Improving Problem-Solving in Manufacturing

Introduction

Problem solving throughout many manufacturing organizations is broken. Downtime issues won't go away. Quality problems persist through several attempts to address them. Even the goals for improvement remain small - many plants have OEE improvement goals of just 2-3 percent for an entire year worth of effort.

In this article, we will talk about how to improve problem-solving in manufacturing. We will start with a quick look at current practices, then look at how they can be improved using smart factory solutions.

Problem Solving Processes

Toyota’s 7 Step Practical Problem-Solving Process

What is the problem-solving process? The graphic on the right side from Jeffrey Liker’s seminal book, The Toyota Way, walks through Toyota’s “Practical Problem Solving” process.

It begins with the general identification of a large issue within the plant. The next several steps work to clarify the nature of the problem itself, and then perform root cause analysis to identify the reasons the problem is happening.

At this point, solutions are identified and put in place, then evaluated and standardized.

An example would be where people feel there is too much downtime on a particular machine.

Solving the problem would then be clarifying the nature of the downtime, measuring the amount of downtime is being experienced, and locating the point of cause in the machine that is failing. The team would then perform a “5 Why” root cause analysis and determine countermeasures to eliminate the root causes of failure. They would then evaluate the performance after the fixes to see if they had the desired impact. At that point the fixes would become part of the standard work for that machine and potentially be replicated to other areas of the plant, where applicable.

Problem Solving Process Comparison

As we can see, there are quite a few “standard” problem solving processes used within manufacturing. There are many commonalities between the methods, and they largely contain the same steps – just grouped and organized a bit differently.

One question I often get is, “Which is the best (or right) technique?” While there are some differences between the different systems, they have more in common than they have differences. My recommendation is that if the team has already been trained on one technique and is familiar with it, then that’s the best one to use.

At that point it really becomes a question of how to improve the execution of the process and how to support the personnel as they proceed through each step.

2 Minute Lean

Another question I sometimes get is to differentiate between small continuous improvements and larger improvement events. In his excellent book “2 Minute Lean”, Paul Akers makes a compelling case for an emphasis on the small improvements that only take minimal resources and time to make incremental, repeated steps in performance.

While the primary focus for today’s presentation is with larger issues that arise in manufacturing, the concepts we will cover about smart manufacturing solutions can apply to these quick changes, too.

In this case, the focus is on giving operators what they need to implement small improvements at their stations:

• Context about performance at the station and its impact on the overall flow
• Summary and detailed information about the problems (who, what, when, why, etc) in a way that’s quick and easy to consume
• Measurement and feedback on whether the change was successful

Broadly, everything we will cover for larger projects applies here, as well. It is simply more focused and formatted for small, continuous improvements

Steps in the Problem-Solving Process

Next, we will look at the steps in a typical problem-solving process and how those can be impacted by smart manufacturing solutions.

Combined Steps in the Process

In the problem-solving process we’ll walk through for the rest of the webinar, I’ve simply combined the steps from the different practices into one list. This is not meant to be a recommendation for yet another workflow, but to show how smart manufacturing solutions can support all the steps of each of the practices.

As we walk through the steps in the following slides, we will see how they can be improved through:

• Historical performance information that captures production rates, uptime & downtime information, quality data and much more
• Detailed process data showing pressures, temperatures, vibration information, PLC tag data, etc
• Thorough contextual information for each one of these events: Who was running the process, what was being manufactured, time of day, day of week, and much more
• Augmented intelligence capabilities through artificial intelligence & machine learning
• Well-designed workflows that help users through the problem-solving steps
• Structured communication platforms that keep the team on the same page and facilitates group productivity
• Advanced data visualizations to help make the information easy to understand for all participants

Step 1: Initial Identification of Problems

“The first step is admitting you have a problem.”

On any manufacturing floor I’ve ever been to, there has never been a lack of issues to be addressed. There are always problems to fix, processes to improve, and so forth.

