Quality Measurement: Multitasking with Multisensor Measurement
Multitasking and rapid changeovers are not just concerns for process engineers anymore. Today, they are important considerations for quality engineers too. As manufacturers continue to move quality control from the lab and integrate it into flexible production lines, inspection devices must do more per setup with the least human intervention possible. Consequently, many quality engineers are designing or specifying automatic measurement devices that deploy more than one sensor to take as many measurements as possible while the part is located in a fixture.
Such was the case for a test bench that Captronic Systems Pvt. Ltd. (Bangalore, India) built for a manufacturer of handheld power tools. The manufacturer's inspectors had been checking the performance of drill guns, angle grinders, miniangle grinders and marble cutters with a variety of stand-alone instruments and recording their measurements by hand. Not only was the process slow and prone to human error, but it also made trend analysis quite cumbersome.
"Since all the measurements were taken manually in an asynchronous manner, it was difficult to relate each parameter with the others," explains Mondeep Duarah, senior manager of the System Engineering Group at Captronic, a builder of custom automation and test and measurement equipment.
The automatic test bench solved these problems by acquiring data from a variety of sensors, which plug into a chassis designed to accept them and their signal-conditioning units as modules. To hold the various power tools coming to the bench for inspection, the engineers designed a fixture for each tool. Each one contains a unique array of t-type thermocouples and accelerometers at positions appropriate for the particular tool. The bench contains the rest of the sensors: a Hall-effect sensor, microphone, infrared temperature sensor, proximity sensor and dynamometer.
Once the operator selects the type of tool to be tested from a list on the controller's screen, enters its serial number and places it in a fixture, the bench performs various measurements automatically. It runs the tool for a short time without a load, measuring current, voltage, power, vibration, noise, speed and temperature. It then applies the specified torque gradually using an eddy-current dynamometer and coupler and measures torque in addition to the other parameters. A buzzer tells the operator when each test is finished, and a window appears to display the results.
Software, the keystone
Meanwhile, LabVIEW software from National Instruments Corp. (Austin, TX) records the results in a historical database that engineers can access and analyze later. The software inside the bench does more than collect, display and store data, however. It also consolidates the multisensor measurements and their individual analyses into one application that makes them available through an easy-to-use interface.
"Without the ability to configure, acquire, analyze and present multiple types of sensor and signal inputs from a single software interface, this system would be very impractical to use," says Ryan Wynn, product manager of data acquisition systems at National Instruments. In Captronic's case, the underlying technology is Virtual Instrumentation, software from National Instruments that creates the necessary definitions and environment for accepting input from disparate sources.
The evolution of such software is crucial to both the effectiveness and practicality of multisensor measurement.
"The increasing pressure to push products to market faster and less expensively and the continuous evolution of sensing and computing technology (such as IEEE 1451.4 plug-and-play sensors) have driven multisensor measurement systems," says Wynn. "At the heart of this evolution is software."
Software, as well as the base, also is crucial for the kind of multisensor measuring machines offered by Werth Inc. (Old Saybrook, CT) and its competitors. Because Werth's engineers base the company's machines on optical sensors and supplement them with other sensors, they developed a series of four standard designs that use plug-and-play software and hardware. Each machine then allows operators to plug a touch-trigger probe, touch scanning probe, laser gage or other sensors into the base machine to add capacity to the optical sensor that comes with the machine.
"Our multisensor machines use an optical sensor as the primary, or reference, device," explains Ralf Herzog, vice president of engineering. "The reference sensor gets mapped to the machine's axes, and all the other sensors are mapped to the reference sensor."
Consequently, not only can the sensors communicate with each other and work together, but operators also can use them interchangeably. "Basically, you can write a program just by selecting different sensors from the same software," says Herzog. He adds that this integration is usually most cost-effective with the group of sensors that the builder either owns or offers regularly. Otherwise, "using third-party sensors will increase the risk of costly software integration."
The superstructure of the machine also contributes heavily to making this class of multisensor measurement machines practical. The reason is that between 70 and 80% of the cost of a precision measurement machine is in the base, actuators and controllers, according to Herzog. The actual sensing mechanism contributes only 20 to 30%. Consequently, an additional sensor might make buying a second machine unnecessary and save the cost of redundant elements.
And that does not include the savings that accrue from putting parts in fewer fixtures and having the correct tools handy on one machine. "It's always about flexibility," says Herzog. "With more sensors, you have much more flexibility to measure more features [while the part is held] in a particular orientation."
An explosion of sensors
Although software is key to making multisensor measurement work, the growing number of sensors available today also contributes to the success of the concept. "There has been just an explosion of technology in the sensor industry," says Randy Adair, president, Berran Industrial Group Inc. (Kent, OH), a builder of custom machines that work plastic parts made primarily for the automotive and bottling industries. "Vendors keep coming out with better ways to sense goods. They also are making them more programmable without any change in pricing. Five years ago, for example, many applications of cameras would have been cost prohibitive."
An example is the custom leak testers that Berran builds for inspecting ducts for clean-air intake systems in automobiles. Because of the drive for 100% inspection in the automotive industries, "leak testing has evolved into more than just leak testing," explains Adair. "Because it has to occur at the end, it's a logical place to attach other kinds of sensing devices. In essence, a leak-testing machine becomes a QC machine for testing any external devices that are attached." That is why engineers have added vision sensors to the mass airflow, pressure, vacuum and proximity sensors already on the machines.
