the torque converter control solenoid to engage or release the torque converter clutch. All such sensors must be designed, manufactured and assembled into the powertrain and maintained during vehicle life. For PCM shift scheduling purposes, it is also desirable for the computer to have data concerning torque converter slip because it is a parameter that significantly affects shift timing.
Torque converter slip is the difference between the torque converter input and output speeds. Torque converter input speed is equivalent to engine speed because the converter cover is bolted to the engine fly wheel and turns at engine speed. Data concerning engine speed is important to many PCM engine control functions, and a crankshaft position sensor is used to provide such information. Torque converter output speed is not measured, but in some powertrain designs a transmission input speed sensor (TISS) is used. Transmission input speed is the same as torque converter output speed, and when a TISS is available, the PCM can calculate torque converter slip whenever that data is required for shift control. However, a TISS is another device that must be assembled into the transmission, adding to the complexity and cost of the powertrain.
As increased and less inexpensive computer capacity becomes available for powertrain control use, it would be preferable, where possible, to estimate parameters such as torque converter slip from other available data rather than provide another expensive and, perhaps, vulnerable electromechanical device in the powertrain. Accordingly, it is an object of this invention to provide a virtual sensor for determining torque converter slip.
SUMMARY OF THE INVENTION
A virtual sensor is an estimator which uses measured quantities from a system to estimate another unmeasured quantity from this same system. This invention involves the use of neural network techniques to design a virtual sensor for torque converter slip.
Neural networks are an information processing device usually executed on a computer. They are a compact, well-defined mathematical structure for implementing nonlinear systems, and they utilize a number of simple modules or neurons. Information is stored in the structure by components that at the same time effect connections between the neurons. While they may vary in complexity, neural networks can all be represented by a basic structure. They comprise an input layer of neurons, one or more hidden layers of neurons and an output layer of neurons. A network may have only one layer of input and output neurons, but may feature one or more hidden layers. Each layer may feature any number of neurons.
The precise mathematical calculations performed on the data input to a neural network are a function of the specific network design. This design is characterized by two main components. The first component is the overall network architecture which specifies the number of layers of neurons, the number of neurons in each layer and the specific input and output signals to the system. The second component consists of the specific network weights and activation functions which govern the interaction between neurons. In devising a suitable neural network, it is crucial to determine an appropriate overall network architecture for any given problem or application.
The internal architecture of each individual neuron in a neural network is identical, regardless of the layer in which it resides. This architecture consists of two separate parts: an algebraic operator which computes an input signal to a neuron based on a specific linear combination of output signals from the previous layer of neurons, and an activation function which converts this input signal into some output value. Thus, the output of a neuron in any given layer of the network (except the input layer) is completely determined by the output signals from the previous layer.
An advantage of the neural network approach to devising a virtual torque converter slip (or transmission input speed) sensor is that a network may be constructed and trained to estimate slip even when the mechanics or dynamics of torque converter slip may not otherwise lend themselves to mathematical modeling. In other words, a trial neural network for a virtual sensor may be constructed and its output, predicted torque converter slippage in rpm, may be compared with experimentally measured torque converter slippage of the powertrain system of interest. Engine and transmission operating parameters affecting torque converter slip are selected for input neuron data. The construction and evaluation of the network is assisted using available software. The network is revised until its output suitably simulates the test system.
In accordance with this invention, it is found that a single neural network-based torque converter slip estimator design performs well for some, but not all, operating conditions of the engine-torque converter-transmission combination encountered in vehicle operation. A feature of this invention is development of a single, composite slip estimator which utilizes different neural network designs for different operating conditions of the vehicle. A specific neural network-based slip estimator design can be tailored to perform well for limited specific sets of powertrain operating conditions. An important aspect of this invention is to partition the vehicle's operating envelope into a number of distinct subsets. Then, for each subset, a different slip estimator is designed which performs well under those specific conditions. The final, composite design is obtained by seamlessly transitioning between these differe

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Neural network-based virtual sensor for automatic... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Neural network-based virtual sensor for automatic..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Neural network-based virtual sensor for automatic... will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-2445053

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.