Authored by:
John Ramming

President
George Sakona

Sr. Mechanical Engineer JR3 Inc.
Woodland, Calif.
Pat Byrne
Comco Inc.
Burbank, Calif.
Edited by Robert Repas
robert.repas@penton.com

Key points:
• The world of 6DoF load cells is characterized more by customization than by standards.
• Multiaxis load cells measure force and torque in multiple dimensions.
• Foil strain cells change their electrical resistance with the application of force, making them a good candidate for 6DoF load cells.
Resources:

Comco Inc

JR3 Inc.

Multi Axis Load Cell Design

Any load cell can sense along a single axis. Simple load cells that handle only one axis are called one degree of freedom (1DoF) load cells. Multiaxis load cells, those that sense more than one axis at a time, currently account for only one tenth of 1% (0.1%) of the overall load cell market, with six-degree-of-freedom force-torque (6DoF) designs forming the smallest subset. The multiaxis market also contains more customization than standards for sensor types and models. The core technology developed for multiaxis load cells took a truly evolutionary, rather than revolutionary, path.

While each 6Dof sensor manufacturer uses its own proprietary designs, all models simultaneously measure force and torque along three orthogonal axes: X, Y, and Z. Moving up to 12DoF sensors adds linear and angular acceleration to the force and torque measurements. All are collectively referred to as multiaxis load cells. Typically, these cells are so complicated that they almost always need their own support electronics.

Some applications in industry for multiaxis force/torque sensors include product testing, robotic assembly, grinding, and polishing. In medical research, they’re found in robotic surgery, haptics, rehabilitation, and neurology as well as in many other areas. Often they’re called upon to work in extreme environments such as space-exploration robotics and critical monitoring of deep-sea oil drilling.

The physical size of the sensor varies depending on factors such as force and torque ratings and mounting dimensions. Most come in a wide variety of load ratings and bolt patterns. Sensor orientation usually places the X and Y axes at the horizontal midplane of the sensor body and the Z axis along the sensor central axis. This places the reference point for all load data at the geometric center of the sensor.

Foil strain gages typically make up the sensing elements in heavy-duty, multiaxis load cells. They can sense the smallest deflection of any sensor technology providing a good magnitude of measurement below that of semiconductor strain gages. While optical and semiconductor sensor technologies are quite accurate, strain gages provide the most accurate and flexible force measurement for multiaxis load cells.

Semiconductor-type sensing elements have an advantage because they embody a higher unit resistance and a strain-multiplier effect. But their greater sensitivity to temperature variations and tendency to drift are liabilities in multiaxis load cells. In addition, semiconductor elements possess a nonlinear resistance-to-strain relationship, varying 10 to 20% from a straight-line response.

Optical sensing elements need a higher deflection to operate than standard foil gages by a magnitude or more. Higher deflections reduce the frequency band dramatically as well as allowing some mechanical deflection that some applications may not tolerate. Most need what is called a “stiff system.” A system that’s too flexible means possible oscillation and loss of precision.

Foil strain gages are not without their drawbacks. One primary problem concerns the cost of putting them on a sensing element — locating, bonding, and testing the gages to verify proper operation. As some 6DoF load cells contain 32 or more strain gages, mounting the gages along with the associated wiring and assembly can account for 50% or more of the labor cost in producing a multiaxis load cell.

Low signal strength was a challenge of early strain gages. But that is no longer an issue with today’s electronics. The issue of hysteresis error has also dropped by the wayside, running less than that of semiconductor strain gages.

In operation, a load applied to the working surface of the transducer changes the electrical resistance of the strain gages. The internal electronics monitor the change in resistance of each gage to produce an output voltage proportional to the force applied. Measurement of this voltage reflects the amount of force.

The design of a 6DoF load cell starts with the selection of a solid round from one of three possible materials: 2024 aluminum, 15-5PH stainless steel, or 6AL-4V titanium. The desired bolt pattern and load rating determine the diameter and thickness of the round. Most 6DoF sensors range in diameter from 2 to 20 in. Force ratings range from under 10 to greater than 25,000 lb with moment ratings from 2 to 150,000 ft-lb. Weight and machining costs give aluminum the edge, but higher loads need titanium or stainless steel.

Normally, a 4-to-7-in. sensor contains three or four load-carrying elements known as strain rings. These rings span the cell from top to bottom. But it is not uncommon to see custom sensors of up to 20-in. diameter with up to 16 strain rings developed specifically for a special application.

