Quality Control: SPC, QC, QA, Six Sigma

Our Experience

Quality has many different philosophies and meanings to companies and personnel.  Therefore, it is difficult to define universal truths about quality when there is such a wide range of processes and viewpoints in use today.  Our goal is to show you some great websites about quality control that have the training / tutorial information you want.  We try to give you practical information that you will not find in books and training. 

Our Experience

Quality control should be designed into every process -- not some after thought, add-on, or special program with a fancy name. 

The process engineer should be responsible for a process’s output, quality, and cost (faster, better, cheaper).  However, a process engineer cannot be an expert in everything.  Therefore, an in-house quality expert is as important as in-house mechanical and electrical experts are.  They support the process engineers to make the process engineer and their process better. 

We do not like sampling programs on out going products – we think every finished product should be thoroughly tested as early in the production cycle as possible.  A sampling program (on out going product) is saying – “We know X percent of our product is bad and we are okay with sending X percent of bad products to our customers!”  Where as 100% testing says “We know X percent of our product is bad but we are preventing bad products from getting to our customers. 

However, sampling happens, and we even use it in some of our business practices where we don't think there any problems -- but you never know you have problems unless you actively try to find them. 

We look at quality as continuous improvement.  First, you have to start testing just to find out what your quality level is.  Then you start solving problems and comparing new process data to the old data to check your improvement.  The new information then gives you a new set of “biggest problems” which you then solve, and your process continues to improve, over years, that defects per million and six sigma starts to become a reality. 

Acceptable Level of Quality

We have the seen the difference in products and systems designed to minimize problems but their costs can be enormous.  If designing a product or assembly system is X then designing a great quality system can easily be 10X.  Typically what happens is that you try to reach some acceptable level of quality for a reasonable price and then depend on the operators and maintenance personnel to do the rest.  For example, look at the differences in the space shuttle, commercial planes, and ultralights.  Extending this analogy to manufacturing would be:

1. Space shuttle: triple redundant, failure is not an option.  The best process examples would be nuclear power plants or other plants that are handling hazardous or dangerous chemicals. 
2. Commercial planes:  this is where you don’t expect to anticipate and prevent every error condition but you try to contain and recover from them.  There is some redundancy for critical systems.  You plan for the most critical failures. 
3. Ultralights:  “Don’t worry be happy”.  Minimal feedback and control, don’t worry about problems. 

In the space shuttle, commercial plane, ultralight analogy, we typically design for the “commercial plane” standard.  Some level of reasonable cost versus error handling. 

Another factor in defining an acceptable level of quality is how much waste you are generating and how much that waste costs.  For example, if we have high tolerances and are catching every reject, then how much money are we throwing away per year?  If a higher quality machine costs X dollars and that will reduce scrap by Y dollars (assuming no other benefits) then is spending X dollars less than Y dollars? 

People that are extruding recyclable materials are in a good position because they can bust up the bad product and dump it back into the raw materials.  Whereas companies manufacturing very expensive products have to spend more to verify quality in every step of the process. 

Another concept of quality is what market are you targeting with this product.  For example, if you are in the low cost market then high quality is not expected.  However, if you are in the high-end market then your product must be of the highest quality. 

Measures of Quality

There are several measures of quality:

1. What is the quality of the raw materials
2. What is the quality of the finished goods
3. What is the quality of the tools, machines, people, and systems that convert raw materials into finished goods

You can easily measure the quality of the raw materials and finished goods.  Few process engineers can do anything about in-coming poor quality except complain to purchasing.  However, we have worked with some chemical companies that are able to measure properties of their incoming raw materials and alter their recipes to improve the quality of the finished product. 

It’s usually the system converting the raw materials into finished goods that the process engineer needs to make faster, cheaper, and better.  I would think that in most cases the quality of the tools and machines are directly proportional to the quality of the finished goods (or the system that converts raw materials into finished goods). 

