Why SPC isn’t enough in 2022
Most precision manufacturers rely on software to automate or measure specific operations. For example, most automotive manufacturing plants use some form of Statistical Process Control (SPC) to measure and control processes and production methods.
What’s interesting, however, is that many of these same facilities seem to fear newer technologies like machine learning. Even when they see the value that newer technologies provide, they don’t necessarily understand the science behind them and that makes them wary.
In this blog, we provide basic definitions of both statistical process control (SPC), which is already familiar to you, and machine learning, which might be something you’re just starting to explore. You’ve probably already guessed that we’re biased towards machine learning solutions for adding value to manufacturers: we do specialize in artificial intelligence and machine learning solutions that help precision manufacturers improve part quality after all.
What is SPC?
We’ll start with the easy one: statistical process control. It’s a cost-effective process control technique that is especially useful for machining operations.
Manufacturers typically use SPC to analyze processes of key features that need to fall within a certain threshold. The idea is to capture any statistical variation in a process by establishing a reasonable sampling quantity and frequency based on a capability or process study. This allows operators to react to issues if X-bar and control charts begin to show evidence of statistical deviation.
SPC is an integral part of most manufacturing facility’s technological ecosystems. However, it has limitations. SPC requires manual interpretation and analysis, meaning that operators spend a lot of time monitoring the lines and also filling out charts. There is much manual work that needs to be done for SPC to be executed correctly.
SPC also has limits in terms of what it can detect, predict, or analyze. These limits are exactly where machine learning can fill in the gaps.
What is machine learning, and how does it differ from SPC?
Machine learning is a type of artificial intelligence (AI) in which computers perform complex processing, reasoning, and decision making on data. Essentially, machine learning refers to the type of AI that can learn from and adapt to new data without human intervention.
Machine learning is capable of performing data analysis that’s impossible for humans to perform due to data volumes, complexity, and the speed in which it needs to be analyzed. In most manufacturing operations, companies collect a surplus of data and much of it is not being utilized to its fullest capacity. Machine learning can help these companies make sense of their complex manufacturing data.
Machine learning automates many of the analytical steps in assessing manufacturing signal data that would typically need to be performed by humans manually using SPC as a foundation. Machine learning can uncover anomalies and draw conclusions in minutes, compared to hours and days that it may take humans to analyze the same amount of data.
Machine learning supplements traditional SPC
Machine learning and statistical process control (SPC) are not mutually exclusive. In fact, machine learning has been proposed as a method for augmenting the construction and interpretation of SPC charts and we couldn’t agree more.
In this post, we focus on one small portion of the manufacturing data puzzle: the data collected by machines regarding part and process quality, i.e., the data that is commonly analyzed with SPC.
Even when a company has dozens of SPC charts showing good statistical control and in-process data, they still need to run end-of-line tests. The fact that every pump, torque converter, transmission, and other part also requires an EOL test to ensure its quality should indicate that SPC alone is not a perfect solution to reducing scrap and rework in your plant.
Machine learning won’t, and shouldn’t, replace SPC checks, but it can help you answer questions like, “Why do my parts fail my EOL tests if they are all technically within specification limits?” You might wonder why we can’t just use SPC on EOL tests (i.e., apply statistical tools to aggregate data) to get the same insights that machine learning could offer.
Machine learning manufacturing software solutions like our LinePulse product look at the relationships between signals in tandem. Unlike statistical process control, machine learning doesn’t just look at the trends of one signal to see if it’s trending towards a control limit. Machine learning, with its complex computing power, can find connections between the time series of the signals themselves and, more importantly, between each other.
A single transmission can output over 500k time series data points in an EOL test. The value of machine learning is in finding the relationships between all these data points and comparing them to the data from all the other transmissions to find common abnormalities. This advanced ability to detect patterns and anomalies in relation to one another is unmatched by any current tools or manual intervention.
Can you imagine one person parsing through 500k datapoints and attempting to draw parallels and understand the relationships between them?
Even if you were to employ SPC techniques on some of the more important signals (which necessitates identifying which signals are important), you would only know if you were trending in a certain direction around your control limits. But that does not necessarily mean you’re dealing with a defective unit.
Looking at all the connections between signals rather than focusing on the direction in which a single signal is trending over time is what differentiates machine learning from SPC. It’s also what enables machine learning to yield actionable insights. SPC can only tell you that your process is trending in a certain direction but not what to do about it.
From an automotive industry perspective, it’s best to use SPC under very specific conditions, e.g., if you have a feature on a part that affects assembly downstream but you do not feel the need to gauge this feature 100% of the time based on a capability study. You know that your process is relatively capable on this feature (between 1.0 to 2.0 Cpk) so you employ SPC and have a sampling frequency to keep the feature within statistical control to avoid assembly issues downstream.
With machine learning on the other hand, you have a complex model that understands your entire process, which helps provide insight into root cause and further protect your product from the process, in addition to SPC.
SPC and machine learning are complementary
Machine learning is the logical next step in response to the growing volume of manufacturing data that most plants generate. In the same way that the invention of the Internet built upon and expanded the capabilities of the computer, machine learning adds exponential possibilities to draw connections, detect patterns, and connect seemingly disparate data together to form a whole picture of your production.
Software such as LinePulse integrates data from SPC with the power of machine learning. The development of these types of cloud-based SaaS solutions present the next generation of manufacturing analysis, and with their expanded capabilities, create a new category of quality management.
Are you ready for the next generation of manufacturing analysis?
If your plant’s process checks tell you that your parts should be good, but you’re still running into assembly problems downstream, it might be time to supplement your SPC with machine learning.
To put it another way, if you’re dealing with univariate data in a normal distribution, SPC might well be sufficient for your needs. On the other hand, if you’re dealing with multivariate data that does not conform to a normal distribution, you might need machine learning.
Generally speaking, the prerequisite for machine learning is having a large data set, and these days most manufacturers do. You could try to apply traditional statistics in such cases, but you also need to remember that the data was generated without any kind of statistical experimental design. So, when you draw your conclusions, you’ll need to account for the fact that your data is imbalanced. The high resolution that comes with big data means that any small difference might seem significant, even when it’s not.
In addition to the theoretical issues, there’s also the actual practice of SPC to consider. As with any task involving a human element, SPC runs the risk of being poorly executed. In many cases, it’s quite literally an operator with a calculator and some graph paper, sampling parts, calculating the average and looking for trends. Compare that to machine learning, where data that’s being generated by the production line is already processed and visualized to provide additional value.
LinePulse combines machine learning with SPC data and visualizations to do complex data analysis for you, without the calculators and graph paper. The cloud-based software connects your manufacturing lines and allows you to keep a constant view on your manufacturing data from one platform. It detects anomalies that predict future quality issues, so that you can intervene early in the process and prevent scrap and rework down the line.
If you are interested into furthering your Industry 4.0 initiatives and supercharging your SPC, read more about LinePulse.
Share on social: