Automated Inspection and Quality Assurance

Stanley L. Robinson, Richard Kendall Miller

• Publisher: CRC Press
• Year: 2017
• Print ISBN: 9780824780029
• eBook ISBN: 9781351465076
• Edition: 1
• Topic: Technology & Engineering
• Subtopic: Operations
• Index: Technology & Engineering

New concepts for gaging, inspection, checking, machine vision, and robotic testing. Includes guidelines for installing complex electronic and computerized systems and a directory of commercially availalbe computer software, as well as distributors' names and addresses. Annotation copyright Book News

1

Historical Overview

Quality in manufacturing has always been of interest to American industry. It has only been within the past decade, however, that computers have been applied to automating the quality assurance task. The batch processing of mainframe computers of the pre-1970 era was not applicable to real-time operation, and these computers were much too costly for dedicated industrial operations. When the development of minicomputers opened the door to manufacturing applications of computers, inspection was one of the first tasks considered. The minicomputers of the early 1970s offered only a fraction of the capabilities of today’s personal computers at several times the cost. However, they were sufficient for researchers to begin to develop the first machine vision systems and other technologies which today form the basis of automated quality assurance. This chapter reflects on some of these early efforts which matured to a wide range of commercial inspection products in the late 1970s and early 1980s.

The National Science Foundation (NSF) undertook a Production Research and Technology program between 1973 and 1982. The objective of this program was to develop a new technology called “programmable automation” to advance automation in batch manufacturing of discrete products for both economic and social reasons. Two of the primary technologies developed were machine vision and robotics. Among the primary contractors were Stanford University, SRI International, Purdue University, and the University of Rhode Island.

The NSF sponsored program at SRI International, begun in April 1973, was the driving force behind the commercialization of machine vision systems nearly a decade later. The research utilized a 128 × 128 solid-state camera and a DEC LSI-11 microcomputer with a 28K word memory. While these components offered only a fraction of the capabilities of today’s commercial systems, SRI developed some important techniques. A software routine called “connectivity analysis” allowed the analysis of objects (e.g., dimensions, area, etc.) and the extraction of local features (e.g., holes and corners). A method was developed for using gray scale to inspect registered objects for pseudocolor. Structured light was applied to inspection of three-dimensional objects by comparing model lines with the inspected lines of intersection of the light plane and indexed three-dimensional objects. All of these techniques of now in popular use, achieving production line speed capabilities due to advance in (VLSI) technology in the decade following the SRI research.

The industrial robot was invented in 1951 by George C. Devol, and commercialized in 1961 by Joseph Engleberger and his company, Unimation. Early industrial robots, however, were controlled by mechanical switches rather than computers. They were heavy-duty machines designed for transporting large loads in the harsh environments of foundries, forges, and metal fabrication plants. The era of computer-controlled “intelligent” robots, applicable to such sophisticated tasks as inspection and assembly, had its roots in university laboratories, primarily sponsored by the National Science Foundation program.

The first program in robotics and computer vision was started at Stanford University in 1965 by John McCarthy. Several landmark accomplishments in the robotics field were recorded in the early years of the program at Stanford. The Scheinman Stanford robot arm (1970) was the forerunner of modern robots used for industrial assembly. The WAVE programming system (1971), developed by Richard Paul, was the first to provide several important capabilities of modern robot programming systems: predictive Newtonian dynamics, automatic planning of smooth trajectories, rudimentary force and touch sensing used in control, and a macro library of assembly operations. This led to the first computer-integrated assembly in 1973, the assembly of an automotive water pump from 10 component parts. By the late 1970s, the Scheinman arm had evolved into the commercial PUMA robot, and WAVE evolved into the commercial VAL programming system. The age of industrial robotics had begun.