Typically, each strain ring contains four or eight bonded foil strain gages. The gages attach to an electronics board within the sensor that amplifies the signals and transmits them as either analog or digital signals.

The heart of a load cell
Contrary to what most think, the sensing element in a load cell is not the strain gage. The true sensing element makes up the primary structural element of the load cell. Typically, this is a precision machined block of material. The application of a compressive or tensive force to the sensing element produces a strain effect on the material, deforming its original shape. Within specific limits, the amount of deformation coincides with the amount of force applied.

Strain gages merely measure the amount of that deformation through a change in resistance. Through the bonding of the strain gage to different sensing elements, the same gage can measure a wide range of displacement, force, load, pressure, torque, or weight.

Each foil strain-gage material has a characteristic gage factor, resistance, temperature coefficient of gage factor, thermal coefficient of resistivity, and stability. The most widely used metals for strain gages are copper-nickel and nickel-chromium alloys. Foil elements come in unit resistances from 120 to 5,000 with gage lengths from 0.008 to 4 in., available commercially. The three primary considerations in gage selection are: operating temperature, the nature of the strain to be detected, and stability requirements. Other factors that determine the success of an application include the carrier material, grid alloy, adhesive, and protective coating.

Passive components such as resistors and temperaturedependent wires compensate and calibrate the bridge output signal. With a compressive force, the length of the strain gage shrinks and its resistance drops. Tensile forces lengthen the gage, forcing its resistance upward.

Surface preparation
Robust attachment of a strain gage demands having a sensing element with an absolutely clean surface. Surface cleaning removes any oils, grease, and other chemical or organic contaminants. A one-way cleaning method using a compatible solvent handles this chore.

Once cleaned, texturing the surface prepares the future location of the strain gages for bonding. The optimum surface finish for gage bonding depends somewhat upon the nature and purpose of the installation.

The strain gages are aligned to the sensing element as dictated by their design. Then they are mounted and bonded with either cyanoacrylate, an epoxy adhesive, or similar bonding material.

Surface abrading is one method of texturing the surface. It also removes any hard-to-clean contaminants such as rust, oxides, and so forth, while creating a surface that offers a “good hold” for the bonding material. Past texturing methods involved the use of both machine and hand tools such as tool grinders, disc sander, files, and handapplied 600 grit sandpaper.

Grit-blasting, or microabrasive blasting, replaced many of the earlier hand-preparation steps. The still-manual process uses work chambers specifically designed to capture spent abrasive while providing the operator a highly visible, clean work area.

Wet abrading provides another option. A mildly acidic solution “cleans” the surface during the abrading process and also acts as a gentle etchant. However, wet abrading is incompatible with many material surfaces.

Moisture contamination creates several problems for the older dry microabrasive blasting process. Aluinstallation minum oxide has been one of the best materials for basic abrasion, but it is highly sensitive to moisture. Moisture in the air lines would form clogs of abrasive in the system, forcing an extended down time to clear the clog.

This issue has virtually disappeared with modern blasting systems. Microabrasive blasting systems, like sensor designs, evolved and improved over the decades.

Three key variables affect the blasting process; the type of abrasive, the size of the abrasive, and air velocity. The process ejects regulated air blasts and abrasive mixture within a sealed work chamber attached to a dust-collection system.

A modulation system that mixes air and abrasive gives greater consistency to the abrasive flow over older venturi and shaker-type systems. The improved air-abrasive mix has led to more consistent results.

The introduction of Simoom Technology as used in the microabrasive blasting systems from Comco Inc., Burbank, Calif., highlights a more-recent advance. This abrasive mixing method uses an improved modulator design, an aerodynamic mixing chamber, and a PowderGate valve. The PowderGate eliminates pinch mechanisms, a major source of maintenance, for starting and stopping the abrasive flow.

The improved modulator and aerodynamic tank boosts efficiency of the abrasive stream for faster parts processing with lower abrasive consumption. Blasters equipped with Simoom Technology can handle a wider range of abrasives than previous designs, opening up new process applications older systems could not perform reliably.

One such application is in the surface preparation for multiaxis load cells. The operator prepares the sensing element surface for texturing by masking those areas that do not need abrading. The microabrasive blasting system controls air pressure and volume of abrasive in the airflow as preset by the operator. Optional timers prevent overblasting by the operator. This ensures a perfect, working bond for even highly customized load cells.

A final surface cleaning with acetone or lacquer thinner after abrading assures removal of any leftover abrasive and any oil or grease contamination that occurred during the abrading process.

