Robotic assembly of HP DeskJet printed circuit boards in a just-in-time environment - Hewlett-Packard - technical

Robotic Assembly of HP DeskJet Printed Circuit Boards in a Just-in-Time Environment

THE VANCOUVER DIVISION of Hewlett-Packard designs and manufactures personal workstation printers. Advances in printer design and more efficient use of printed circuit board space have made major contributions to improving the printers' price/performance ratio. The number of printed circuit boards in each printer has declined significantly, reducing both size and material cost.

The printed circuit boards in the HP DeskJet and Rugged-Writer 480 printers are hybrid boards composed of a mixture of surface mount and through-hole components. Assembly of these boards is done on an automated high-volume assembly line capable of processing the boards for both pritners in a mixed-mode production environment. This robotic printed circuit board assembly system is carefully designed to fit into the Vancouver Division's just-in-time (JIT) manufacturing system.

A component mix of many small surface mount components coupled with a few large odd-shaped parts is a key factor in the design of the surface mount assembly line. A high-speed pick-and-place machine is used to place the majority of the small components. The remaining odd components are placed by the slower but more flexible robot workcell. The line is designed to allow mixed-mode production of different products with a minimum batch size of one. Each printed circuit board is automatically identified by the robot. Board loading information is communicated automatically to the robot controller via an RS-232-D connection to an HP 9000 Model 320 Computer.

Manufacturing Strategy

A just-in-time manufacturing philosophy has been effectively implemented at the Vancouver Division over the past five years. Work-in-process inventory has decreased dramatically since he implementation of the JIT system. The throughput time to process a complete printer from start to finish has dropped from over one week to less than four hours. On-line inventory space has been reduced to a level that allows twice the value-added shipments from the same floor space.

In a JIT system, demand for each subassembly exists only when the subsequent assembly creates the demand. Pulling a finished printer from the end of the line ultimately creates the demand for each preceding subassembly. For this reason, JIT production is termed a "pull" environment, with each assembly pulling the production from the previous workstation.

Extensive cooperation with the R&D design team is a requirement for products destined for automated assembly. Often the addition of a small feature that has no effect on the product's function or part cost can greatly simplify the automatic assembly of the product. Design changes made to improve manufacturability by automated equipment generally simplify hand assembly as well.

Robustness and flexibility are important features for automated assembly equipment used in a JIT environment. In a serial-flow JIT assembly line, the failure of any piece of equipment will cause the entire production line to stop.

The robot workcell described below is designed to learn the location of all important points in the work envelope using sensors mounted in the robot end effector. This feature has paid for itself many times over by greatly decreasing the reconfiguration time when adding or changing part feeder locations. A complete robot changeover has been accomplished in four hours with the aid of the self-teaching features incorporated into the robot software.

Use of Automated Equipment in a JIT Environment

It should be pointed out that automated assembly and JIT manufacturing are two entirely separate ideas. Each can be implemented independently of the other. By no means does one imply the other. Some of the most successful JIT assembly lines employ little or no automated assembly equipment. Highly automated factories exist that operate strictly in a traditional batch-type production climate with large work-in-process inventory levels. A key step in the Vancouver Division's manufacturing strategy is to integrate automation into the JIT environment.

Automation for automation's sake usually results in automating waste and the production of scrap material. To avoid this, each automation project is considered on the basis of improving quality, lowering production costs, and eliminating difficult or tedious manual tasks.

The precision assembly tolerances of [plus-or-minus]0.005 inch required for surface mount component placement makes this task difficult or unattainable in a manual workstation. A decision was made early in the robot workcell design to employ a machine vision sysem to ensure the placement accuracy required for zero-defect assembly. Quality levels have improved significantly since the implementation of the automated surface mount placement line.

The Robot Workcell Hardware

A two-phase process is used to place surface mount components onto printed circuit boards. A high-speed pick-and-place machine places all components whose size and packaging are compatible with the machine parameters. The larger PLCC (plastic leaded chip carrier) and SOIC (small-outline integrated circuit) components, which do not fit the size and packaging requirements of the high-speed machine are placed by the robot workcell. Several components are available in packaging compatible with either the robot or the high-speed machine. These parts can be switched to either machine to balance the machine cycle times for changes in product mix.

The robot workcell (see Figs. 1 and 2) consists of a Seiko RT3000 robot and controller, a machine vision system, a printed circuit board conveyor sysem, electrical interconnect hardware, and control computers. The workcell accommodates up to 16 tube-type parts feeders. Each feeder has space for up to four tubes, dependent on the component size. One machine operator is able to run the entire surface mount placement and solder assembly line. the operator's main task is to place printed circuit boards screened with solder paste on the input conveyor of the high-speed placement machine. The operator is also responsible for filling the component feeders as they are emptied. All placement operations and the movement of printed circuit boards between machines occur automatically.