What smart factory solutions help with here is to understand the scale of the problems. The historical performance information shows exactly how big each problem is, how they are trending, and more. This helps to determine which should be the highest priority issues to work on. The systems allow the creation and tracking of a project funnel for the facility.

Instead of simply chasing the last major event or the squeaky wheel, the team can prioritize the most impactful work to be done.

Step 2: Form the Team

Once it is decided which problem is going to be addressed, a team must be assembled. While smart factory systems won’t help with forming the team itself, they are fantastic for facilitating teamwork and communication.

• Well-designed workflows help users through the problem-solving steps, showing the relevant information at each step, tracking inputs from multiple users, and consolidating information in a single platform. Most projects today have information spread through multiple systems and, even more, in multiple excel spreadsheets that have access and version control issues. All that can be eliminated.
• Structured communication platforms that keep the team on the same page and facilitates group productivity. In addition, modern platforms facilitate communication by integrating tools like Microsoft Teams or Slack. Instead of having communication happening across multiple emails and spreadsheets, communication can be facilitated right into the workflow and analysis platform.

Step 3: Clarify and Describe the Problem

One of the key benefits to these systems is the contextual information captured for each one of these events: Who was running the process, what was being manufactured, time of day, day of week, and much more

One of the most basic methods of finding the cause of problems is to ask the five different “W” questions (plus one “H”). I believe I learned this one all the way back in elementary school. When you want to something out, ask these basic questions to get to the bottom of things.

This method can be very beneficial if done correctly. For example, if you are preparing to create charts of possible causes or perform a statistical analysis, this method can help you identify possible causal factors. It can be used as a brainstorming tool to be somewhat comprehensive in the approach. However, it tends to lead to simplistic analysis as it does not try to get the practitioner to drill down any further than top level (or proximate) causes.

For example, it is not enough to know that my first pass yield went from 97% to 93%. Just knowing that my FPY dropped is not enough to take action to fix it. The information is not actionable. It is a good start, and I will know that I need to pay attention to it. That is good to know! But until I know why it went from 97 to 93, there is nothing I can really do about it. To address that in the old days, we went out there with a clipboard and stopwatch and we started watching the line for what was causing the defects. We used tick sheets collect information about what things were happening, where they were happening, who was involved and how frequently they were happening. These methods of data collection not only lacked accuracy, but they also lacked detail, they were frequently biased, and they were always wasteful.

There are better ways to do that now.

When performing this analysis, Industry 4.0 provides capabilities to automatically capture accurate information about each of these potential cause categories. Keeping with the example about first pass yield, information can be captured directly from the testing machines on *how* the part failed the test. The systems can know *who* is logged into the stations. By knowing what machines were used for which parts, the system can know *where* the problem occurred. They will also know *when* that defect occurred. Finally, by tracking the detailed process variables as the part was processed, the system can track *what* happened during the production.

Using this data to understand what high-level causes are correlated to the problem can identify areas for further investigation.

A quick note on integrating artificial intelligence or machine learning systems into this process…

Up above, we talked about the benefits of the contextual information allowing the team to better clarify and describe the problem. It is also possible to augment the team analysis with AI capabilities.

Classifier systems can automatically group similar events across a range of different dimensions. Other types of AI can perform other tasks automatically, as well.

This means that the team does not have to do all the heavy lifting with the analysis – the software can help them out with that process.

Step 4: Contain the problem

By understanding what things are contributing to the problem, we are now in a position to implement a short-term fix. While we should not be satisfied with putting a band-aid on the problem, this can contain quality or safety issues while a permanent fix is sought.

Having a good analysis prior to this point will help determine the proper containment measure to implement.

Step 5: Root Cause Analysis

There are many frameworks for performing root cause analysis. In other webinars, I’ve gone through details on each of these different methods. Today, I’ll give more of an overview how each can be impacted, as well as drilling deeper into a few of the methods discussed here.

