This is the glue pot on the IMA edgebanding machine.
A customer of mine asked me to design a system that will control simultaneously three different heating resistors from 400 to 1000 Watts each, and will get temperature feedback from 3 different J-type thermocouples. The system will be used for a maintenance table for glue pots from IMA edge-banding machines. These machines have a unique feature that allows the operator to completely remove the glue pot from the machine for maintenance or servicing reasons.
The service table must have a temperature controller for the 3 (or 2) heating resistor sets of the glue pots. The glue must be kept in operating temperature (about 180oC) or in stand-by temperature (about 140oC), otherwise the glue in the pot will be solidified. Glue pots have 2 or 3 sections: The bottom section which has also the glue roller, the middle section which is called pre-melting, and an optional top section which is a pre-melting expansion. Each section has its own heating resistor set (800 to 2800 watts in total), as well as a J-Type thermocouple to read the temperature.
New Features and Technologies
The controller that i design will have the following features:
It must have very good power filters because it will operate in a very noisy industrial environment
It will have 3 outputs, each one capable to drive at least 15 amperes of resistive load
It will have 3 inputs for the three J-type thermocouples
The controller design must be solid and compact, able to stand some amount of vibration
A 20 by 4 LCD and 3 buttons will be used for user interface
A switch will be used to select between operating and stand-by temperature
The operator must be able to set the P gain,I gain and I time for all three channels from the user interface
Most important though, i plan to go into new technologies and techniques: This will be my first project that i will use an SMD PIC. Things are very nasty for me because i will use the 16LF1939 QFN/ML package, which actually has no pins on it's sides. Instead, all pins are underneath the chip. I watched several videos and read many articles explaining how to solder QFN packages, yet i still do not feel very comfortable. But i am absolutely thrilled to see the results of my first QFN/ML soldering job. I will use my brand new Aoyue 968 hot-air station for the job.
The QFN package is ridiculously small! Its tiny!
All 44 pins are underneath the chip, therefore a soldering iron is useless.
Additionally, the chip has maximum 3.3 volts supply, therefore i have to find ways to interface it with the LCD which uses 5V. Finally, this will be my first stackable board layout. What i mean is that the PCB that i will design will have a 16-pin female connector on top right and 4 holes on it's corners. This layout will be identical to the layout of a 20 by 4 LCD display PCB. I will use 10mm metallic spacers to fix and connect the LCD board right above my PCB - one board will be stacked above the other. Hopefully, my PCB will have the same dimensions like the LCD board.
K type thermocouples are very popular for one main reason: The temperature range. A K-type thermocouple can measure up to 1250 oC. But in lower temperatures, a K-type thermocouple is not very accurate. For this reason, the machine maker (IMA) decided to use J-type thermocouples. This type can measure temperatures up to 750oC and its wires are made from Iron and Constantan. I will use the Maxim MAX31855J Cold-Junction Compensated Thermocouple-to-Digital Converter chip. This chip is the replacement of the MAX6675 that i used in my DIY Soldering Station project.They have the same pinout and they use the same protocol to interface with the microcontroller. Additionally, the MAX31855 series (suffix "J" stands for J type) provides more information to the microcontroller like the cold junction temperature, and it can also detect basic problems such as thermocouple connection break or short-circuit to earth/supply.
25 Amperes are enough for output!
Each glue-pot has 3 sets of heater/thermocouple pairs, each pair with different setup:
Pair 1: Glue roller. It has 2 x 1000 Watts resistors and 2 x 400 Watts resistors, a total of 2800 Watts - about 12.8 amperes
Pair 2: Pre-melting unit. It has 2 x 400 Watts resistors, a total of 800 Watts - about 3.6 amperes
Pair 3: Pre-melting expansion (optional). It has 2 x 400 Watts resistors, a total of 800 Watts - about 3.6 amperes
12.8 amperes are too much to drive them through the PCB directly. There are of course methods to increase the maximum current through the PCB wires, but i prefer to use external load relays. I plan to use power bursts to control the output power, therefore the relays will be armed and disarmed many times. For this reason I chose to use Solid State Relays (SSR) to drive the loads effectively. I found a very interesting offer at ebay...
This project is probably a stretch. Heater control is a big deal in industry and there are already many complex solutions to heating control. Each situation is different and the designer must know the parameters and tolerances he is working with. It is not possible for the equipment operator to be expected to adjust the PID parameters on the fly because that is the engineer's job and a very complicated one at that. Plus, the indiscriminate adjustment of various parameters on the fly can easily cause system instability that is difficult to understand and correct unless all the system details are known. You will probably NOT want to use 3 discreet resistors for the heating. Typcially a pot like this is heated using banded silicone surface heaters wrapped around the pot. You must define the allowable temperature variation under operating conditions and how quickly the temperature must stabilize. From that data you can specify how much power the controller must be designed for. The smaller the allowable temperature variation, the higher the required power. Heating control is a bit simpler than, say, a robotic arm or other mechanical device and I doubt you will need derivative factors to do it. Just the capacity to change the temperature of the pot - under various states of fill - should do it. Remember that as soon as liquid is poured from the pot, the applied power will need to be reduced immediately to avoid overheating the remaining fluid. I suggest checking into alternative sensing devices such as IR optical/temperature sensing to avoid breakdown of the tiny thermocouples. You will also get more accurate readings on temperature than a t.c. placed on the outside of the pot. These solutions are difficult and expensive. Have a look at the control circuitry that exists at the industrial level and you will get an idea of how complex it can be. Programming the PID parameters is an iterative process and cannot be done from a set of tables. Good project - not many out there that can do this work. Good Luck.
@Bartek thyristors would certainly work. But as i explained, i do not want to put high currents on the board for 2 main reasons: First this will increase the size of the board. I want the board to have the same size as an LCD 20 by 4 board. And second, to drive 12 amperes a thyristor will need a large heatsink. Solid state relays are not that expensive after all, and there are nice large heatsinks for SSRs suitable for electric cabinet installation.
But most important is that this controller will be used in an industrial application (in a big factory that makes wooden furniture), which means that the most vulnerable parts must be quickly and easily accessible and replaceable.
@Bartek I have already make other PID controllers. Therefore, for me, it is not a big deal to triple the program and make a triple PID. I may not use the D term at all, the reaction time of the system is slow and kinda stable.
But After all, it is more fun to make a PID controller -even if it is actually PI. And as i always say, something that is not fun to do it, its not worth doing it.
Hello, control engineering student here.
Although PID controller is very fancy, i believe in this sort of application it\'s an overkill. As long as you don\'t have to control temperature very precisely and the reference trajectory is not going to change rapidly too often, I\'d recommend using a simple two-state regulator with hysteresis. With PID, you have to take care of noises(I believe you know what happens when you try to calculate the derivative of noisy signal), also, you should implement anti-windup mechanism. Lot of coding and lot of possibility of problems. Two-state regulator requires just a few lines of code.
And instead of using a solid state relays, maybe a thyristors would be a cheaper and more reliable solution?
@George yes indeed, but Microchip offers this package only in 1000 reel :/
@Cheerio I felt exactly the same when i received the package. With the hot-air it is very easy to solder it. And in comparison with the PDIP package, this one needs at least half of the time to solder it.