Temperature control of industrial ironer proof of concept

Part of the series University projects (0 posts total).

My 5th semester project was in collaboration with JENSEN, a company specializing for washing, drying, ironing and folding textiles on an industrial scale. These machines are typically used in hospitals, hotels or aboard cruises, where a large volume of clothes must be cleaned at fast rates. JENSEN had made a project proposal: How can their industrial ironing machine automatically guarantee a similar finish on all clothes, despite uncertainties such as type and wetness of clothing.

The machine

Alongside the proposal, we received a proof of concept machine which was to act similarly to the ironing machine. The ironing machine works by having a heating system that constantly heats liquid which enters a pipe known as the chest. Clothing enters from the feeder (on the right) and is rolled in between the chest and a smooth fabric roll akin to an ironing board.

The received prototype worked a bit differently. Instead of clothes, the heat exchange was to be performed between the liquid in the chest, and cold water in another pipe. In both instances, the controlled variable is the input temperature of the chest and the flow thereinto.

The three flat panels on the right represent three chests, and the open spouts are to be connected with a steady stream of cold water representing clothes. Inside each chest, there are three layers connected by the pipes on the outside. This cold water exchanges heat with the hot water from the rest of the system through a thin sheet of metal inside the panels. This hot water is stored in the tank (the big white box), pumped around with the pump (large gray cylinder) and heating with an electrical heater inside the tank (not pictured).

We were the first ever to use this prototype and it had been built specifically for this project. It took a long time to properly figure out the controls and workings of this machine in tandem with our point of contact at the company, since it included adding new sensors, configuring sensors, connecting pipes and programming the accompanying programmable logic controller (PLC). The PLC was a B&R PLC and was programmed using structured text (ST) and C. In some cases, we also used ladder diagrams as not all functionality of the PLC was supported by ST. We programmed a human-machine interface (HMI) to interact with the system without reprogramming the PLC.

The model

A very extensive model of the system was created to properly capture its dynamics and ensure adequate control. In particular, we wanted to control the temperature of the hot water supplied to the chest to be constant, regardless of the flow or the flow and temperature of the clothes (cold water). The system is modelled as two separate masses of water located in the water tank and in all the chests combined with constant flow in between. The heat exchanger is modelled as a multipass crossflow heat exchanger.

The energy exchanged in the chests is related to the change in temperature which is often calculated by a logarithmic relation, which is highly non-linear making it difficult for control design. A linear relationship is chosen as an approximate for this heat exchange. All the parameters of the model are found using isolated direct experimentation by examining the cooling of various elements within the system. The entire model is constructed in the block diagram tool Simulink and simulated and compared to measured responses of the system. The most important fit is the top left, i.e. the temperature of the hot water supplied to the chest.

The controller

Two controllers are designed. The first is a simple on-off controller with hysterisis. It is theorised that such a controller would work well for this system as it has little inertia as the supply water temperature stops increasing very shortly after the heating element inside the tank is turned off.

The second controller is a combination of a heating controller, which attains a working region of temperature, and a local controller which attains the reference smoothly and efficiently. The heating controller is on-off with hysterisis as before. The local controller is realized as a single pole system and a gain (essentially a PI-controller) with a lead component to ensure stability in the feedback loop. This is done in accordance to the found model parameters and the specifications from JENSEN.

The controller is digitalized. The Nyquist-Shannon theorem states that aliasing will occur if a continuous signal is sampled with a frequency less than twice as the bandwidth of the signal. The bandwidth is given from the specification of the temperature sensor but it is too large and therefore causes aliasing, which is fixed with an anti-aliasing filter. Both filter and controller are discretized using the bilinear method which causes nonlinearities in high frequencies, which is luckily small enough to ignore.

Difference equations are posed for the controller and filter and they are constructed in simulation in direct form II. Due to the bilinear transform, the system is now in the Z-domain and its poles are quite close to 1 indicating that the system is very close to unstable. If the direct form is not calculated with enough accuracy, the system will indeed become unstable. This is guaranteed by examining the poles for differing amounts of decimal points and it is concluded that 5 decimal points is needed to ensure stability.

We did another test on the prototype to see the responses of these controllers.

As expected, the on-off controller has a difficult time approaching a steady state, while the other controller has a large overshoot. Less overshoot requires either a smaller gain which results in a slower response or a higher order controller which may give stability issues, in particular wrt. the numerical issues presented earlier. Still, it seems that the on-off controller is perfectly adequate to control this system as originally hypothesized due to the nature of the system.

What did I learn?

Sometimes you must admit that your engineering solution is subpar to the pre-existing one. As long as your learn why, it is still a learning experience, and I learned a lot:

• Collaboration with a company in project-oriented work
• How to model heating systems and in particular heat exchangers
• PLC programming using structured text and C (and some extent ladder diagrams)
• Building an HMI using said PLC
• Constructing a small-signal controller
• Discretization of continuous systems by bilinear methods (other methods were also considered)
• Implementing discrete controllers/filters by difference equations and direct forms
• Stability analysis of discrete systems in the z-domain

We also wrote a whole manual for the machine for the groups to come. Sadly, no one worked on the project for the next two semesters, and the prototype was sent back to JENSEN to make space for other machinery.