Real-Time Production Monitoring: the 2023/24 implementation playbook

In the last year, we have seen a strong trend towards real-time monitoring and predictive maintenance. In this article we take a brief look at the current and future best practices, and discuss opportunities for increasing your process engineering ROI.

Production monitoring systems in real time with Artificial Intelligence

Production monitoring has always been one of the key requirements to ensure KPIs are met in manufacturing companies. When we talk about KPIs in manufacturing, we basically refer to metrics related to 3 aspects:

  • Productivity: it can be measured in different ways. Often, production efficiency is referred to as OEE, which stands for Overall Effectiveness Efficiency, which is a measure of the level of utilization of a productive asset. Also, more simply, productivity can be measured in terms of delivered good per unit of time;

  • Delivery time: the time required to produce a good, from warehouse withdrawal to shipping. In this case we can take under consideration metrics like the lead time, that is the overall time required to produce a good. Also, it is possible to subdivide lead time into cycle times (time required for a single operation) and set-up time required to retool a machine or work area. Another metric often used by production engineers is the takt time, that is the cadence of the production line required to guarantee the satisfaction of the deliveries;

  • Product quality: very often statistical metrics are used such as process capability, in terms of Cp and Cpk, DPU (defects per unit) or number (or cost) of scraps parts.

All of these metrics substantially impact the concept of value to the customer, i.e., what the customer expects from the supplier based on the price paid.

It is clear how these three aspects impact on the value of the goods produced, in a synergic way.

In this post we will describe the importance of implementing a production monitoring system capable of constantly verifying compliance with manufacturing KPIs. This system should also be able to identify those activities considered value-added and report those activities considered non-value-added, i.e., that do not bring a benefit to the end customer. Next, we'll discuss the importance of implementing such a system in real time: in fact, timing often plays a key role in KPI management, and even just a few minutes' delay in corrective action could result in additional costs for the manufacturer. Finally, we will discuss in more detail how some new technologies, and in particular solutions based on Artificial Intelligence, can be exploited for the monitoring of production processes.

Why we need a monitoring system for production

To deliver more value to the customer, it is necessary to ensure that production takes place in a repetitive manner based on company standards.

To ensure greater productivity in manufacturing cells, it is important that tasks follow Standard Operating Procedures (SOPs). Standard Operating Procedures ensure that activities are carried out repetitively based on the best designed methodology, specifically a sequence of operations and activities using the most appropriate tools to accomplish the required operations in the least amount of time and with the best possible quality. In order to ensure that operations are carried out according to standard operating procedures, therefore, compliance must be monitored.

Another example to understand the importance of implementing a production monitoring system refers to compliance with quality requirements on a statistical basis. In this sense, the most used methodology in manufacturing (but not only!) is Six Sigma. Six Sigma was developed in the 1980s in the United States, at Motorola, based on the teachings collected over the decades by the so-called fathers of quality, from W. Edwards Deming to Joseph Juran. Quite simply, Six Sigma is based on the use of data collection to maximize the potential of inferential statistics, thanks to which it is possible to monitor and predict the behavior of a population based on data from a statistically significant sample. Without going into too much detail, this continuous improvement methodology is essentially divided into 5 distinct phases, referred to by the acronym DMAIC:

  • DEFINE: implies the initial phase of the project, the collection of requirements and objectives to be achieved

  • MEASURE: implies the collection of data to analyze the process from an exquisitely quantitative point of view and thus have a starting point from which to carry out all subsequent analyses

  • ANALYZE: involves the analysis of data collected in the previous phase, in order to identify the main causes of the problem and possible solutions

  • IMPROVE: implies the verification on a statistical basis of the goodness of the identified solutions, in order to guarantee from a quantitative point of view that it is not a fortuitous event in the short term, but that the solution will bring the expected benefits also in the medium-long term

  • CONTROL: the last phase implies the control of the improvement actions introduced through the use of control charts in order to ensure process stability and control.

It is important to note that the implementation of a production monitoring system, the last of the 5 DMAIC steps, represents a fundamental phase to ensure compliance with the quality requirements of the final product and the associated production process.

Value Added vs Non-Value-Added activities

A production monitoring system is essentially a data collection system. Once analyzed, data can provide engineers with essential information for optimizing the production system. One of these essential pieces of information is the identification of value-added and non-value-added activities. Let's take a closer look at what this means.

Value-added activities are all those actions that add value to the product. For example, this category includes all activities that transform the product throughout the production cycle, such as machining, additive manufacturing, assembly operations or heat and surface treatments. Clearly, the ideal would be to have only such operations within your processes and to focus only on improving them. However, in reality, priority should be given to eliminating non-value-added activities. In this case, it is necessary to distinguish two subcategories:

  • non-value-added but necessary activities: that is, all those actions that do not add value to the product but are nevertheless deemed necessary;

  • non-value-added and unnecessary activities: this category includes all other operations.

As regards the first sub-category, we can consider, for example, many process controls and inspections that are required by the client or by international or industry standards: even if they do not add value directly to the product, they are nevertheless necessary to maintain. In this case, therefore, they cannot be eliminated, but a new way must be devised to ensure that they minimize waste or can be incorporated with value-added activities.

