Ceramic Component Metallization with Custom Robotic Dispense Systems
Automating as much of the ceramic metallization process as possible can drastically reduce costs, decrease variability, and improve process compatibility.
Ceramic metallization enables the joining of ceramics to other ceramics or metals via eutectic or pure metal alloy brazing processes. Ever since the advent of ceramic-to-metal brazing in the 1930s, the brazing industry has relied primarily on hand application for the refractory metallization pastes needed to create the braze seal. For a complex joint, skilled technicians (artists, really) painstakingly apply the paste using a variety of methods, typically a paint brush, pen or spray. Operators will often spend several minutes on individual parts to ensure that the applied paint meets all the dimensional, positional and thickness requirements needed for a hermetic joint.
For complex parts in quantities of up to 50-100, hand painting by skilled operators is an efficient process. For high-volume, simple parts, where automation could provide multiple benefits, hand painting is still the norm in our industry. Although semi-automation tools such as rotating fixtures are often used to process parts more quickly, each operator has their own application technique, which leads to high part-to-part variability in film thickness, quality, uniformity, and application time. Ideally, a single coat is applied; however, the paint must be free of voids and pinholes so that the subsequent Ni plating required for brazing is uniform and light tight. This may require two coats due to the inconsistent nature of hand application.
In order to ensure that the right amount of paint has been applied, parts are often weighed after painting, resulting in even longer processing times for reworking to either remove or add paint. Even with semi-automated fixtures, parts are still handled individually. The monotonous nature of repeatedly painting the same part requires that operators take frequent breaks to prevent repetitive stress injuries and boredom.
Automating as much of the process as possible can drastically reduce costs, decrease variability and provide opportunities to apply statistical analysis to improve processes. When designed correctly, automation enables high-volume manufacturing of components with multiple diameters and paint compositions, providing the opportunity to increase daily production from hundreds of parts per day to thousands.
Automation can provide superior part-to-part paint thickness consistency and uses paint more efficiently due to accurate and repeatable dispensing systems. Very little paint is wasted as a result of rework and inconsistent application. In many cases, the ability to precisely locate and apply paint eliminates the need for masking parts.
The best candidates for automated metallization are round components that are symmetrical around a center axis, don’t require masking and are metallized on their outer surface. These properties allow for fast loading of blank parts into cassettes, provides more options for end-effector grip tooling, and makes final smoothing simpler.
System and Robot Design
Metallized and nickel-plated ceramic parts are sold in high volume (up to 100,000 pieces per month) and have a geometry ideal for automated metallization. When early internal automation efforts did not provide the required efficiencies to produce these parts, a robotic workstation was developed to provide the following capabilities:
- User-friendly human-machine interface (HMI) to enable machine operation, process tuning and error recovery
- Multi-level safety interlock system to prevent operator injury
- Easy cassette loading of parts to be coated
- Systems capable of applying multiple paint formulations for both metallization and glazing
- Flexible computer-controlled dispense valve systems for accurate paint metering
- Position control for reproducible stand-off height for paint application to three-dimensional surfaces
- Flexible smoothing system to ensure paint uniformity
- Dual, identical, sealed robots with 6-axis rotation and accurate position control of less than 0.1 mm
- On-board heated drying/storage stations
- Computerized vision system for in-process inspection
- Exit conveyor for painted parts loaded onto firing setter plates
These design requirements dictate that many correct choices need to be made early in the design of the robotic system. The most significant are dispense valve technology, compatibility with multiple ink formulations, and robots capable of high throughput. Because metallization pastes are being produced, the ability to tune paint viscosity, particle size distribution and solids loading was incorporated to enable repeatable dispensing at the application frequency required by high-volume manufacturing.
Dispense Valve and Paint Rheology
It is vital to be able to control paint rheology in order to enable the rapid metallization of smaller components. Prior to sintering, applied paint thickness is characterized by component weight gain and calibrated X-ray fluorescence (XRF) measurements. For high-volume, reliable dispensing, the paint needs to have an adequate pot life to enable reliable fluid properties for over 8 hrs of application time. Shifts in paint solids loading or agglomerated particles can disrupt production or inadvertently change the composition of the paint while it is in the dispense apparatus. Some paints or glazes require continuous agitation or stirring to produce reliable dispense repeatability.
