Using IoT And Machine Learning For Production Optimization
Optimizing manufacturing processes for efficiency can have a significant impact on your bottom line. Production optimization refers to the set of initiatives that is aimed at driving this efficiency. Any action that reduces waste throughout the production cycle – such as reducing Takt time or optimizing first pass yield, can contribute to production optimization.
Production optimization is rarely a one-off effort towards a short-term objective but rather an ongoing set of actions aimed at delivering business goals. These long term objectives create a considerable competitive advantage by reducing the cost of manufacturing, delivering better profitability and increasing the number of products produced per unit.
Find the following in the read below: What Is Your Optimal Point Of Production, IoT For Production Optimization, Machine Learning For Production Optimization, AI For Production Optimization, Get Closer to Product Optimization Today.
What Is Your Optimal Point Of Production?
Optimal production level is the ideal output level where the marginal revenue derived from a unit sold roughly equals the marginal cost to produce it. Profits can be maximized at the production level where the marginal revenue gained from selling one additional unit equals the marginal cost to produce it. A business should continue to increase output as long as its marginal cost is less than the marginal revenue gained from selling the product.
Assume you want to maximize your profits as a small coffee mug manufacturing plant and are studying all the competing factors involved. Suppose your market climate accepts a $10/unit price. That number allows you to calculate the cost to produce one additional mug and therefore estimate the number of mugs you can produce. Let’s say an additional mug cost $9.55 with a $0.45/unit profit – this is sensible! However, if it costs you $10.25 for an additional mug with a loss of $0.25/unit, it would be economically inefficient to manufacture this additional uint.
Earlier we talked about marginal revenue and marginal cost. So, from the above example it is clear that the marginal revenue is the fixed market price ($10.00), or the revenue gained by selling the mug. The marginal cost is the cost involved in producing the next much and is helpful in deciding whether or not to continue production.
Maintaining the marginal cost levels lower than the optimal production level can be influenced by a wide variety of factors. Reducing fatigue driven errors and inefficiencies through pick and place robots can improve throughput and hence optimize cost of production. But, so can route planning combined with ergonomic jigs and fixtures guided by intuitive assembly instructions for floor labor. Similarly, a firm can choose between hiring personnel to haul supplies around a factory in carts and forklifts or investing in guided vehicle robots. Hence the optimal point of production can be a subjective affair and their implications vary vastly from factory to factory.
How To Use IoT For Production Optimization
In the words of Lord Kelvin, “That you cannot measure, you cannot improve.” The first step towards improving production efficiency or optimizing the production process is to measure all influencing parameters.
With the advent of IoT and low-cost sensors, it is now possible to gather and measure intelligence from different aspects of the production environment. When combined with traditional data gathering systems like SCADA and DCS, this produces volumes of information. This information can be effectively used to take decisions and implement initiatives that will drive production optimization.
Using IoT, production can be optimized in several ways and at different levels of the ISA 95 framework. For instance, OEE can be optimized at the node level such as a specific motor on a machine. IoT extends the scope of data gathering and data handing over unimaginably wide areas eliminating the distance barriers that constrained DCS and SCADA. This centralization can be achieved at the plant level by optimizing routing as well as the enterprise level through strategic initiatives like Kanban, 5S or Lean manufacturing.
With the help of IoT it is now possible to observe and respond to production environment stimuli from remote locations. IoT is powered by the internet and hence proximity is no longer compulsory for operations, With the correct infrastructure and provisions in place, IoT sensors and actuators tied to smart phones create endless possibilities for production optimization, eliminating constraints of vicinity to ensure production efficiency.
Health And Safety
One of the most used applications of IoT is the identification of possible operator fatigue. Aspects like position of the operator with reference to potentially hazardous equipment or environment, and the relative ergonomics of machine usage in a production environment can be closely monitored. All these parameters can be easily tracked with data from IoT wearables like belts, cuff and rings used by factory personnel. These wearables not only alert potential health hazards, but also come with situational alerts or feedback mechanisms that can notify the user or operator before incidents occur.
If an operator becomes fatigued in the middle of successive shifts, an automated workflow will detect closing eyelids or nodding heads. This detection will then automatically trigger a vibration to a wearable wristband or alert the line manager of the floor personnel’s fatigue.All of this is possible through the power of IoT enabled wearables and guide frameworks of safety that are accessible through cloud.
IoT embedded devices not only enhance safety but also empower manufacturers to embrace the future of smart manufacturing.
In-line or end-of-line IoT sensors can detect deviations from specifications of WIP material allowing for agile in-process changes. This can greatly help reduce wastage and end-of-line scrap. Hence monetary savings are achieved by reducing waste and eliminating labor, energy and other resources consumed in wasteful processing of off-spec material.
How To Use Machine Learning For Production Optimization
Humans are able to learn from mistakes whereas machines or computers strictly do what they’re told to. A computer will continue to execute a routine or procedure as many times as instructed regardless of the validity of outcome. In other words, computers work along the lines of ‘if-then’ and ‘do-while’ loops and require detailed step by step instructions on exactly what actions to take and not take. Now, this is where machine learning comes into the picture. Machine learning is a way of getting computers to learn from the data of past experiences. It provides machines the ability to learn and improve from history without being programmed each time.
