Monthly Archives: September 2014

You are browsing the site archives by month.

Screen Shot 2014-09-25 at 4.28.39 pm

Philip Sage, Principal Engineer

An “unreliable” manufacturing process costs more money to operate.

Management “always says” we need to improve.

Individually, we know that “You cannot improve what you do not measure”.

So we must conclude if we want to make our process “reliable” we must measure the process reliability.

The search for measurable data that can be utilised may seem hard.  Equally hard could be a high level understanding of when a process is reliable, and what specifically must a process exhibit to be deemed reliable?

Fortunately for many, the daily production data is often the most accurate data found in a manufacturing plant.  It is also easy to find a source for this data. After all, this one daily number represents the efforts of the entire team across a 24 hours total. With just this one number, your management represents the relative health of the manufacturing process to their peers.

But how do you convert this “number” into “reliability” and what does it show us?

For starters, we can calculate the “lost opportunity” costs due to process variation. We can also identify the causes and assign actionable tasks to reduce these losses. While care must be taken in the use of certain production numbers, generally speaking, once understood, the number your organisation tracks can be utilised to determine your process reliability.

A world class “reliable process” is one that is operated at or near its functional capacity with low variability. Day in and day out the process delivers as promised.

A manufacturing process that can routinely deliver the expected result is one that is considered “reliable”. The daily output number can be measured in output tons, widgets per hour, or really any other measurement that your organisation has adopted for production success.

The key to sustaining high levels of manufacturing process reliability is the ability to recognise when the asset is not fully healthy while it still may be producing “acceptable levels” of throughput. Next, you must learn methods to diagnose which portion of the asset or sub system is responsible for the degraded performance.

A well accepted process to analyse and visualise production reliability was first published by Paul Barringer. We will explore this process more thoroughly at the Reliability Summit on the Gold Coast where we will hear from the leading users of this method in Australia.

So, how reliable is your manufacturing process?  

We will use a plants daily production output to explore just enough of the process to tease you, but not enough to explain it fully, as that detailed explanation would take more space than this blog post allows.


This plant is reliable only about 50% of the time. Within this time of “reliable operation”, you should expect the plant to produce between 70% and 95% of the process maximum capacity.

That is a profound statement!

Conversely stated, 50% of the time you should expect this process to result in less than 70% of the capacity.

Equally profound.

The plant exhibits a significant loss when it is operating in an “unreliable” fashion. In fact, almost 30% of the capacity of the plant is lost each year.

ANALYSIS Specifics

This (Red Arrow) straight line represents how your manufacturing process has performed over time.

The (Blue) dashed line estimates on the vertical axis the “Reliability” you should expect from your process over time.

The (Yellow) dashed lines indicate the “range of throughput capacity variance” one should expect when operating reliably (~35% of capacity variance is expected).

Screen Shot 2014-09-18 at 10.36.53 am

Figure 2 – Exploded View Plant A Manufacturing Process Reliability

The area shaded in (Pink) accounts for 3.6% of the capacity loss. This occurs when the process is not operating reliably due to special causes that can include significant breakdown events.

The area shaded (Blue) accounts for 26% of the annual lost capacity. This is the difference between the inherent reliability and the demonstrated process reliability.

The area shaded (Light Green) accounts for 7% of the annual lost capacity which is generally considered “unrecoverable” due to the inherent reliability of the process.

The important point is that over 30% of a hidden plant exists within the existing plant structure which is not being converted to sellable product annually.

Join us on the Gold Coast for Reliability Summit 2014 to learn more about manufacturing process reliability. We will have industry leaders sharing their experiences and experts showcasing cool methods that you can use to advance your organisation with targeted improvements.

This question was posed to a discussion group and it got me thinking how do you grade an investigation?

The overall success will be whether the solution actually prevents recurrence of the problem.  One definition of Root Cause Analysis is: “A structured process used to understand the causes of past events for the purpose of preventing recurrence.” So a reasonable assessment of the quality of the analysis would be to determine whether the RCA addressed the problem it set out to fix by ensuring that it never happens again (this may be a lengthy process to prove if the MTBF of the problem is 5 years, or has only happened once).

Are there some other tangibles that can help you assess the quality of an RCA?  RCAs use some sort of process to accomplish their task. If this is the case then it would stand to reason that there will be some things you can look for in order to gauge the quality of the process followed. While this is no guarantee of a correct analysis, ensuring that due diligence was followed in the process would lend more credibility to the solutions.

What are some of these criteria by which you can judge an analysis?

  • Are the cause statements ‘binary’? By this, we mean unambiguous or explicit. A few words only and precise language use without vague adjectives like “poor” since they can be very subjective.
  • Are the causes void of conjunctions? If they have conjunctions there may be multiple causes in the statement. Words such as: and, if, or, but, because.
  • Is there valid evidence for each cause? If causes don’t have evidence they may not belong in the analysis or worse yet solutions may be tied to them and be ineffective.
  • Does each cause path have a valid reason for stopping that makes sense? It is easy to stop too soon and is sometimes obvious. For example, if a cause of “no PM” has no cause for it so that the branch stops, it would seem that an analyst in most cases would want to know why there was no PM.
  • Does the structure of the chart meet the process being used? If it is a principle-based process then it should be easy to check the causal elements to verify that they satisfy those principles. These might be causal logic checks or space-time logic checks or others that were associated with the particular process.
  • Is the chart or analysis completed? Does it have a lot of unfinished branches or questions that need to be answered or action items to complete?
  • Is the chart or analysis completed? Does it have a lot of unfinished branches or questions that need to be answered or action items to complete?
  • Are the solutions SMART (Specific, Measurable, Actionable, Relevant, and Timely)? Or do they include words like: investigate, review, analyze, gather, contact, observe, verify, etc.
  • Do the solutions meet a set of criteria against which they can be judged?
  • Do the solutions address specific causes or are they general in nature?  Even though they may be identified against specific causes if they don’t directly address those causes then it may still be a SWAG*.

If there is a report, is it well written, short, specific and cover just the basics that an executive would be interested in? Information such as cost, time to implement, when will it be completed, a brief causal description and solutions that will solve the identified problem are the requisites.

These are some of the things that I currently look at when I review the projects submitted by clients. I’d be interested to know about other things that may be added to the list.

* SWAG =  Scientific Wild Ass Guess