Getting started with smart industry: the inside of the factory, production line and machine
The spiral, which is actually a system of circles together, is a mechanism for making improvements in a manufacturing organization. These improvements can be in the areas of efficiency, cost, quality and even environment. Which improvements are most relevant of course depends on the KPIs that are labelled as most relevant. In many manufacturing companies, the main focus is on production capacity and consequently more efficient manufacturing, but it is also possible that sustainability is more important, looking at issues such as energy consumption and reducing waste.
In all cases, it is necessary to know what is happening on the shop floor, gather information about it, and make changes based on that information - which are then monitored.
"When a company is looking for efficiency improvements, questions come up regarding the amount of product that passes through a workstation, the time it takes to process that amount and the downtime of a workstation. The answers to those questions can be obtained in several ways."
Data, and more data from machines and workstations
The innermost level of our spiral is the machine, or rather the workstation. The latter term is more appropriate because, particularly in SMEs, there is a fair amount of manual labor involved in production, with a production worker combining manual actions with machine operations. Think of things like manually adding small amounts of ingredients in compound feed production, setting and adjusting machines for bending or sawing metal parts, or assembling products using a soldering robot.
When a company is looking for efficiency improvements, questions come up regarding the amount of product that passes through a workstation, the time it takes to process that amount and the downtime of a workstation. The answers to those questions can be obtained in several ways. After each production step, i.e. after each workstation, so-called WIP, Work-in-Progress, is created. By counting these and combining them with the hours spent by a workstation (based on running hours, or by combining WIP numbers with the employees' time clock), it is possible to derive relatively accurately how efficient a workstation is.
When it comes to sustainability and energy consumption, it is already more difficult. Especially with electrically driven equipment, an energy meter per machine is not available as standard. The same applies to the waste produced, where applicable - there is little record per machine there either.
For data collection, it can be useful to add additional sensors to a machine. The energy meter for power consumption is a simple example, but we can also think of a sensor that, for example, determines the actual number of productive hours of a machine on the basis of movement signals. There are relatively cheap sensors available today that can be placed near, on or in a machine without having to modify the operation (hardware and software) of the machine itself. Those sensors can be linked to a network and the data collected can be stored and used centrally. Even the time clock, or a counting pedal or scale used by production workers can be seen as a sensor, allowing the non-automated side of production to be measured and analyzed as well.
"There are relatively inexpensive sensors available today that can be placed at, on or in a machine without having to modify the operation (hardware and software) of the machine itself."
Reactive as proactive
The data collected by these sensors can be used both preventively and proactively. Reactive use applies, for example, to the issues of improving efficiency and sustainability. The sensors collect data, which is analyzed and then used to make adjustments in, for example, the configuration of a machine. Proactive use is found in areas such as preventive or predictive maintenance. By keeping track of how long a machine is running, how accurately a machine makes certain movements, and how often errors occur, maintenance can be planned before a machine actually stops.
Production lines are more relevant than individual machines
Optimizing the operation of individual machines in a plant is not always useful. In many cases, those machines are part of a production line, in which they perform successive processing steps. So in that case it is much more about the combination of what the different machines accomplish together. The sensors we mentioned above can be used to analyze where in the production line bottlenecks occur, and how they can be solved. In the first article in this series, we already referred to Goldratt's work from the 1980s, which focused on the problem of bottlenecks. As in those days, Goldratt emphasized the collection of data from production lines, because only on the basis of that can it be determined where the process is stagnating. At that time, around 1984, it was still far from common to have computers in a factory collecting all the data and making it available. With the current state of the art, and the sensorics mentioned above, we have two good opportunities to do these analyses better.
Dashboards and Digital twins
These two are dashboards and digital twins. A dashboard can be compared to a live reporting system: data collected from sensors is used to show on a screen which machines are running, which are idle, what and how much is being produced within a production line, and where there may be problems. This is similar to what happens on the control screens of MES and SCADA systems, but the emphasis in our case is more on the sensor data collected as part of improvement than on the operational data generated by the machines themselves.
A digital twin is a digital representation of a production line, with all machines. It can be used to simulate the operation of the production line, using the data collected in the real plant to feed the simulation. This makes it possible to replay certain scenarios, and test adjustments without making an actual change in the factory. Such a digital twin can be implemented at different levels of detail. The most basic form operates at the process level and uses global parameters, such as machine throughput to simulate production. The most advanced form, which is also used in engineering (we'll come back to that in a later article), simulates all the details, movements, and physics of individual machines to do a much more detailed analysis.
These dashboards and simulation are widely used to identify and implement all kinds of improvements and adjustments. The simulations are particularly interesting because they reveal, without delay in production, relationships between production steps that would otherwise go undetected.
"The possibilities with the datasets that arise from these combinations are almost endless."
The plant as a whole
The spiral as we have introduced it has five levels, with the factory itself being the middle level. In factories, although by no means everywhere, SCADA and MES software is used to plan and monitor the production process. These systems also generate the necessary data, often in the form of logging - in which events are recorded. Think of events such as 'order started', 'batch completed', 'ingredient added', with of course the corresponding details such as time, order number, etc. When this data is combined with the sensor data we described earlier, it becomes possible to simulate entire factories in a digital twin, but also to identify correlation between production planning and bottlenecks, or between production planning and machine failure. The possibilities with the data sets that arise from these combinations are almost endless.
The big challenge here for many manufacturers is twofold: on the one hand, smaller parties in particular do not always have a central MES or SCADA system where this data is collected; on the other hand, bringing the right data together is not always trivial. Fortunately, over the past 20 years there have been the necessary developments in the field of data warehousing, big data and (the newest name for this branch) data science. The software to do these analyses is available, which can open up a world of information for any manufacturer, large or small.
And the human being?
At the beginning of this article, we also mentioned production workers. They are affected in two ways by what we have mentioned here. First, the question of privacy arises when we start collecting data about the work they do. Nobody wants to be judged by the information we collect. In addition, they run the risk that their work will change or even disappear if we start introducing smart industry solutions. Therein lies a task for their employers and for the providers of smart industry solutions, the parties about which the third and final article in this series will focus.