• 5W & 1H – while this was discussed earlier in regard to properly clarifying and describing the problem, this method is sometimes sufficient to get at root causes for performance.
• Graphical Evaluation – In this method, data about the progress is put into graphical form where the users can evaluate it visually. At Visual Decisions, we are big fans of this type of analysis! Smart manufacturing systems are fantastic for this type of analysis. Because they are so data rich, most of the information you’d like to know about the process can be charted and shared for analysis.
• FMEA – In another webinar, I’ve discussed how to create an “ever-green” FMEA (failure mode and effect analysis) that is continually updated with information from the shop floor. Based on the symptoms observed on the shop floor, this tool can be used to see if there are known causes for failure to investigate.

We will look at the rest of the methods below.

5 Why

The most common type of root cause analysis done in lean manufacturing environments is 5 Why analysis.

In this method, we keep asking why until we get down to the root cause for the problem. In the example shown, our problem is that the customer is refusing to pay for the leaflets we printed for them. Our top-level reason is that the delivery was late, so the customer could not use the leaflets. Instead of settling for that answer, though, we continue to ask why. In this case, the next question is “Why was the delivery late?” Eventually we reach the root cause of the problem that we ran out of ink and couldn’t order additional supplies in time. At this point, we have reached a cause that we can act on!?We can put in a countermeasure to find a backup ink supplier we can use in emergencies so that we can continue to keep our normal stock minimized, but have a place to go for more during periods of surge demand.

The biggest trap I’ve encountered with this method is that people only ask why once at each level. In the example given, we only have one answer to each why question. In reality, there is usually more than a single cause for each question. Instead of ending up with a straight line, you should end up with something that looks like a tree.

The other key problem that I’ve seen is that this is usually not done with data. It is usually done by gathering people into a conference room and asking people to go through the process anecdotally.

5 Why analysis can be greatly facilitated by having data at each one of these levels. In addition to the traditional brainstorming, the team can investigate the data to see what it shows as each question is asked.

Fishbone, or Ishikawa, Diagram

In this method, you show the problem (or effect) to the right side of the diagram (shown in the triangle on the right side in the inset picture). Then you try to determine all the contributing factors to the problem. It is typical (but not required) that the first “branches” in the analysis are man, machine, material, method, management, and environment. Then you ask what it is that is causing the issue. Continue asking why until root causes are reached. Essentially it is 5 Why’s with a structure that encourages asking “Why” more than once at each level.

This process is usually done in a highly manual fashion that is completely disconnected from the process.

Another weakness is that while the diagram gives an idea of contributing causes, it does not show how much those causes contribute to the problem.?When data collection is performed, it is usually done manually after the initial analysis has been completed. One way it is done is to give the operators tick sheets to note how often a failure on each of these branches occurs. The other way is to have someone go out to watch the process and record that information so that the operator can continue their work.

One problem with collecting data by watching the process in person is the Hawthorne Effect. People behave differently when they are being watched – think of how drivers slow down when there is a policeman on the side of the road. This leads to the problem of bias in the data – the data that gets collected does not accurately reflect how the process normally operates.

With Industry 4.0, we always collect data. This allows us to create a pareto diagram of how much each of the causes is contributing to the issue.?Then we can begin the Kaizen event with a very strong data set on all of the different causes contributing to the problem and what is most important to address first.

Statistical Analysis of Factors

Finally, we close out this section with a brief discussion of the gold standard of data-based root cause analysis. Some of the forms of statistical analysis that apply here are:

• Analysis of Variance (ANOVA)
• Design of Experiments (DOE)
• Linear and Multiple Regression
• T-Tests
• Chi-Square tests

These methods are at the heart of Six Sigma analysis.

It is beyond the scope of this particular webinar to dive deeply into each of these statistical methods. I do want to cover a few important points, though. These generally represent “Tests of Hypothesis” where the hypothesis represents a potential cause for the problem. In many cases, the preceding methods can be used to identify the candidate causal factors for the problem, while these methods can be used to confirm whether those factors are truly correlated or causal to the outcome.