Is the quality of the operators important?  That depends more on the product being produced.  If you are making fine china then, yes, I would think that the quality of the people in the system has a direct effect on the quality of the finished goods.  If you are turning out a mechanical part that is cut and drilled – then you can get a high quality machine to do the operations and the operator is simply loading and unloading the machine.  So different processes require different amounts of artistic ability from the operators. 

The level of automation is also directly proportional to the need for high quality operators.  For example, a mostly manual operation will require higher skill workers than a mostly automated process.  And the automated process should have lower variability. 

Design for Quality

One great idea is known as design for manufacturing / quality / automation.  Many designers design what they think is best without any consideration for what it takes to build their design.  It helps the process engineer and quality expert to hold design reviews to work with the designer that can significantly make the product faster, cheaper, and better. 

What customers say versus what they want

We have designed quality measurement systems that inspect what the customers specify.  Then when the inspection system is put into production the system finds more problems than the customer wants. 

Some "quality challenged" customers say they want an inspection system.  What they really want is a system that passes all their parts.  They do not understand how much effort goes into making high quality parts. 

We have had unscrupulous competitors that go in and propose systems that we know cannot meet the inspection requirements and pass many bad parts just to make the customers happy and get the business.  Fortunately for us, unfortunately for the customer and competitor, it catches up with them eventually. 

What we prefer, and try to provide, are adjustable limits that the process engineer can work with.  It’s one thing to only allow good parts out the door, but if you’re not getting many parts out the door, then you can’t stay in business.  Therefore, what we encourage the process engineer to do is to initially set their limits to some values that are rather wide but allow them to identify and work on problems while they are still getting product out the door.  Then over time and after improvements, keep tightening the limits. 

There was a great analogy in one book.  Unfortunately, we cannot remember the name of the book to refer (and give credit) to.  However, the analogy was that you are in charge of a river that is used by boats for transport.  You want to make the river as safe as possible by eliminating under water obstructions.  One way to start would be to go through the river and remove all of the obvious obstructions – i.e. anything sticking out of the water.  Then -- lower the water level by a foot and then go back through to find all the new obstructions sticking out of the water.  Then lower the river another foot, remove obstructions, then another foot, …

This is the same as tightening the limits on different parts of a process.  You start out with obvious problems.  Then tighten the limits to produce problems that are more obvious.  Then tighten the limits some more to expose more problems.  Soon you have one safe river (we mean -- great product). 

Ratio of False Positives to False Negatives

A false positive is something that is passed that should have been rejected.  A false negative is something that is failed that should have been passed.  You can tell how serious someone is about quality by their acceptance of their ratio of False Positive to False Negatives.  We have zero tolerance for False Positives (passing something that should be failed).  Production managers typically don't like this.  Again, that is what adjustable limits are for.  Open your limits if you want more to pass -- don't fake it.  Sooner or later that will be discovered. 

We are more acceptable to false negative.  Meaning, we would rather have several parts we reject as bad that turn out to be good, than allow one bad part to pass. 

When do you check for quality?

Here is another interesting phenomenon that we see.  Customers that have not been doing any serious inspection and then all of a sudden buy a correctly operating 100% inspection system to check the finished product.  They are horrified at all of the rejects!  How can their product be so lousy?!?  It must be that the inspection station is not working correctly!

If your part then has to pass through many operations and you don’t inspect until the end of the line then  the probability of many failures is high.  For example, let’s say that the probability of a part failing a single process is one percent and there are fifteen steps to manufacture the product (which is not unreasonable for a production line that is using inspection for the first time).  The probability of a good part making it all the way through the fifteen steps is .99 * .99 * .99 * … * .99 = 86 percent.  So if you are not pulling the failed parts until the end you will be putting 14 out of 100 parts into the reject bin.  Nobody will be happy with that.  However if you inspect at each process and are able to pull every reject from every station then the number of rejects coming off the line appears to be only 1 out of 100. 

Another reason for checking for quality throughout the production process is adding work to defective products.  For example, if an incoming sub assembly is bad but no one has detected this (because it is not tested until after final assembly) then everyone keeps spending time and money assembling it.  Whereas, if the sub assembly is checked at each station and bad parts are removed, later processes are not spending time and money working on bad assemblies. 

