Imagine we are in a paint factory. In the factory there is an automated production line, and next to that we have two warehouses. The first warehouse contains ingredients for paint: thinner, binder, colourant, and additives. In the second warehouse we store cans of paint produced on the production line.
The ingredients come in from two other factories, owned by the same company that runs the paint factory.
Paint is made to order, as well as in bulk production: our factory provides paint to specialist painting companies as well as wholesalers.
The bulk production runs 6 days per week, for 6 different kinds of paint and is interrupted whenever a special order for a specialist company needs to be produced. What the management starts noticing is that a number of issues occur over time. First of all, it happens regularly that special orders have to be delayed, because ingredients are not in stock. Second, production levels are not in line with the calculated capacity of the production line. Also, there appears to be a shortage of some of the additives at certain times.
The management of the factory wants to do something about this, and decides to start collecting data from various sources to analyse what is going on. As we saw last month, data sources can be ERP and MES systems, but also sensors, as well as PLCs and other control systems. When we make a decision to collect data from there, it takes time to get everything in place, but let’s assume that in this factory we can do it instantaneously.
Before doing so, however, we need to establish what information we want to have available in the end. In this case, that leads us to expect to need the following.
What do we need?
In order to understand why material is not available for special orders, we want to analyse the size and frequency of these orders against the production and logistics process for getting the materials to our factory. This means we have to collect order sizes and frequency, and compare that to production and logistics planning (and actual execution) for the ingredients.
In relation to that, we also want to investigate why the additives are running short. For this, we need to check the logistics and production plans for additives against the production plans and actual usage of the additives in production of our paints.
Finally, we want to analyse why the production line does not meet it’s capacity. For this, we mainly want to look at running and idle times of the different machines, and the causes for the various idle times.
Where to get the data?
To get the right data for these analyses, we need to tap into multiple sources.
For material availability, we need to get order information from ERP (sizes and frequency), from MES (bill of materials for each order) and the MES and logistics systems for the ingredient factory. We need this in order to get clear how much material is needed, when its needed and when it is produced and delivered to the main factory. This included planning as well as execution data: we need to know when ingredients are planned for delivery, as well as when they actually arrive.
Similarly, for analysing the shortage of additives, we need data from MES about which amounts were needed, available and missing during order execution. Since these additives are partly handled by hand, we also need data about actual usage, spillage and measurement data from the equipment used to measure the doses of additive. This comes from the PLCs controlling the equipment and possibly additional sensors installed there.
For the production capacity analysis: we need from MES the expected capacity of each production machine, while feedback data from PLCs and sensors can tell us exactly (from when to) when the machines were working, idle, blocked etc.
Collecting and using the data
Collecting the data from all these sources is handled by a data gateway. Typically, this is an IIoT gateway, that allows all systems and devices to connect to it using standard internet protocols (TCP/IP), on top of which we can run messaging protocols like OPC-UA and MQTT. From the gateway, the data goes to a data store, either on premises or in the cloud, and we run our analysis software and dashboards on top of that.
Analysis can be done in a cockpit like dashboard (on a tablet or a normal screen), but also in Excel sheets if needed, or by means of feeding the data to Machine Learning models. The results can be used to improve the way of working in the factory.
And the reasons were…
In our imaginary paint factory, the analysis led to some interesting conclusions.
First of all, it turned out that the ingredients factory was given a production and logistics schedule based on bulk orders only. As a result, some ingredients, mainly additives were not produced in the quantities needed for the special orders. This was solved by modifying their production strategy to include an estimation of what is needed for special orders. The initial estimate is based on data from the past year, in a next improvement step we will add the possibility to also take into account the actual current orders. That was not done immediately, because it required a special coupling between the ERP system and the MES system for the ingredient factory.
The shortage of additives was only partly explained by this. It turned out also that during manual handling of certain powders a significant amount of material got lost. This was eventually traced back to the fact that some of the containers used in production were not completely dry after their bi-weekly cleaning cycle. This led to the material clogging to the container and getting lost. This was detected in two steps: the data analysis showed that a lot of material was reported as ‘lost’, physical inspection at the work station showed why.
The lack of production capacity had two causes. One was the absence of material for the special orders. The operators would reconfigure the system for a special order before checking if the material was available. That led to a double set up for machines in case the material wasn’t there: change to special order, then switch back to the previous product.
The other reason was the bulk production planning, which did not take into account set up changes in relation to bill of materials. For some paints, the colour mixer needs to be cleaned before switching to a different colourant. The schedule included extra cleaning cycles (a part of set up) that could be avoided by changing the order in which colourants were used, eliminating 20% of set up time.
Does it have to be digital?
All of these are thing that can be investigated and found without digitalisation but are likely not to be found soon that way. The reason is that collecting the required data is not always easy by hand, and the other is mindset. If data and optimisation are not first class citizens in a factory, people will mainly focus on ‘getting the work done’.
That way, the continuous improvement cycle shown in the figure that comes with this article will not come to life.
Source: Bits&Chips, https://bits-chips.nl/artikel/painting-by-numbers/