As for the second sub-category, the only action possible is total elimination as they are wasteful. This category includes all activities considered waste, or Muda, according to the Japanese terminology used in Lean Manufacturing. For completeness of discussion, it’s possible to identify a total of 8 categories of waste:

  1. Transportation: these are the recurring costs related to the excessive transport of material / semi-finished products with related associated equipment (non-recurring costs), such as lifting equipment, trolleys, cranes, etc.

  2. Inventory: costs associated with excessive storage of material, from space to raw material costs

  3. Motion: similar to transport, it represents the costs associated with the unsolicited handling of the material, for example during an assembly operation due to a non-optimal workplace organization

  4. Waiting: these are the costs associated with waiting times for example of a semi-finished product waiting to be processed

  5. Overprocessing: are the costs associated with unsolicited processing, i.e., that does not add value to the product or that can be eliminated through optimization activities

  6. Overproduction: are all those costs associated with a PUSH rather than PULL production system, with greater WIP, therefore material in work, space, storage, which in reality is not required by the customer

  7. Defects: non-quality costs, in particular related to management and rework

  8. Skills (lack of): the eighth waste refers to the costs associated with the lack of skills to perform a job correctly

Real-Time Monitoring Systems

One of the main functions of production monitoring systems is to identify the behavior of the manufacturing system and thus identify which are the value-added and non-value-added activities we described earlier. As anticipated, such a system involves a non-negligible expenditure of resources. For this reason, such production monitoring activities are carried out by engineers and production workers on an occasional basis. For example, one of the tools used for flow monitoring is the Spaghetti Chart, in which the path of production batches is mapped and traced to check for any inefficiencies within the production flow. Or, the SMED methodology is used to monitor the setup activities of industrial assets in order to eliminate waste and optimize value-added activities. In this way, standard operating procedures can be produced that must be followed step-by-step by operators to ensure process KPIs are met.

However, how to be sure that processes always perform repetitively? For this reason, it becomes necessary to implement real-time monitoring systems that allow you to keep the production process under control at all times. In this way, it becomes possible to react to potential anomalies in a timely manner, in order to prevent the defect from propagating downstream of the processes, exponentially increasing the probability of greater waste.

Today, there are several technological solutions that can help processors in their production monitoring activities.

For example, RTLS (Real Time Locating Systems) are able to monitor in real time the position of moving assets, typically production batches. In this way, it becomes possible to know in a precise way the flow that each batch follows inside the factory or the supplier, to know exactly the cycle time or if the parts perform operations out of cycle. This technology also makes it possible to monitor the position of factory personnel in order to improve, for example, the layout of the workstation so as to minimize position changes within a cell.

Another technology that enables real-time monitoring systems is the use of Computer Vision, i.e., cameras able to process the acquired images through Artificial Intelligence algorithms.Flowbase analyses and annotates video automatically, producing VA/NVA, Production Time Breakdown and Overall Production Efficiency reports, plus custom metrics and alerts.Never spend another minute tediously watching and logging items.

How to use Artificial Intelligence for process monitoring

When people talk about Artificial Intelligence (AI), they tend to refer generically to the development of complex algorithms capable of performing tasks that approach human abilities. In fact, Artificial Intelligence is the ability of a machine to exhibit human abilities such as reasoning, learning, planning, and creativity.

Artificial Intelligence allows systems to understand their environment, relate to what they perceive and solve problems, and act toward a specific goal. The computer receives data (either already prepared or collected through sensors, such as a camera), processes it, and responds.

One of the main applications of Artificial Intelligence is what is called Computer Vision. Basically, as the human brain is able to process information and make decisions based on information received from the eyes, i.e., the vision sensor of the human body, in the same way Artificial Intelligence is able to process information received from a camera and provide inputs accordingly based on it.

Therefore, through Artificial Intelligence and the use of cameras, it’s possible to monitor industrial processes that are within the field of view of one or more cameras. Going even further into the merits of the matter, a system of this kind will be able to monitor in real time, with extreme accuracy and flexibility a series of activities within production processes that only the presence of trained personnel would be able to do, with all the associated costs.

Some examples in this sense are:

  • Monitoring personnel tasks during setup activities: it is possible to analyze videos by identifying which non-value-added operations are being performed. Then, through Artificial Intelligence algorithms, it is possible to signal whenever an operation is performed out of standard. This application can also be used in training, to help workers to learn how to perform standard operating procedures and thus greatly reduce the learning curve;

  • Monitoring production bottlenecks: when batches tend to accumulate in a certain area in an unplanned manner, it means that a bottleneck is being created and immediate action must be taken. However, it can often take hours or days before someone reports the problem. Artificial Intelligence can assess when the buildup reaches a certain threshold;

  • Monitoring of the production flow: when a batch performs out-of-flow processing or remains in one location for too long, the anomaly can be reported in a timely manner;

  • Monitoring of manual tasks: within a production cell, there are many opportunities for waste, especially in terms of excess or unnecessary movements. Computer Vision can monitor and detect such movements and report them to the appropriate personnel.

These are just a few of the applications that Computer Vision Systems are capable of performing in real time.Another advantage of these systems is cost, or capex. In fact, from the hardware point of view it is required simply to install cameras costing a few hundred dollars, while the main expense concerns the development of the monitoring algorithm. However, even in this case the cost is accessible to small to medium sized businesses, also by virtue of the fact that it is a one-off cost. As a result, at the state of the art, the examples on the field show extremely interesting ROIs and payback periods, depending on applications and margins for improvement.

No cookies, no trackers. Built with care in 🇪🇪
+372 5559 7089