Some machine processes, such as presses, can require extreme accuracy in applying and holding force on an object. A popular way to measure force is via load cells. But what do you do when the accuracy required by a particular application is higher than that guaranteed by the load cell manufacturer?

Figure 1. A schematic view looking down on the pressure plate. The four motion controllers each control three cylinders (1-4), evenly spaced along the plate, via a single proportional valve.

One can use multiple load cells and average the results. That way, small differences in load readings between the cells, which can occur at different times and under different environmental conditions, are averaged out. A side benefit of the averaging process is that noise signals on individual load cell outputs would also be reduced by the averaging process. Another benefit of using multiple sensors to measure a single process parameter is that one sensor can fail and the system can still function, albeit at a possibly lower level of accuracy.

But how does one manage gathering inputs from multiple redundant sensors and processing them to achieve the above benefits? The motion controller that is used must be programmed to do checking and averaging of load cell data before that feedback is provided to the control loop. Parker Hannifin (Marysville, OH) did this using feedback programming of the RMC150 motion controller from Delta Computer Systems.

Load Cell Precision

Parker’s customer needed to maintain a large amount of downward force on a 50,000-pound metal plate to a very tight tolerance of ±25 pounds against a total force of over 150,000 pounds. Also, because of the critical nature of the process, it was important that the system be protected should any force sensor fail to provide correct readings.

To ensure adequate redundancy, one load cell was attached to each of three hydraulic cylinders that were controlled by a single motion controller operating a single servo valve. Further, to make sure that adequate force was being applied, there were four groups of three cylinders arranged along the plate (Figure 1), operated by four total motion controllers with servo valves.

Figure 2. The RMC150 can control up to eight motion axes, and can process inputs from up to 16 different transducers.

Each load cell output was compared to the target force, and the status bits of each load cell were monitored. If the motion controller detected a problem with one of the load cells, that load cell was instantly eliminated from the calculation, and the outputs of the other two were averaged. If the output from a load cell drifts too far away from a target value, the motion controller starts a timer. If this difference maintains long enough to keep the timer going, that load cell is eliminated from the average. This allows the system to filter out step changes in target force to avoid false positives (i.e., false indications of load cell failure).

The motion controllers handle the pre-processing of data from the load cells using custom feedback, which allows the motion control loop to use feedback from a source other than directly from a hardware input. Instead, axes defined with custom feedback obtain feedback values to be used by the control loop from an internal CustomCounts register that can be written to by a user program that continuously calculates the feedback value and assigns it to this register. In the press application, the controller runs a software program that tests and averages the sensor data from each load cell in the group of three connected to it, and the resulting average is used as feedback to precisely control target force on the pressure plate.

By taking a simple average of the load cell readings, it is possible to reduce the noise of the system. The average force value becomes a solid line on a graph of force versus time. The controllers are supported by a software tool set called RMCTools, which includes software that can be used to verify this (Figure 3).

Besides looking at the load cell force outputs individually to decide whether the sensor values are believable, software running on the controllers also looks at status information coming from the transducers. If it looks, based on that info, like a connector is unplugged or a load cell has failed, the user program running on the motion controller automatically eliminates that cell from the next scan, and issues an alarm to the machine operator.

In addition to the load cells, the system is also wired with pressure sensors in the cylinders on either side of the piston. The difference in pressure values between these two provides another measurement of force that can be used for diagnostic purposes.

Controlling the Controllers

Figure 3. A plot showing actual force (black) versus target force (yellow). Red represents position (it hardly changes as force is increased to a higher level). Green is the servo drive signal to the hydraulic valve. The top line (orange) represents system pressure.

The motion controllers are connected to the PLC via Ethernet. When a motion controller is first turned on, a “first scan” program is run. Then the controller looks at status bits and if they are within parameters, the motion controller informs the PLC, which sends a command telling the motion controller to start controlling the force. Each PLC looks at what each of the motion controllers is doing. Should one motion controller produce an output that disagrees with the others, the PLCs turn it off and shift its work to the other motion controllers.

Besides averaging outputs from redundant sensors, custom feedback can also be used to switch between transducers of different sensitivities. This would be useful in applications such as testers where controllers need to maintain precision while spanning a wider range of inputs than a single transducer model can handle.

In fact, feedback doesn’t need to come from physical transducers at all. It can be a variable. For example, a user program can assign the custom count register to “anything,” where “anything” can be a mathematical function.

As a result, machine designers who have special challenges that are not met by individual off-the-shelf transducers should consider employing custom feedback processing techniques in order to improve their machines’ operation.

This article was written by Peter Nachtwey, President of Delta Computer Systems, Battle Ground, WA. For more information, Click Here .