In vacuum mass-airflow testers, the designers specify two kinds of vision sensors from Banner Engineering Corp. (Minneapolis). Each performs a different task. A low-cost PresencePLUS P4 GEO simply verifies the presence and orientation of stainless steel hose clamps on the ducts, and a more sophisticated PresencePLUS Pro uses pattern recognition to verify that a temperature sensor inside the duct has been installed correctly and is not damaged.
Inspection occurs in two phases. First, the P4 GEO looks for the presence and position of a clamp that is attached to the duct, and afterward the tester conducts a vacuum mass-airflow test to check for leaks. After the leak test, a shuttle transfers the PresencePLUS Pro into position to inspect the temperature sensor inside the duct. If the duct fails any of the inspections, the device's human-machine interface (HMI) alerts the operator, who either corrects the problem and retests the duct or determines it to be irreparable and rejects the duct.
The vision sensors suit these tasks well because the clamps are left open for downstream assembly. Because some clamps are open slightly further than others, their slightly different orientations can confuse proximity switches and other sensors that focus on a specific point. "Even though a clamps are in the right place, a proximity switch might not detect it because the part might be a sixteenth off," says Adair. "The beautiful thing about cameras is that they can tell where an object is" and find the features that they are looking for. So the features need be within the camera's field of view, not in exactly the same place every time.
Sensors for mass production
Multisensor measurement is not limited to audits or low-
volume production. It also can automate inspection on mass production lines. Marposs Corp. (Auburn Hills, MI), for example, has been combining linear variable differential transducers (LVDTs) and eddy-current technology to inspect automotive parts inline for years. The transducers measure the dimensions of various features, and the eddy-current detectors check for cracks during nondestructive tests. Depending on the required accuracy and precision, some machines contain a few extra transducers to monitor the mechanics of the machine and compensate for position
errors introduced by the mechanical actuators.
In one crankshaft-inspection application, an automatic machine performs more than 100 measurements within roughly 30 seconds. After comparing the measurements to the control limits, the controller tells the machine either to move the crankshaft along to the next operation or to direct it to a reject station and inform the operator which journal, pin or face is bad.
Recently, engineers at Marposs applied the lessons that they learned from crankshaft inspection to checking disc-brake rotors. They designed an automatic machine that not only measures thickness, height and other dimensions dynamically on the braking faces with LVDTs but also looks for cracks on the braking surface with eddy current. When rotors arrive, the machine puts them in the gaging station, locates them there, and performs the measurements and eddy-current checks, activating the appropriate sensors at the right time. When the machine releases the part, it sends good parts down the line and diverts bad ones to a reject station.
The strategy for handling rejected parts is an important consideration on mass-production lines, according to Gary Sicheneder, manager of new market development at Marposs. Some manufacturers want to segregate rejected parts based on the type of failed inspection check-for example, rejection due to a crack in the part or a dimensional measurement. If all of the rejected parts are diverted to only one bin, then a person will have to sort through the parts to try to determine the type of failure. So manufacturers that want to trace rejects automatically need to put some thought into a scheme for separating them.
Despite rising interest in automatic gaging systems that perform several measurements with a variety of sensors, Sicheneder notes that a large number of applications in mass production still need electromechanical and hard attribute (go/no-go) gages for audits on the factory floor. Many factories relying on this kind manual intervention, however, want to automate data collection as much as possible. Consequently, they are asking gage manufacturers to link the various manual gages on their inspection tables to computer networks.
A case in point is transmission production at one of the automakers' plants. Engineers there had asked Marposs for a set of manual audit gages dedicated to checking bores in an aluminum part for one of its transmissions. In the resulting bench-mounted system, inspectors use three electromechanical plug gages to measure the diameters of three bores. Some of the eight pneumatic plugs on the bench also are dedicated to specific bores, but others measure more than one diameter.
The pneumatic plug gages are necessary in this application because some of the bores have very smooth surfaces that cannot tolerate any contact marks that mechanical inspection might leave. "So you can't touch them," says Sicheneder.
All of the gages feed their measurements to a Windows-based E9066 industrial controller that Marposs developed for computational intensive applications. Not only does the controller display results and record them for troubleshooting and continuous improvement studies, but it also leads operators through the inspection process, telling them which features to check and what gages to use. In normal use, the controller will not accept measurements taken out of sequence or with the wrong gage. The operator, however, can switch to "casual mode" to work outside the normal routine to troubleshoot a recurring problem in manufacturing.
Sicheneder reports that a number of users are exploiting the capabilities of the PC-based controller further. After computing the statistics in real time and displaying them locally at an inspection station near the production machinery, the controller can upload data to a network for monitoring Cp, Cpk, and Ppk values. "In some cases, the customer makes the correction from a computer terminal right at his desk, instead of having to send a technician to the floor to make an offset," he says. This is the ultimate way to multitask with multisensor measurement. Q
• Many quality engineers are designing or specifying automatic measurement devices that deploy more than one sensor to take as many measurements as possible while a part is located in a fixture.
• Although software is key to making multisensor measurement work, the growing number of sensors available today also contributes to the success of the concept.
• Multisensor measurement is not limited to audits or low-volume production. It also can automate inspection on mass production lines.