An adjustable-width printed circuit board conveyor system connects the two pieces of placement machinery. Boards are moved along the conveyor using antistatic belts which grip the outer edgs of the boards. The belts are driven by step motors controlled by the Model 320 host computer. The workcell design will allow a second robot to be put in place as capacity and product mix needs dictate. Conveyor control and electrical interconnect for the second robot are already in place.

Pressure sensitive safety mats are used to guard the front and rear of the robot work envelope. These mats have proven effective in maintaining a safe zone around the robot while still allowing access and visibility. Facade plates mounted on both sides of the workcell guard against dust and enhance the aesthetics of the production environment. The facades are easily removed to allow access to electrical interconnect hardware.

Automatic Transfer of Component Loading Data

Component placement coordinates are downloaded directly from the printed circuit design computer system to the controllers of the placement machines. This capability greatly simplifies prototype builds of new boards and revisions to existing boards. Prototype builds, which previously were done manually, can now be run through the standard placement assembly line with minimal impact on production throughput.

Robot Operating Sequence

Boards arriving at the robot loading position of the conveyor are detected by a reflective sensor. The width of the loading section of the conveyor is narrowed to hold the board for component placement. This approach eliminates tooling pins and other associated tooling fixtures, but requires some extra software development to accommodate the lower-tolerance board fit in the clamped conveyor. A combination of vision and board graphics supporting the extra software overcomes the looser fit at a considerable savings in tooling costs.

The robot initializes the I/O lines and begins a search for the reflctive "landing padc located on the leading edge of each board. Printed circuit board design guidelines specify that all new boards will have a landing pad located att the same coordinate along the leading edge of the board. A reflective sensor in the robot end effector located the reflective pad. Along with the coordinates of the clamped conveyor edges, the landing pad provides a refrence point for the robot to locate critical board artwork details.

In the next step the reflective sensor is used to read a binary code incorporated into the board artwork, which tells the board type. The ID code dimensions and location are specified for all new boards. In most cases the component type and location coordinates can be called directly from the robot controller memory. When prototyping a new board, the robot will query the host computer for the file that contains component loading data.

Feeder locations and part parameters (number of leads, part height, etc.) reside in a data base in the robot controller memory. The component data base also contains vision system parameters necessary for the camera to identify and evaluate each part. The vision system will reject any part that does not have the proper number of leads. This feature has proven very successful in guarding against placing parts with bent or contaminated leads.

When the robot controller has identified the board location and type, the component loading sequence begins. The loading program was specifically developed so that the same program can load any board after the board has been identified by the robot. The board ID code points to a file that contains the loading information for each board. This file includes the number of components, the type of component, the feeder number, and the location of each component on the board.

Each part is picked from its feeder using a vacuum nozzle in the robot end effector. Each feeder is turned on for a fixed time before and after each pick. Feeder controler is through the Model 320 host computer. Four robot input/output lines are used to send control signals to the host computer for each of up to 16 feeders. Each part is moved over an upward looking camera which takes a machine vision image of the part. The part centroid and rotation are calculated from the reflections of the leads tips. A screening algorithm is used to block out the center section of the part and avoid confusion between the part leads and the writing that appears on the bottom of the part.

While the vision system computer is processing the part offsets, the roobt moves to the ideal placement location. The robot arm adjusts for the x, y, and theta (rotational) offsets specified by the vision system and presses the part into the solder paste. When all parts have been placed, the robot signals the conveyor control program to unclamp the board and move the board onto the exit conveyor. The fully populated boards exit directly from the robot workcell into the paste cure oven.

Workcell Design

An HP ME Series 10 computer-aided design (CAD) system was a key factor in the successful completion of the workcell design project. Design proceeded from a rough layout drawing, and each piece was added to the ME Series 10 drawing as the layout proceeded. Detailed drawings required for the fabrication of each piece of custom hardware were generated from the layout drawing after the layout was frozen. The importance of having a layout drawing that includes both the electrical and mechanical hardware cannot be overemphasized. The additional accuracy of the CAD tools leads to significant quality improvements and time savings in any complex design project. The integration of electrical interconnects and safety systems is greatly enhanced by including these features in a single layout drawing.

Acknowledgments

The success of this project would not have been possible without the contributions of many people. Special thanks to Marc Hartquist, Britt Freund, Ken Wade, and Rick Hughes for their participation in the workcell development team.

COPYRIGHT 1988 Hewlett Packard Company
COPYRIGHT 2004 Gale Group