Obviously, these methods are also very heavily dependent on data. The more data, the better. The richer and more contextualized that data can be, the better these methods will work.

Step 6: Create Solutions

Another big impact of Industry 4.0 is to broaden the solution space. For many problems in manufacturing, the fix will be some alteration to the physical equipment involved. Going back to the fishbone diagram, these fixes primarily relate to the “Machine” branch. However, when the problem is related to Man, Material, Method, Management or Environmental the fix is often process or system based.

Just as a quick example, many times the problem can be traced back to a lack of standard work or a lack of adherence to the standard work. In all those cases, smart factory systems can be an integral part of the solution. In many quality-related problem-solving efforts, technology can be a key part of error-proofing, automated inspection, and other solutions.

For these issues, the technologies of Industry 4.0 can provide tremendous benefits.

There are additional ways that Industry 4.0 solutions can help create solutions.

The usual way to go about creating solutions is to look at many different alternatives. You want to brainstorm and utilize cross-functional teams. You can incorporate vendors, partners and customers when beneficial. This may mean getting input from people at remote locations. In 2020, it can help to have augmented reality capabilities to get remote visibility into the process for those external experts.

Another key capability is being able to track what has been done before and utilize those best practices. Being able to see if this machine has had this same type of failure before, what did we do then? Was it successful? How long did that fix last? Why did it break again? Was there a fix done on a similar piece of equipment that worked better or lasted longer? For maintenance repairs, this information might be in a typical maintenance management system (MMS). However, there must be a way to capture changes like a process fix, additional training, changes to standard work, etc. That way there can be one central place to look for potential solutions that have been tried in the past.

Another key trap to avoid is, “That’s just the way the line works.”

As an example, I worked with a customer to improve a shampoo bottle filling line. In the data I noticed that there were a lot of short stops attributed to the filler. When we drilled into the data, we found that the overhead bottle feeder to the filler was jamming. The current solution to the problem was for the operator to grab a broomstick and poke that until the jammed bottle until the line started feeding again.

When I discussed this with the team, they were a bit defensive about it. They said it only took a few seconds to fix the jam when it happened. They said that the line just worked that way!

I led them through the data to show the real impact of this problem. That jam was sometimes occurring over 20 times a day. It was true that the operator could clear that jam in a few seconds – they got very skilled with that broomstick! But it was a long line with a single operator. The filler would often be stopped for a few minutes before the operator would see it and get to the jam to clear it.

That silly little jam was actually the single biggest source of downtime for the process.

We gathered the process engineer, the operator, supervisor, and the maintenance team to the line to brainstorm solutions. We put in two key fixes. The first was to fix the alignment of a join in the feed that was causing the bottles to jam. The other was an andon light to let the operator know right away when the filler (the gating process for the line) stopped for any reason. Problem solved!

Step 7: Testing the Chosen Solution or Solutions

If you are implementing technology as part of your solution, you want to make sure that you adapt the process and provide adequate training to people for that process change. Simulation provides another key capability here from an industry 4.0 perspective. If the potential solution is something that requires a capital investment, simulating that solution in advance of making the investment can provide evidence that the solution will have the desired impact.

A large portion of what we work on at Visual Decisions is implementing I-IoT solutions within our customers. These platforms have tremendous flexibility to address common (and uncommon!) issues in manufacturing. These solutions tend to have some development time associated with them. A big recommendation is to utilize Agile approaches to development of these solutions. Agile development techniques were based in part on the Kaizen philosophy of continual, stepwise improvements. For testing the solutions, once the minimal viable product (MVP) has been produced, it can be tested to ensure it is going to fix the problem it is supposed to address. Only then can additional development work be committed to enhancing the solution.

Another technology that should be noted here for testing is Augmented and Virtual Reality. Not only can these technologies be a solution to many manufacturing issues, but they can also be used to test out a wide variety of other solutions. They can be used to give insight on how workers may interact with those potential solutions before making heavy capital investments putting them into place.

