‘My Foam Plant looks at thousands of parameters at the same time and makes the correlations between them,’ Vinas said. The Human brain can understand the correlation between two, three or four parameters, but not between 200 parameters. Our algorithm can calculate which parameters have the greatest impact and in which direction.’
This can feed through directly into the bottom line. ‘The raw material cost is a large driver for our customers and that’s related to efficiency. My Foam Plant will help foamers take control of that process,’ he said.
For each slice of block, you can understand the quantities of raw material used, the physical parameters of the foaming machine, and the external conditions during the process such as the humidity or the pressure, the evolution of the temperature and of the shape during the curing and finally the final shape that will be delivered to the converting department,’ he said.
‘When you connect the process into the product is when you can start the optimisation and the correlation between the process parameters and the final output.
IPF uses two of the most cutting edge industrial ideas: Big data and machine learning to take people and their prejudices out of the equation.
To do this with machine learning, the process learns how to change the inputs of the foaming process to make better quality foam in more consistently shaped blocks.
‘Nowadays, when companies want to understand the foaming process they need experts,’ Vinas explained.
‘Because they know the chemistry and the physics the experts can understand the process. My Foam Plant connects process with results using different technologies, and it is not so important now to understand physical chemistry or the relationship between chemicals and process parameters, because the system finds the correlations. It’s similar to the software used by companies such as Google and Facebook.’
‘The IPF model, collects and stores a lot of inputs, makes correlations, and then the system detects which parameters have the most effects on the final results. We can start asking My Foam Plant questions.
‘In our latest version, we are focusing on the dimensions of the block, because it relates to quality and efficiency of the cutting process. We try to statistically understand what the impacts of input and processing parameters are on the final dimensions of the block and on the final shape of the block. We have already seen interesting facts, such as…
- Output differs between a 5-15% compared with planned block shape;
- Block height is not constant, having important variations that only when are tracked you can get the best from the cutting;
- Block stabilisation is much longer than we normally assume, and depends on the type of foam and other process conditions; and,
- Automatic efficiency measurement differs with manual methods, its more accurate and especially important, all blocks are tracked.
By using a statistical approach, and by taking many measurements per second as the block moves along the conveyor, it becomes possible to accurately link the properties of a portion of the block at as it moves along the conveyor away from the mix head with what was happening at the mix head when it was poured. This way the effects of flow rate, proportions of chemicals and so on can be connected to the cell structure and shape of a slice of block.
In addition to foaming process info, the blocks themselves are also tracked with information such as date of manufacturing, its readiness for cutting, how efficient the foaming process was, or type of foam, he said.
For example, normally professionals talk about shrinkage in very general terms, Vinas said, adding his system measures every block, and the results are quite surprising. Even in the same run, results might be quite different, he said.
To human vision a good block looks flat, but the laser scanning process shows that there are ripples and the height of the block moves in waves along its length.
‘We can measure the impact of the flows, the impact on each wave, which parameters cause or are related to the gain or to the cold flow of the block before and after the curing, when the reaction is complete,’ Vinas said.
As well as ripples, in extreme situations ‘in blocks which are designed to be square sometimes the lower part becomes wider and the top part becomes narrower,’ Vinas has found.
Vinas said that the interest in his product extends to raw material makers as there is now an understanding of the impact of changing raw materials during foaming and after storage. ‘We can also look at the effects of changing raw materials in a formulation,’ he added.
‘When customers talk to materials companies about shrinkage or defects, they often show simple samples or a formulation; with our process, materials companies can really track this information and at the same time the parameters of the foaming machine at the moment of the incident. The raw materials supplier can then understand what happen when processing his raw materials,’ he said.
Open source collaboration
IPF, located in the Basque region of north Spain, collaborated with several other companies to develop open-source software that can connect to a wide range of other machinery operating systems. ’We have a team of people with different backgrounds working on My Foam Plant inside IPF,’ he said.
‘We don’t talk about versions, we talk about the range of solutions available. They go from a very simple version where you just connect machines. The simplest set up (WGO: what´s going on) – shows which machines are connected, how many parts are being made and it shows you in real time what is going on in the factory. This most simple platform shares the same architecture as the most sophisticated version we produce,’ he said.
The WGO version use graphs and numbers to illustrate what’s happening at the factory. Plant managers can see online or study later when a machine started, how many metres have been produced, how many kilos have been foamed without going into great detail.
Vinas explained that intermediate version, currently unnamed, looks at individual machines. ‘So instead of looking at the whole process, we focus on certain machines. If you have a machine which controls the output of your converting line, we compare the shifts, we compare the efficiency of the machine, we can detect the defect of the machine at early stage to plan a recalibration or a maintenance. We can really analyse a great deal of information from machine in detail,’ he said.
‘We have been developing this platform for the past 3-4 years. At the beginning, our goal was to connect machines to report the performance of the process from the raw material to the trimming. It is interesting, but this wasn’t giving very much added value to the customer and we started to focus on what was going to give real benefits’.
The key version of the platform, known internally as ‘the Brain’ but lacking any other designations, ‘collects information from production, ‘we connect process information into the product cm by cm,’ he said. Now you can focus on an individual block and overview the whole process, the DNA of the block,’ he said.