Proper milling of the paints is also required to prevent dispense valve clogging or internal damage from paint particles. Therefore, having a well-controlled and characterized paint manufacturing process that considers high-volume dispensing is critical for automated ceramic metallization. At a minimum, it is important to use a well-milled paint that is verified by checking high- and low-shear viscosity, in addition to milled particle size distribution and LOI (solids content).
Dispense valves come in many types, volumetric flow ranges and levels of sophistication. Most valves are pressure driven, while some more advanced valves are volumetric. Pressure-driven dispense systems with manual valve activation can help each artist meet their own unique style of dispensing the required amount of paint for each component.
Due to the inherent variability of human-activated pressure-based dispense valves, it is difficult to maintain better than 3 mg part-to-part paint mass reproducibility. This correlates to a part-to-part paint dispense thickness variability of 10%, or 1 sigma. This dispense thickness variability is also operator- and takt time-dependent, since different operators use different feed pressures and triggering times for applying thixotropic paints at high volume. Since paints are shear thinning, operators who operate their dispense valves at higher feed pressures can be dispensing on a completely different fluid flow regime than others. Application time can vary widely, from a few seconds to up to 1 min for the same component. This influences both final paint thickness and overall productivity (parts per hour).
Operators typically have a personal mix of valve pressure and dispense timing that works for them, which ultimately means that operators do not dispense the paints the same way and therefore do not produce identical parts. Thus, it is not possible to use Six Sigma process techniques to optimize human-based paint dispensing for metallization.
Automated dispensing, or fluid jetting, using the properly selected dispense valve provides more reliable part-to-part paint thickness and weight gain. A graph illustrating the performance of an automated Mo/Mn ink dispensing system can be seen in Figure 1. A 10x more consistent paint dispense mass reproducibility can be achieved at a dispense rate 50% faster than the most capable high-speed painter.
Compared to artists, computerized paint dispensing can easily be tuned to achieve high process capability at high production rates for overall improved repeatability. In addition, since valve on/off time is uniform, feed pressure can be established, and part-to-part dispense mass (paint thickness) can be controlled by tuning paint rheology and LOI. Paints can be prepared, characterized, and stored well in advance to provide a consistent and well-controlled process over long production runs.
Qualifying the Process
During the qualifying process for the automation cell, a modified Turing test was used to determine if the robot’s output was as good as or better than that of a human. The Turing test was developed by Alan Turing in the 1950s to grade a machine’s capability to replicate a sentient human in completing a task.
A custom test was developed that evaluates a total of 13 different criteria to compare robot-made components to those made by skilled artists. Ironically, many process improvements have come from implementing skilled operator techniques into the code used to drive the robot kinematics.
Process development requires multiple component runs to optimize Turing test scores and ensure a metallizing and/or glazing process that is far more capable than a human artist. These criteria range from quantitative (e.g., the mass of paint deposited, throughput and yield) to qualitative (e.g., the surface finish of the paint after smoothing, part-to-part consistency and ease of load/unload). For both robots, a letter grade of A-F was given for each of the 13 criteria; a total score for each robot was used to assess how it was performing. The robots were treated individually, since one robot completes tasks in a left-handed manner and the other robot in a right-handed manner. A grade of C is awarded if the robot meets human capabilities.
In general, it is easy for the left and right robots to exceed human performance in terms of part-to-part paint thickness control, process capability to specification (CpK), and overall component throughput. Challenges arise if components fed into the robot are non-ideal, variable in size and have inconsistent geometry (e.g., press steps, shrink variability). Figure 2 shows an example of a test process evolution/qualification, where scores start below human levels and eventually far exceed the performance of an artist.
Having well-controlled incoming parts for metallizing helps enable easier automation. Figure 3 shows a comparison of robotic- and artist-prepared metallized components. In most cases, the components to be metallized are controlled in position to less than 0.002 in., so the robotic process can manage the variability. Experienced artists have an amazing adaptability to manage proper paint application to non-ideal parts. When considering automated, high-volume metallization, however, incoming components must be well controlled. For low-volume and complex components, automation does not make sense unless very precise dispense repeatability (less than 0.5 mg, 1 sigma) is required by the application.