Automatic prognostic qualification of manufacturing products

Patent metadata

Field	Value
Publication number	US-11209345-B1
Application number	US-201916425644-A
Country	US
Kind code	B1
Filing date	May 29, 2019
Priority date	May 29, 2019
Publication date	Dec 28, 2021
Grant date	Dec 28, 2021

Abstract

A system includes a void filter that receives sensor data employed to produce or inspect a manufactured part, the void filter generates a void data subset indicating voids detected in the manufactured part. A stress analyzer processes the void data subset from the void filter and determines coordinate data and force data for the respective detected voids in the manufactured part. At least one deterministic model analyzes the coordinate data and the force data from the stress analyzer determined for the detected voids from the void data subset. The deterministic model analyzes failure of the detected voids with respect to time and generates deterministic output data indicating failure over a deterministic timeframe. A prognostic analyzer processes the deterministic output data from the at least one deterministic model and generates a failure prediction for the as manufactured part.

First claim

What is claimed is:

1. A system, comprising:
one or more computers executing computer executable components from a non-transitory computer readable medium, the computer executable components comprising:
a void filter having executable instructions to receive sensor data employed to produce or inspect a manufactured part, the void filter generates a void data subset by identifying voids that indicate a potential defect in the manufactured part, the void filter assigns void parameters to the identified voids to quantify and qualify the potential defect in the manufactured part, the void filter identifies the voids based on comparing the assigned void parameters of the identified voids to a void defect threshold, the void filter generates a void location tag in the void data subset representing coordinates on the manufactured part where the identified voids exceed the void defect threshold;
a stress analyzer having executable instructions to process the void data subset from the void filter and to generate stress output data for the respective identified voids in the void data subset, the stress analyzer to associate a stored force vector having a force magnitude and direction to the coordinates specified by the void location tag, the force vector to model expected stress forces to be applied at the respective coordinates of the manufactured part, the stress output data generated as force vector and coordinate subsets by the stress analyzer representing the expected stress forces to be applied at the respective coordinates of the identified voids;
at least one deterministic model having executable instructions to analyze each of the respective force and coordinate subsets from the stress output data generated by the stress analyzer, the at least one deterministic model determines a curve for each of the respective force and coordinate subsets where the force vector is applied at the associated coordinates over time, the at least one deterministic model generates deterministic output data for each of the respective force and coordinate subsets over a deterministic timeframe representing failure of the manufactured part at the specified coordinates at a given point in time, the deterministic model compares a point on the curve to a failure threshold where failure is predicted for the manufactured part at the specified coordinates based on the applied force vector at the given point in time exceeding the failure threshold, such that a minimum time that failure is not likely to occur is based on the applied force vector at the given point in time exceeding the failure threshold; and
a prognostic analyzer having executable instructions to process the deterministic output data from the at least one deterministic model and to generate a failure prediction output for the manufactured part based on computing a collective failure probability from the deterministic output data with respect to the predicted failure of the identified voids over the given timeframe, wherein the prognostic analyzer determines the collective failure probability by analyzing a probability estimate for identified voids and associated force vectors to generate a likelihood of failure estimate at a time before the failure in the deterministic timeframe to a prediction of the minimum time that the failure is not likely to occur.

2. The system of claim 1 , further comprising a process monitor having executable instructions to monitor a plurality of sensors to generate the sensor data, the plurality of sensors coupled to a machine to produce the manufactured part or to a process variable affecting quality of the manufactured part, the sensors include at least one of a voltage sensor, a current sensor, a power sensor, a position sensor, a velocity sensor, an accelerometer, a strain gauge, a surface quality sensor, a camera sensor, an x-ray sensor, and a microscope sensor.

3. The system of claim 2 , wherein the process monitor includes a processor and a program memory storing the executable instruction and a random memory to store the sensor data, the process monitor includes at least one of an analog-to-digital converter (ADC) to convert received sensor signals to the sensor data.

4. The system of claim 2 , wherein the void parameters characterize the sensor data that further characterizes a surface of the manufactured part, wherein the void parameters exceeding the void defect threshold indicates that the void is present in the manufactured part.

5. The system of claim 1 , wherein the void filter executable instructions generates a location map indicating a location for the respective detected voids with respect to a correlated position on a surface coordinate of the manufactured part.

6. The system of claim 5 , wherein the void filter executable instructions includes a size analyzer to determine an area value for the respective detected voids and a density analyzer to determine a density value for a cluster of voids detected within a predetermined region of the manufactured part.

7. The system of claim 6 , wherein the void filter executable instructions selects the respective detected voids in the void data subset by comparing the area value to a predetermined area threshold and selecting voids that are greater in size than specified by the area threshold or by comparing the density value for the cluster of voids to a cluster region threshold and selecting void clusters that are greater in size than specified by the cluster region threshold.

8. The system of claim 7 , wherein the stress analyzer executable instructions includes a force analyzer to apply at least one force vector specifying the force magnitude and force direction to the respective voids identified in the void data subset, the stress analyzer executable instruction includes a location analyzer executable instruction to process the void data subset from the void filter to associate the at least one force vector with the void locations coordinates specified in the void data subset, and to generate a stress map indicating applied force vectors at the identified void coordinates.

9. The system of claim 8 , wherein the at least one deterministic model executable instructions processes force data from the stress map and generates the deterministic output data indicating failure and over a deterministic timeframe by generating a curve of the force vector at the respective void coordinates with respect to time, wherein failure is determined by an executable instruction comparison to a point on the curve to a predetermined failure threshold.

10. The system of claim 9 , the prognostic analyzer further comprising executable instructions that include at least one learning model to generate the likelihood of failure estimate based on a statistical determination of the learning model, wherein the learning model includes a classifier or a neural network to perform the statistical determination based on the probability estimate to the collection of selected voids.

11. The system of claim 1 , further comprising a concurrent uncertainty management system having executable instructions to receive the failure prediction output from the prognostic model, determine a product life estimate based on processing the failure prediction output, and update executable instructions or data of at least one of the void filter, the stress analyzer, the deterministic model, and the prognostic model based on the processing.

12. The system of claim 1 , further comprising a real time data generator to update at least one of the void filter, stress analyzer, deterministic model, and the prognostic analyzer based on vehicle data received from the real time data generator, where the vehicle data is collected from

Assignees

• Madsen John S
• Nardiello Jerrell A
• Anagnostou Elias L
• Crawford John A
• Engel Stephen J
• Northrop Grumman Systems Corp

Inventors

• Madsen John S
• Nardiello Jerrell A
• Anagnostou Elias L
• Crawford John A
• Engel Stephen J

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G05B23/0283

  Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] (preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06) · CPC title

• G01N3/06 Primary

  Special adaptations of indicating or recording means · CPC title

• G06N5/048

  Fuzzy inferencing · CPC title