Locating the strain gages
The normal method of accurately locating and orienting a strain gage on the sensing element surface starts by first marking the surface with a pair of crossed reference lines at the point where the strain measurement is to be made. These reference or layout lines were typically made with a burnishing tool rather than a scribe which could raise a burr or create a stress line. On many surfaces, a simple 4H drafting pencil was considered a satisfactory and convenient burnishing tool.

However, graphite marks are carbon and have a corrosive affect on aluminum. Many manufacturers now make special devices to hold and fixture strain gages. Therefore, rather than using pencil or other burnishing marks, the strain gages are actually put down on a rubber pad used to apply bonding pressure. Typically, this takes place under a microscope with crosshairs.

The tiny fixture that holds the strain gage also has reference datums that let the operator off-load the gage and tack it down with the glue used during the bonding process. When inserted into the sensor it maintains the alignment of the gage.

Load-cell geometry has several controlling elements. It must fit where it is going to be used. It must carry the applications worst case loading. And it must be adequately sealed to survive the application environment.

Current advances
Load-cell makers have always wanted to compensate for acceleration effects when using force and torque-sensing devices. Today, most force and torque-sensor families with integrated electronics possess acceleration compensation.

Force and torque loads come from acceleration and deceleration due to gravity, starting, stopping, and change in direction of a mass moving through space. Often there’s a need to measure contact loads while a tool or part is in motion. Until now, it seemed impossible to distinguish contact loads from forces and torques caused by changes in motion.

Simultaneous measurement of acceleration, force, and torque lets the sensor distinguish contact loads from the other forces. This permits control of contact forces and torques even in the presence of apparent loads.

The sensor integrates the signal-conditioning electronics for the many force and torque sensors into its body. The electronics includes amplifiers, analog-to-digital converters, EEPROMs for calibration data, and RS-485 serial drivers. A typical sensor outputs two serial data streams: a 2-Mbps stream for forces and torques and an additional 2-Mbps stream for accelerations. Both streams contain complete six-axis data sampled at 8 kHz and readable by serial receivers.

Acceleration-compensated force and torque sensors have the same load ratings and bolt patterns as their parent 6DoF sensor families. Typically they can measure linear accelerations up to 5 g on the X, Y, and Z axes and angular accelerations up to 200 rads/sec2.

There is a caveat with multiaxis , 6DOF/12DOF load cells. Even if the application only needs two axes, the load cells still come with a full complement of six axes at a minimum. It’s simply not economical for a multiaxis load-cell manufacturer to pull out strain gages. Also, calibrating all signals within the sensor improves the accuracy of the signals the application does need.

Coupling, reading, and calibrating all axes quantifies the accuracy to correct any minor errors. Therefore, even if the application reads only two axes, the processing of all six for accuracy makes the data from the desired two more accurate.

While all measurements start from an analog base, most interfaces today rely on a form of digital communications. Network-capable devices are now appearing with USB-compatible interfaces, following in the footsteps of their older siblings, the one-degree- of-freedom load cells.

Dongles are another advancement. While the basic sensor doesn’t change, this electronic adapter at the end of the cable takes standard digital output and adapts it to that needed by the interface. Thus, dongles can adapt and change communications protocols over the decades, keeping the sensor valid and current. There is no reason to change proven protocols within the sensor itself, so this external interface is one way to improve connectivity without affecting the actual sensors.

Issues with multiaxis load cells remain unique to the fields custom sensors address, with the exception of noise reduction common to all sensors. Applications are constantly changing and new ones added, always aiming at end user needs.

Where other technologies pursue gigahertz speeds and video-processing bandwidths, force/torque sensing remains relatively slow. Whenever things are touching and moving, there is mass involved and mechanical systems need time to respond. This technology nominally falls into the kilohertz realm, gathering data at a rate of thousands of samples per second.

Sensors today oversample at 8,000 times/sec for all six loads — three forces and three torques. The industry is getting ready to move that up to 10,000 times/sec. Not knowing what each application needs, multiaxis load-cell manufacturers always handle the base line of 5, 8, or 10 kHz.

The technology for 6Dof/12DoF load cells is continually evolving, supported by simultaneously improved technologies that make the manufacture of these devices easier and more economical.

All in all, the name of the game is customization. When designing for any application that measures force, torque, and acceleration, the main challenge comes in understanding exactly what will be needed from the sensor. The size, performance level, and often the materials that go into the load cells must match the level and environment of use for a perfect fit.

© 2011 Penton Media, Inc.