In the manufacturing sector, ML allows manufacturers to uncover hidden insights and enable predictive analytics. The insights drawn from these analytics are invaluable in predicting the Mean Time Between Failure (MTBF) of machines and equipment. An early prediction of downtime can greatly help plan for redundancy and continuity. Machine learning finds a variety of such applications in the modern factory.
Historians, distributed control systems, SCADA and all other data gathering systems create volumes of historical information about the production environment. This combined with the power of Machine Learning can deliver useful details that can be used to train machines to predict potential future failures.
The key prerequisite for a true predictive maintenance application is to have enough data. The rule of thumb is you need ten times the number of variables you are looking to predict. This means that a pump on a machine will need to fail ten times before machine learning can predict that pump will fail. Gathering this data is time consuming and often not readily available. Condition-based monitoring; however, monitors operating conditions and alerts operators to any abnormal scenarios including low pressure or high temperatures. This approach can accelerate your time-to-value with a predictive maintenance solution.
Digital Thread And Simulations
Machine learning can be used to train engines or algorithms to gather information and develop a digital replica of the manufacturing environment. This replicated environment can be used to run simulations for multitude of scenarios such as load bearing capacity, exploring lean manufacturing options, studying crisis handling and incident response, to mention a few. These simulations can help prepare for a scenario long before it occurs.
Agility In Response To Fluctuations
Matured manufacturing organizations have historic information about capacity utilization and its dependence on market demands. The connectivity between enterprise applications like CRM, ERP, SCM and MES have an inherent lead time because of interdependence.
Assuming the market demand and consumption behaviors are changing rapidly, there will be an impact on the orders in the CRM. The data from the CRM will then impact the ERP, which will in turn impact MES. This will eventually reflect in the production instructions for the factory.
Depending on the lead time and amount of throughput, there arises a possibility of surplus or deficit in finished goods. This can have undesirable results such as unsold finished goods or unrealized sales.
Hence, it is possible to simulate historical data through machine learning algorithms to develop and detect potential fluctuations in demand. This ability gives more real time manufacturing intelligence to make quicker decisions.
Information from machine learning algorithms can also predict peaks and troughs in demands. This intelligence can be used to plan resource allocation accordingly.
Machine learning can help understand potential bottlenecks in plant routing and can act as a decision support system for the production manager to decide how to balance the load across different lines.
When volumes of data are consistently tracked through machine learning algorithms. It tends to capture information around potential deviations that are normally not visible to the naked eye. Minor variations in aspects like turning shaft, feeble fluctuations in pump output and anomalies in the energy consumption patterns can easily go unnoticed. Algorithms can be trained to identify such deviations and suggest interventional or recalibration activities in a timely manner to prevent wastage and avert potential incidents.
How To Use AI For Production Optimization
The fairly recent regard and recognition that AI (artificial intelligence) has been receiving makes it easy to assume that AI is a new discovery. In fact, the concept of AI has been around since the early 1950s, almost a decade ahead of the production of “Star Trek: The Original Series”. The lack of technology available then had it shackled to the shelf of “interesting ideas”.
Operators today continue to heavily rely on their experience, intuition and judgement. This reliance on experience makes it difficult to scale and replicate the wisdom of such operators. The variations in operators’ experience and qualification can impact both performance and outcomes. This makes AI’s ability to retain, enhance and standardize knowledge all the more important. AI’s ability to aid making operational decisions can be leveraged to drive predictable and consistent outputs.
With the growing volume of data in the manufacturing environment, AI tools and ML platforms no longer confine their applications to just visualizing intelligence and allowing the user to make decisions. The platforms today have reached a “Star Trek” level of sophistication and can now suggest possible decisions and prioritize them based on alignment to business objectives.
AI has innumerable applications in the form of vision intelligence. Vision intelligence can be used to check geometry conformance to minimize wastage. By extracting data about the dimensions of WIP goods, it can assess the conformance to prescribed quality standards. AI can also potentially identify and direct to the point in the manufacturing process where the deviations have occurred. In scenarios where the pipeline throughput is of highly valuable material, vision intelligence can be used to identify material removal or misplacement. This can help avoid unnecessary losses due to theft or mishandling of property.
Vision intelligence can also be used to ensure safety. AI engines can closely monitor for unwarranted or unnecessary human interventions in a biohazardous production environment.
Manufacturing Process Optimization
AI applications can run simulations of current and future alternatives for manufacturing processes. These simulations help identify the most viable and optimal manufacturing process. What Oden calls “The Golden Run.”
For instance, an AI system analyzing motor fed conveyors can suggest the replacement of motor fed conveyors with gravity fed conveyors. The replacement will help not only eliminate the expensive motors and spares, but also minimize the cost of energy consumption involved.
Manufacturing Assistance denotes the close collaboration between AI systems and factory floor personnel in the manufacturing environment. The AI system can assist the operator in competently executing their roles and responsibilities.
A simple example of this arrangement could be robotic welding arms guided by personnel to identify the spot of welding. The robot then decides the right amount of weld fuse and arc to be used. This can help not only optimize energy consumption but also drive better efficiency in the production process.
Get One Step Closer To Production Optimization Today
Production Optimization in manufacturing is key to ensuring efficient, cost-effective, desirable outcomes that also assure sustained competitive advantage. Industrial IoT software, machine learning and AI can come together to deliver unseen benefits through optimization. With the right platform that connects all the three, your manufacturing line can become very profitable.