Browse Prior Art Database

TECHNIQUES FOR A MACHINE-LEARNING-BASED APPROACH FOR CODE-CONTEXT ENRICHMENT FOR INDUSTRIAL INTERNET OF THINGS SYSTEMS

IP.com Disclosure Number: IPCOM000249408D
Publication Date: 2017-Feb-23
Document File: 13 page(s) / 1M

Publishing Venue

The IP.com Prior Art Database

Related People

Vinay Kolar: AUTHOR [+3]

Abstract

Presented herein are techniques to model and associate contextual data with lines of code using machine learning algorithms. These techniques may be used for providing services for development and consultation in Industrial Internet of Things (IIoT) equipment manufacturers and product manufacturers.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 19% of the total text.

Copyright 2017 Cisco Systems, Inc. 1

TECHNIQUES FOR A MACHINE-LEARNING-BASED APPROACH FOR CODE- CONTEXT ENRICHMENT FOR INDUSTRIAL INTERNET OF THINGS SYSTEMS

AUTHORS: Vinay Kolar Qihong Shao Gyana Dash

CISCO SYSTEMS, INC.

ABSTRACT

Presented herein are techniques to model and associate contextual data with lines

of code using machine learning algorithms. These techniques may be used for providing

services for development and consultation in Industrial Internet of Things (IIoT) equipment

manufacturers and product manufacturers.

DETAILED DESCRIPTION

Heavy manufacturing industries, such as the Industrial Internet of Things (IIoT),

use machines that are capable of manufacturing custom parts based on a program that is

fed into the machine. These machines are programmed by Computer Numerical Control

(CNC) programs, which automate the manufacturing process by precisely directing the

machine to perform actions that are carried out in a sequence. A manufacturing company

may have multiple machines which produce hundreds of parts in parallel. Bugs in the code

can lead to waste of materials, unforeseen downtime, and loss in revenue.

This is exacerbated by the fact that some parts of the code may contextually lead to

faulty manufacturing. For example, cutting a hard metal for more than 30 minutes may lead

to problems only on older machines when the humidity is low. Such contextual errors are

harder to identify during regular testing cycles because such identification requires

exhaustive testing under a combinatorial number of contextual parameters.

As such, provided herein are automatic techniques to model and associate

contextual data with lines of code using machine learning algorithms. These techniques

may be used to provide services for development and consultation in Industrial IoT (IIoT)

equipment manufacturers and product manufacturers.

Copyright 2017 Cisco Systems, Inc. 2

More particularly, described herein is a machine-learning based approach to

debugging IIoT systems based on the observed behavior of the programs. The system

accepts both real-time and batch data from the enterprise. The data may include:

1. Program Data: This can be the code of the CNC program with delimited blocks and

sequences.

2. Real-time Data from Machine, including:

a. Program parameters being executed, including the program name, the code

block and the exact sequence number of code that is currently being

executed.

b. Sensor data, including machine context data such as vibration, temperature,

spindle positions. Usually approximately 50 to 70 such sensor parameters

may be found in one large CNC machine.

c. Product data, which indicates the product being manufactured, such as the

type of product, product-id, etc.

Usually a row of data indicating the above parameters is emitted once every few

hundredths of a millisecond.

3. Item Inspection Data: The portion being produced is inspected by humans or

machines, and is marked as defective or not. They may also include the type of

d...