Step 8: Measure and Analyze the Results

Much of this webinar has been about utilizing data effectively to identify and analyze issues. This is yet another place within the problem-solving process that acquiring and analyzing data is crucial.

All the topics touched on earlier apply here, as well. In particular, having a system to collect the data 24/7 in an unbiased fashion is critical. Because of the bias and inaccuracy inherent in manually collected data, measuring the success of the solution in that fashion will lead to poor results. In addition, it is usually people that have a vested interest in the solution that would be collecting the data. That makes the situation even more likely to produce biased results.

For a variety of reasons our preference for data collection is to use an IOT platform. First among those is the inherent flexibility. In many situations, an MES system may be most appropriate for gathering the bulk of the data from the shop floor. Our recommendation when evaluating MES systems is to focus on the flexibility and adaptability of those systems even after implementation. Another approach is to augment the data collection capabilities of the MES with an IOT platform to handle some of those edge cases. Overall, the IoT platform allows the agile philosophies to come into play where you can very quickly develop and deploy a solution during a Kaizen process. Since Kaizen is an ongoing process of improvement, adding additional capabilities in this way will improve the overall system over time. This will make it a better and better fit, as opposed to a more monolithic solution.

One of the dangers of using technology to solve issues is the creation of “monuments”. In lean manufacturing, a monument is something that cannot be moved or changed. All other solutions must work around that monument since it cannot be modified or adapted without great effort or expense. A typical system example of a monument is an ERP system. Once they are implemented, it is usually a very difficult process to make changes to the way the ERP system operates. While ERP systems are important to the functioning of modern corporations, they often present roadblocks when it comes to continuous improvement. It is often more effective to extend the functionality of solutions such as ERP or MES with other applications that provide the desired flexibility.

Step 9: Standardize the Solution

In the previous step, I touched on the dangers of systems becoming monuments in the future. However, I will mention here that one of the benefits of software is to standardize the solution. By having a common interface in the software across shifts, machines or even plants, software can help to “systematize” the solution across the company. These systems can also be used to document standard work and share that among the operators. In addition, deviations to the new standard work can be tracked and used as learning opportunities with the staff.

In fact, I cannot emphasize enough the key role of standard work here. When we think about standard work in manufacturing, the first thing that pops into mind is getting all the operators to work with the equipment in the same way every time. And absolutely that is part of the solution to many problems.

But we also must think about the leader standard work. Setting the culture and making a solution part of the process means having the team leaders, supervisors, value stream managers and executives all be a part of the solution.

For example, at several customers we have created a dashboard that acts as a management operating system for the company.?Using the solution on the COO’s weekly cadence call is really what is going to help standardize that solution. This solution not only showed the current values of all the KPI, but also tied into all the improvement projects being run across the sites. This was used to track progress, knock down roadblocks – and to ensure that once solutions were put in place that they became standardized within the organization and best practices were shared across facilities.

Because he used that to drive communication with his plant managers every week, that put the pressure on them to understand the information in that application to ensure the accuracy and completeness. In turn, by using the solution in their meetings with their value stream managers, they ensured the value stream leaders used that solution as part of their daily and weekly processes. That emphasis made the solution part of the standard work that happens on the shop floor shift by shift, day by day, and week by week.

Closing thoughts

Data is absolutely critical to identifying the improvement opportunities, analyzing causal factors, generating solutions and more.

Caution must be used to avoid technology monuments. Do not put something in stone that you will not be able to change later. Situations will change and you will want that flexibility going forward. Choose solutions that are easily adapted and improved.

Finally, when developing solutions on platforms such as I-IoT, use agile concepts to implement incremental improvements and test those solutions when reaching the minimal viable product stage of development.

If you've gotten this far, congratulations! This subject must be important to you to have invested the time to read the whole article. Please share your thoughts in the comments.

?