What is good?  What is bad?

One difficult aspect of inspection systems is when the customer can not give hard numbers to indicate what parts are good and what parts are bad.  In fact we have been in many applications, one operator is passing parts that another operator is failing.  This is one of the hardest problems – accurately specifying what are acceptable parts and what are reject parts. 

Value of Automation

Many people claim that a company is automating for no reason.  Besides a reduction in labor costs, and producing more parts in less time, automation has two major advantages: (1) produces parts more consistently and (2) generates process information. 

Automation should reduce the variability of the parts being made.  When you have different people actively involved in making parts by hand there is more variability.  When the people are replaced by automation the machine should make parts more consistently.  This is even more true for operations that run 24 hours per day, seven days a week.  If the process engineer can reduce their variability then it is easier to spot the problems that should be corrected. 

Value of Process Information

Another great benefit of automation is the information that the automated system can generate for the process and quality engineers.  Not only information like number tested, number passed, and number failed but information like:

• Total number of faults for the process
• Number of faults for each part of the automated process
• Total downtime for the machine / process
• Downtime attributed to each part of the machine / process
• Cycle time of process
• Cycle time of each part of the process

The value of this information is:

• Helps the process / quality engineer identify what their biggest problems are
• Provides data “before” and “after” to identify how successful a change was.  Let’s reiterate that we want continuous improvement and we want to encourage experimentation with process improvements.  Now we have information to provide feedback to show us how well our experimentation did. 

Quality Reports

Our preference is to use Microsoft Office products to produce reports.  For example, we can write data to Microsoft Excel, Access (database), or Word.  We prefer to generate data for Excel since the data is in a format that allows further analysis by the users.  We can do pretty charts where we plot each data point, averages, ranges, etc.  We can compute, report, and display most statistical measures. 

Data Recording

We can record inspection parameters on each part, or summarize the data by shift, lot, or some other group.  We can write the data to a database, spreadsheet, text file, or other formats.  We can read and write databases and data formats to interface to almost any other hardware and software.  We can store the data on a file (data) server, or on CD-ROM for permanent storage. 

One interesting new capability allows you to store the quality data in a database and then create a "web service" that allows your customers to access the quality data at any time. 

Typical programming scenario

When we develop a quality control component our approach is:

1. Clearly define what parameters are to be measured.
2. Clearly define what is good and what is bad.
3. Define what type of quality control we are doing:
   1. 100% inspection
   2. sampling
   3. binary tests: good / bad, pass / fail, go / no go, (no analog measurements)
4. Provide initial alarm and warning limits
5. Specify what statistics you want to track:
   1. Average / Mean
   2. Median
   3. Standard deviation
   4. Variance
   5. Cp
   6. Cpk
   7. Maximum
   8. Minimum
   9. Other
6. Specify what graphs you want to see
   1. Each measurement
   2. X bar
   3. R bar
   4. Pareto
   5. histogram
7. Specify under what conditions the process is warned and shutdown
   1. Warn on any measurement outside warning limits
   2. Shutdown on any measurement outside alarm limits
   3. 3 of 5
   4. 5 of 9
   5. Other
8. Define what data should be logged to disk
   1. Every test
   2. Every lot or batch
   3. Local hard drive, network drive, CD-ROM

Our Quality Software

Our quality software is known as Quality Master.  It is source code that you are free to modify as you please.  Or we can do the programming for you.  Point is -- with our software you can get exactly what you want and need -- not change your operations to work with some software.  It is also a site license so for one price you can use it on as many computers as you want. 

Every System

We try to design this information gathering and dissemination into every machine we design.  We want to help the process engineer improve their process -- fast, better, and cheaper. 

We try to offer a fair and balanced opinion on every page of our website.  We would appreciate more information from other users to express their opinions which we will then incorporate.  If you have questions or comments please post them on our message board (see button in left hand column) so that others can read and benefit.