From liquid metal to finished jet engines, makers of sensors and inspection systems are putting in place devices and arrays that can take more-accurate readings faster and over wider areas. Often that is done in the processing line, with no need for sending samples to the lab. But as more and better data are available, vendors and mills are working towards smart feedback integration for corrective action without having to stop the manufacturing process. Ultimately that data will be made available in ways that can be used to optimize process control and give enterprise-wide systems a better window onto the shop floor.
We are the largest and the broadest provider of inspection technology in terms of types of sensors and in terms of industries served, says Holger Laubenthal, vice-president of Baker Hughes General Electric Inspection Technology. Metals is a very prominent segment on its own, as well as in terms of end-use sectors for metal components, notably automotive, oil and gas, and aviation.
BHGE sensor technologies include ultra-sound for pipe, tube and plate fabrication worldwide, as well as radiography in two and three dimensions. In radiography there has been a great deal of technology that has moved from the lab to commercial applications, says Laubenthal. Our first program is being rolled out now to a manufacturing line to provide speed, accuracy, and 100% coverage. Eddy-current sensors and field radiography are being used in testing aircraft-engine components.
The speed, accuracy, and coverage of new sensors and arrays are in response to metals producers and fabricators, but those capabilities do come with new requirements for systems integration, Laubenthal explains. We are talking about 20 gigabytes of data or more in a single scan. That needs automation around it to support the scan times and cycle times.
In a fully integrated system, sensors and fabrication operate at speed in automatic defect recognition (ADR). The core of ADR is always imaging, says Laubenthal, but now big-data analytical software is the essential part of corrective action in the manufacturing process. As we get more sophisticated with the types and number of defects that are detectable, the question becomes What is the source of those defects and can the inspection findings be fed back to the manufacturing process?
The major hurdle in the corrective-action feedback loop is that process control systems and defect-detection systems are not always compatible. These still need to come together, says Laubenthal. As we move from simple binary good/no-good detection and more to qualitative data, we are getting very specific error information. That information can be integrated with process control systems, but much work in that area remains to be done.
At even higher levels, sensors and inspection data are coming to the attention of departmental and company executives. What we are seeing over the past few years is a conversation around inspection that was once among quality managers and now is being engaged at higher levels in the organization, Laubenthal notes.
In one example, new sensors are able to detect corrosion in refineries and also in pipelines, and BHGE is launching a major new corrosion-control system to replace manual measurements. Data will be transmitted to any mobile device or to a process-control system. Laubenthal describes it as much of a production-management tool as a process control or maintenance system. Pilot and some commercial work is under way.
Ultimately it would be ideal for data to make its way back to the material source at the steel mill or aluminium smelter. We do not yet have that capability, says Laubenthal. We are still working on fabrication companies moving out of their silos to horizontal integration. If flaws are detected at step seven of ten, the cause may be at step two or four.
Primetals Technologies has seen a lot of work lately in the areas of in-line measurement of mechanical properties of finished strip steel in lines such as annealing and coating. Our latest development of this PropertyMon technology is for determining the magnetic properties of silicon and electrical steels, says Michael Stiftinger, vice-president of mechatronics at Primetals.
The first industrial installation will be in March. We think the results will be of interest to all producers because this type of testing has been difficult and costly, says Stiftinger. The development work that led to this new sensor system was done around both new processing methods and also the mathematics behind the sensing.
Primetals is also working on a transformation monitor using EMspec technology, developed by the University of Manchester, for hot strip mills. With an array of sensors along the length of the cooling runout table, it will be able to measure the phase transformation during and after cooling to ensure the finished strip is on specification, says Stiftinger. This monitor will be able to prove the process went correctly, to improve the quality of the steel produced, and to enable the mill to avoid further processing or ultimately downgrading.
Stiftinger agrees that automation of testing and integration of inspection data for corrective action is a major push within mills and sensor system suppliers. There is very big interest across the industry, Stiftinger notes, but many steel producers are not yet in a position of defining for themselves what they want to know. The goal is what Stiftinger calls through-process optimization. Under such a system we collect all quality-relevant data in the steel production and downstream processes, says Stiftinger.
There is also some interest in moving sensor findings up to higher levels. Customers tell us the final goal in big plants is to combine all the data in the system: production planning, process parameters and maintenance. Then you can optimize for quality or for maintenance in balance. That goal is definitely yet to come. I have not seen anyone that advanced yet. We are trying to support efforts in that direction.
New and retrofits
SMS Group has sensor systems in several of the newest mills built. Lots of the new systems are measuring magnetic properties, measuring dual and triple phases over the length of the strip, says Markus Reiffersheid, head of R&D for SMS Germany. That is done in a galvanizing line near the exit where the mechanical properties are established. That obviates taking samples and sending them to the lab.
We are monitoring temperatures with fiber optics, Reiffersheid adds. Automation of materials properties measurement is part of the intelligent furnace system. The mill operators can adjust furnace temperatures to homogenize properties across the strip.
Getting data up to higher levels of operation and management involves the same data center, says Franck Adjogble, chief engineer for process control and production planning systems for SMS Pittsburgh. The idea is to capture deviations to feed back so the process can optimize itself. It also generates reports for operators and engineers to improve the process. The same data center is used to display for upper management.
While SMS has already installed systems in some greenfield facilities, Adjogble says that the same product genealogy concepts can be retrofitted into existing plants.
If older plants have at least kept up with basic levels of process automation we can replicate the system, says Matt Korzi, vice-president of electrical and automation systems at SMS Pittsburgh. Perhaps they do not have the same number of sensors in place, and may have to add sensors, but we can integrate with any control system or interface.
More sensor companies are providing combined analyzers. Bruker AXS has new gas detectors that use non-dispersive infrared. That limits overlaps that can create cross-talk over a wide area, says Peter Paplewski, production manager for spark optical-emission spectrometry and combustion-gas analysis.
The systems also now use light-emitting diodes rather than a conventional light source with a chopper wheel. The LED gives a higher output, and also a clean on and off, says Paplewski. With the chopper wheel there was a sinuous light signal. The signal-to-noise ratio is now much better. This is actually a big step. It gives more precise readings and extends the life of the sensor.
Looking farther ahead, Paplewski believes that combustion-gas analysis will become increasingly important as three-dimensional printing with metal powders becomes more widespread. For example, the argon content of the powder is important, but no manufacturer will give original argon content. That number can be determined by using mass spectrometry.
In the same vein, hot isostatic pressing is not a new manufacturing method, but is new in 3D printing. But porosity is tough to detect, says Paplewski, and porosity can cause cracking. That is where the process-gas analyzer is necessary.
UniWest specializes in eddy-current testing, looking for microcracks in materials, and also to assess hardness in critical components in automotive and aviation. Those stresses put the task on us to find microcracks, even pores, says Tom Guettinger, manager for complex systems at UniWest. You cant just pull over at 37,000 feet if there is a problem with the engine. We are using smaller and smaller sensors. We put them together in an array of 128 next to each other and read them all at the same time.
The current capabilities of sensors and testing equipment have put something of a challenge back on metals manufacturers. The manufacturers want to test everything, and they would like to have it all automatic, said Guettinger, but fully automatic testing can drive costs up.
Communication between systems can be another hurdle. Guettinger notes that different clients use varying versions and ages of software.
Ironically, there can still be a challenge even after sensor systems are installed and communicating with the clients operating systems. Once some companies get sensor feedback and find high numbers of rejects, what they do about it sometimes becomes more about politics than about operations, Guettinger cautioned. The sensors dont lie and the manufacturers have to accept that. If something is found to be wrong, the question becomes: What does the manufacturer do about it?