Some important contributions to robotics were also developed under NSF sponsorship by Dr. R.B. Kelley’s group at the University of Rhode Island. Studies centered on methods to enable robots with vision to acquire, orient, and transport workpieces. A method was developed to allow a vision guided robot to acquire randomly placed workpieces from a bin. This “bin picking problem” was listed first among the frustrating gaps in needed knowledge for factory automation, according to a 1976 poll of members of the Society of Manufacturing Engineers. Also developed in Kelley’s program was an instrumented parallel jaw gripper which allowed a robot to “feel” the workpiece within its grip.

The NSF sponsored the development of advanced industrial robot control systems at Purdue University, under the direction of Richard Paul and Shimon Nof. Among the accomplishments were advanced position, velocity, and force controls, and a technique for simulating a robots work routine. Simulation subsequently has become a routine practice for engineers in developing robotic work cells.

The application of this research to automated quality assurance and computer-integrated manufacturing began in 1978. In that year, Machine Intelligence Corporation, founded by researchers from SRI International, introduced the first machine vision used for inspection. Also in 1978, General Motors demonstrated the CONSIGHT vision system, used in conjunction with a Cincinnati Milacron robot for sorting metal castings.

The November 1981 AUTOFACT conference in Philadelphia is recognized as the event which launched the widespread recognition of machine vision technology. At this event over a dozen companies demonstrated commercial vision inspection systems and over 5,000 engineers left with the knowledge of how this technology could be utilized for automated quality assurance within their companies. The following year saw IBM, Control Automation, Intelledex, and other companies introduce robot models with extreme precision and computer vision capabilities for sophisticated inspection and assembly tasks. Automatix was recognized for the marriage of robotics with a statistical process control software package. Lord Corporation and Barry Wright Corporation introduced advanced tactile sensor pads for robot grippers. The computer-based techniques of machine vision were expanded to other types of sensors for inspection. Based on research at MIT, Cochlea Corporation developed a commercial computer-based ultrasonic system for the inspection of small parts.

Today, machine vision systems for inspection and industrial robots have become commonplace. Over 10,000 computer vision systems and over 20,000 industrial robots have been installed in industries of the United States. But this is only a beginning. Of the 700,000 different quality control tests run regularly in the United States, it is estimated that at least 25% could be replaced by fully automated machine vision inspections. An additional 40% could be more effectively handled by an operator using machine vision as a gauge. A report by the National Bureau of Standards suggested that on the order of 90% of all industrial inspection activities requiring vision will be done with computer vision systems within the next decade.

The potential cost savings achievable by automating the repetitive, boring inspection task through the use of automated inspection is tremendous. Inspection tasks are performed by an estimated 10% of the U.S. workforce. This translates to 400,000 persons. If an average annual wage of $20,000 is assumed, the annual cost of industrial inspection in the United States is $8 billion. The incentive for U.S. industry to implement automated quality assurance is not only to turn this cost into profit through automation, but also to increase product quality as necessary for U.S. industry to maintain its position as world leader in manufactured goods.

2

How Automated Quality
Assurance Affects the World

2.1 Characteristics of the Old World

2.2 Characteristics of the New World

2.3 Wheel of Progress

In order to evaluate the effect of automated quality assurance (AQA) on the world, we will compare two worlds. The present predominant world will be referred to as the “Old World.” The after-automation world is the “New World.”

2.1 CHARACTERISTICS OF THE OLD WORLD

The Old World, before the influence of automated quality assurance, has the following characteristics:

1. A fully-manned, costly, quality control department: managers, supervision, on-line inspectors, sampling inspectors, and data handlers.

2. Use of sampling techniques concerned with outgoing quality level (OQL) which yield a less-than-perfect inventory of finished product.

3. The human on-line inspectors become fatigued and perform in less-than-satisfactory manner.

4. Often destructive sampling is essential, which results in a costly quality control method.

5. Scrap loss is elusive. The tendency is to use broad stroke methodologies, i.e., withdrawals of raw material from warehouse equals input. Finished product equals output. The difference is scrap loss. Not necessarily true, i.e., inaccurate warehouse records, material moisture control, etc.

6. After-the-fact control samples taken and inspections made followed by time-consuming analysis while good or bad product continues to be made.

