Method for predicting impact and an impact prediction system for realizing the same by using neural networks

• US 5,377,108 A
• Filed: 04/13/1993
• Issued: 12/27/1994
• Est. Priority Date: 04/28/1992
• Status: Expired due to Fees

• Current Assignee: Takata Corporation (Ningbo Junsheng Investment Group Co., Ltd.)
• Original Assignee: Takata Corporation (Ningbo Junsheng Investment Group Co., Ltd.)
• Inventors: Nishio, Tomoyuki
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Walker, Tyrone V.
• Application Number:
  US08/045,247
• Time in Patent Office: 623 Days
• Field of Search: 364/424.05, 364/424.01, 364/424.1, 364/194, 364/164, 364/178, 340/436, 340/669, 180/282, 180/274, 307/10.1, 280/734, 280/735
• US Class Current: 701/45
• CPC Class Codes: B60R 21/0132 responsive to vehicle motion parameters , e.g. to vehicle longitudinal or transversal deceleration or speed value
• CPC Class Codes: G06N 3/045 Combinations of networks

Abstract

A method for predicting impact by using neural networks comprising steps of supplying a predetermined crash curve to a first neural network having an intermediate layer to train said first neural network by means of learning calculation and supplying a predetermined air bag deployment limit curve to a second neural network to train said second neural network, supplying data indicative of crash curve obtained by an acceleration sensing device on collision to said first and second neural networks, predicting in said first neural network a time instance at which a threshold displacement is going to reach based on the basis of the training result in said first neural network, comparing in said second neural network data indicative of crash curve on said collision and said air bag deployment limit curve to produce a decision signal of deploying the air bag according to the comparison result, calculating said decision signal and said time instance, to deploy the air bag depending on the impact, and supplying an operation command signal to an air bag deployment operation device.

Claims

1. A method for predicting impact by using neural networks comprising steps of:

supplying a predetermined crash curve as a first group of parallel training data to a first neural network having an intermediate layer to train said first neural network by means of a learning calculation of a back propagation network architecture and supplying a predetermined air bag deployment limit curve as a second group of training data to a second neural network to train said second neural network by a back propagation network architecture;

supplying data indicative of a crash curve obtained by an acceleration sensing device when collision occurs as a set of data having a predetermined time interval to said first and second neural networks;

predicting, based on initial data of the crash curve obtained from said first neural network, a displacement output from the initial data of the crash curve on the basis of the training result in said first neural network, and obtaining, based on the displacement output from the initial data, a time to reach the threshold for maximum occupant displacement in which an air bag can be opened to protect an occupant,;

comparing in said second neural network data indicative of the crash curve on said collision in accordance with the training result in said second neural network and data obtained from the air bag deployment limit curve, and outputting a decision signal indicative of deploying the air bag only when the data for the crash curve are greater than the data from the air bag deployment limit curve; and

supplying an air bag deployment operation command signal only when the time to reach the threshold for maximum occupant displacement is obtained in the first neural network and when the second neural network outputs the decision signal indicative of deploying the air bag.

2. A method for predicting impact by using neural networks as claimed in claim 1, wherein each of said first and said second neural networks comprises a two dimensional self-organizing competitive learning layer in said intermediate layer to perform learning operation in said two dimensional self-organizing competitive learning layer.

3. A method for predicting impact by using neural networks as claimed in claim 1, wherein said first group of parallel training data is used for learning calculation by means of converting the crash curve obtained by calculation in said first neural network into displacement data.

4. A method for predicting impact by using neural networks as claimed in claim 3, wherein said first group of parallel training data is supplied in parallel to an input layer of said first neural network after parallel converting a subsequent displacement values within a predetermined range, and displacement values following to said subsequent displacement values are parallel converted into a desired output data to be supplied in parallel to an output layer of said first neural network.

5. A system for predicting impact by using neural networks comprising:

displacement prediction means where crash curve data obtained on collision by an acceleration sensing device are supplied, as a data set having a predetermined time interval, to a first neural network having an intermediate layer already trained by means of a learning calculation of a back propagation network architecture and supplied with a predetermined training crash curve as a first group of parallel training data, said first neural network predicting a displacement output after initial crash curve data and obtaining a time to reach a threshold for maximum occupant displacement in accordance with the training result in said first neural network;

air bag deployment decision means where said crash curve data obtained on collision are supplied to a second neural network having an intermediate layer already trained by means of a learning calculation of a back propagation network architecture and supplied with a predetermined air bag deployment limit curve as a second group of training data, said second neural network comparing data indicative of the crash curve on said collision in accordance with the training result in said second neural network and data obtained from the air bag deployment limit curve and outputting a decision signal indicative of deploying an air bag only when the data for the crash curve are greater than the air bag deployment limit curve; and

a logic circuit where an air bag deployment operational command signal is supplied to air bag deployment operation means only when the time to reach the threshold for maximum occupant displacement is obtained in the first neural network and when the second neural network outputs the decision signal indicative of deploying the air bag.

6. A system for predicting impact by using neural networks as claimed in claim 5, wherein each of said first and said second neural networks comprises a two dimensional self-organizing competitive learning layer in said intermediate layer to perform learning operation in said two dimensional self-organizing competitive learning layer.

7. A system for predicting impact by using neural networks as claimed in claim 5, wherein said first group of parallel training data is used for learning calculation by means of converting the crash curve obtained by calculation in said first neural network into displacement data.

8. A system for predicting impact by using neural networks as claimed in claim 7, wherein said first group of parallel training data is supplied in parallel to an input layer of said first neural network after parallel converting a subsequent displacement values within a predetermined range, and displacement values following to said subsequent displacement values are parallel converted into a desired output data to be supplied in parallel to an output layer of said first neural network.

9. A system for predicting impact by using neural networks as claimed in claim 5 wherein said first and said second neural networks are coded after completion of learning to be implemented in a substrate of an air bag operation control circuit.