Lean manufacturing is a collection of principles which align to accomplish one major goal — identify and minimize non-value-added activities during the production process. Phrased another way, lean production is about “doing more with less” while remaining highly responsive to customer demand, and its practices have been influential and widely adopted throughout a range of manufacturing industries, including automotive, electronics, aerospace, and elsewhere.
The precursor of lean manufacturing lies in Japanese industrial engineer and businessman Taiichi Ohno’s work on the Toyota Production System (TPS) in the 1980s. Ohno’s approach centered around taking more measurements of part quality, and at more points throughout the entire production line, than other car manufacturers were doing at the time. The benefit was discovering errors in parts and processes before attempting to assemble the vehicles, which resulted in savings of time, money, and materials, as opposed to sending partially assembled vehicles for rework or scrap. In addition, Toyota adopted a general mentality of continuous improvement through measurement paired with reflection and action. Taken together, these mentality shifts quickly put Toyota on track to becoming a global powerhouse in automotive manufacturing.
Implementing lean practices requires both upfront and ongoing effort, but the rewards include increased productivity, increased quality, minimized scrap and rework, and, of course, reduced production costs. The downside is that implementation has typically required human experts (manufacturing engineers) monitoring manufacturing processes and investigating potential improvements hidden among the data, which may take weeks of effort from a highly limited pool of available talent.
Artificial intelligence (AI) and machine learning, on the other hand, have the potential to quickly and organically build models with available data. When algorithms are developed to do the process monitoring, humans can focus their time elsewhere. If something about the process needs attention, the right people can be provided with actionable insights. AI and machine learning are poised to turbocharge lean manufacturing as their continuously-improving models translate into continuously-improving operations on the factory floor.
The rise of AI and machine learning
AI is poised to become a revolutionary technology for businesses of all sizes and industries, especially manufacturing. Some manufacturing companies have already applied AI to improve their processes, for example, by slashing supply chain forecasting errors in half.
AI generally refers to the study of how to make a computer act more intelligently in pursuit of a defined goal, and machine learning is a particular application of AI that gives computers the ability to automatically learn and improve from experience without being explicitly programmed to do so. In other words, this type of system consumes data and transforms it into expertise in the form of statistical models on a level beyond that which humans can easily grasp. Many of the concepts involved are similar to those you can find in the statistical functions of spreadsheet programs, but the power lies in the sheer scale at which the variables and inputs of the model can be linked.
AI and machine learning techniques for inspection and improvement have already been applied in other areas to great effect. For example, botanists have used AI to build “smart greenhouses” that benefit from autonomous control, planning, and monitoring. Meanwhile, machine learning researchers are investigating the use of AI to detect anomalies on train rails in order to stop trains in advance of collisions.
Measurement is the first step in improving any process, but the following steps are not always obvious, especially in a complex production environment — which is where the autonomy of AI can help.
Three ways AI can turbocharge lean manufacturing:
1. Automated QA
Feedback from automated testing and quality assurance (QA) equipment allows AI systems to build statistical models of what normal production “looks like,” giving it the means to identify errors when anomalies occur in a sensor or measurement. Then, because the AI system has access to data from the whole facility rather than just one operation, links can be made to data from upstream in the production process, allowing it to reason about the root cause of the error and preventive/corrective actions.
These insights can be highly beneficial to your overall equipment efficiency (OEE), productivity, and accuracy. You might discover, for example, that a variety of parts which tend to drift out of spec all spend a brief amount of time on your machining centers with the same type of roughing tool. Once that similarity is automatically flagged, it allows you to scrutinize those tools and realize they’re more prone to “built up edges” when cutting your material than originally thought. You can take immediate action such as switching to tools with a different coating, changing your cutting fluid, or altering your process parameters (more on this in point #3 below). The deficiency could have been found through an investigation and deductive reasoning by your manufacturing engineers, but the key advantage AI provides in this example is speed by giving engineers empirical guidance on where to look first.
2. Predictive maintenance
Equipment failures are an unfortunate part of manufacturing, and unpredictable failures generate the last two words anyone wants to hear — unplanned downtime. As complex machines become more difficult to keep track of and pressures mount to meet your production quotas, it’s more likely that you’ll experience a breakage or machine failure that can bring your operations to a screeching halt.
Through the power of machine learning and observing data from sensors in, on, or around a piece of equipment, it is possible for AI to learn signs of when a machine is likely to fail and then schedule preventive maintenance into the facility’s production schedule. This keeps downtime planned and to a minimum. A simple example is monitoring the vibrations of a piece of rotating machinery over time. The machine produces vibrations when it is healthy and also when it is approaching failure, but the key is that the vibrations are different when failure is near. Placing an accelerometer on the machine allows an automated system to perform spectral analysis on the vibrations over time and learn what healthy operation “feels like”. If the vibrations start to become out of the ordinary, someone can be alerted so production can be safely paused for preventive maintenance before any severe failures can occur.
According to consulting firm McKinsey & Company, predictive maintenance can reduce machine downtime by 30 to 50 percent and increase a machine’s lifespan by 20 to 40 percent.
3. Process optimization
AI and machine learning can help establish a feedback loop between planning and execution, continuously improving your manufacturing processes. The factory simultaneously transforms into a laboratory, with every production setup being a new “experiment” that is automatically documented and analyzed. As you manufacture more parts, the insights that you discover will improve, as will your processes and part quality.
Back to the machining example, imagine that your high-mix/low-volume machining centers have cut a variety of stock materials into many different part geometries, which implies many toolpaths and cutting strategies. When those past process plans are coupled with data from automated inspection, models can be generated which predict what optimal cutting strategies, feeds, and speeds might be for the given tool-material combination of your next order. If the next order involves a particular steel alloy you have not yet gathered data on, the predictions can be based on similar alloys as a starting point.
In addition, quoting and forecasting becomes more scientific as the system gathers more data about historical costs of different operations and production steps in your factory.
The time has come
With the benefits of lean manufacturing already established, it’s time for more companies to adopt AI and machine learning to make their processes more efficient, productive, and accurate. The early movers will define the future of lean manufacturing, and AI will play a leading role.