7. Simple manual data processing is used for record-keeping. This limits the degree and extent of data collection. Process of record-keeping is relatively slow and costly.

8. Manufacturing is not a real-time process. To know what is happening while it happens allows for a better decision-making process.

9. If raw material varies, the process no doubt varies. The standard deviation increases without the knowledge of the operators. It can happen in small increments that are not detectable by human eye.

10. Output can go beyond the upper control limits (UCL) or lower control limits (LCL) without human detection.

11. Process control is only a broad stroke method. Data collected by sampling methods is general, nonspecific, and not timely. Process could move in or out of the control limits before knowledge of sample results are available.

12. A process is practical when it has a known margin of error, i.e., you are allowed to deviate from the mean and still produce an acceptable product. Acceptable product may mean a product with some compromises in order to make the process usable and practical. The less the knowledge about the process data, the greater the potential for compromises.

13. The Old World deals in macrocosms. It evaluates end results, and is not capable of finer scrutiny.

14. System (old) is not on-line, real time, user-friendly.

Slow to conclude process adjustments

Slow scrap control

Does not evaluate supplier

15. Specification limits are usually broad due to lack of process insights through inadequate information.

16. Tendency to allow higher work in process to cover for contingency.

17. Higher risk for back order due to less control and predictability.

2.2 CHARACTERISTICS OF THE NEW WORLD

The New World—after the introduction of automated quality assurance—may be characterized as follows:

1. AQA allows better evaluation of suppliers which can lead to just-in-time (JIT). The results are increased inventory turnover in magnitudes of a present situation of four to six increasing to 20 to 40. Consequently work-in-process (WIP) is reduced.

2. One-hundred percent inspection becomes possible. Slow untimely sampling procedures are replaced with high-speed accurate information. Outgoing quality level approaches 100% acceptable product.

3. Analysis in microcosms becomes possible. Inspection points can be increased (e.g., inspection can occur at several points of a machine as opposed to one human inspection at the discharge point).

4. Reduced quality control manpower results in a cost reduction.

5. On-line test of raw materials is possible at points of entry to the process. Supplier evaluations are a convenient result.

6. Automated quality assurance is superior to human effort—less fatigue and less variance.

7. Increased points of inspection before the final product is assembled can eliminate costly after-the-fact destructive inspections.

8. Scrap avoidance can be a built-in factor. High-speed data collection in real time at a greater number of points in the process allows for better decisions.

9. Before-the-fact control or synchronous control becomes possible.

10. Computer integration with other segments of a work cell or within a department becomes possible.

11. Automated quality assurance is real time, user-friendly.

12. Process variation can be tracked in smaller increments or continuously with speedy data collection and interpretation resulting in usable information.

13. Process control takes on a new profile of technology. Managers know what is really happening as opposed to approximation.

14. Specification limits do not have to be as broad. Product quality is improved.

15. The practical margin of error can be identified as well as reduced with AQA.

16. Customer satisfaction should increase due to a better, more consistently performing product.

17. Automated quality assurance can see better than the human eye.

18. Better control of the process requires less work-in-process (WIP).

19....

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• About the Series
• Preface
• 1 HISTORICAL OVERVIEW
• 2 HOW AUTOMATED QUALITY ASSURANCE AFFECTS THE WORLD
• 3 COMPUTER-INTEGRATED MANUFACTURING
• 4 MACHINE VISION
• 5 SENSORS FOR INDUSTRIAL INSPECTION
• 6 ROBOTIC INSPECTION
• 7 SOFTWARE FOR QUALITY ASSURANCE
• 8 INSPECTING THE PRODUCT
• 9 INSPECTING THE PACKAGE
• 10 DEVELOPING THE IN-HOUSE PROGRAM
• 11 ARTIFICIAL INTELLIGENCE: THE NEXT STEP
• 12 CASE STUDIES
• Appendix 1: Glossary
• Appendix 2: Commercially Available Software
• Index