| Problem Statement | Improvement Approach | Benefits |Company Contact Information |
Hewlett Packard needed to increase output on a final inkjet cartridge assembly line with minimal investment.
Hewlett Packard (Ireland) manufactures inkjet cartridges. The final assembly part of the process assembles and tests the pens. These lines run at a rate of around 1500 units per hour. They consist of conveyor loops; with pens travelling around on a limited number of pallets. Often there is a buffer between each loop. At this high rate of manufacture, the equipment tends to go down every hour, normally as a result of a jam or a re-test. Therefore the sizing of conveyor buffering between equipment, on the loop, is critical. For example, assume Mach_2 feeds Mach_1 and Mach_1 goes down, processing on Mach_2 will be blocked if the conveyor buffer is too small. However if Mach_2 does goes down it will starve Mach_1 if the conveyor buffer is too large.
Conveyor logic at junctions also can cause unnecessarily starvation of or blockages to equipment. The problem was to increase output by 10% to 15% by: sizing the conveyor buffers, identifying real bottlenecks, evaluating and modifying the conveyor logic and defining the optimum number of pallets on each loop.
One of the loops on the line
Simulate assembly line using read downtime data. Evaluate HP proposals and recommend a cost effective solution.
HP management decided to use simulation to improve the performance of their lines. In the US they used simulation at a high level but not at equipment level.
Over two projects CIM Ireland was asked to look at the Formula line. The approach for each of these projects was to work with a line engineer to build the model. This model was built using actual downtime logs for each machine based on a good, bad and average day.
The model was validated by: 1) Having the line engineer involved in the building process so that he could understand the model logic and physical details; 2) By comparing actual outs per hour with simulated outs. The model accuracy was 99.47% with a correlation factor of 0.926. See Figure 1.
Having validated the model, the next step was to identify where improvements could be made. The first experiment was looked at adding a new buffer between two loops.
This did not result in a sufficient improvement. The next step was to add buffers in between each piece of equipment, Then run a series of experiments by varying both the number of pallets and which buffers were active. 500 experiments were ran each day.
As a result one location was identified as the main constraint. We modelled an intelligent buffer (where the buffer was used only when it was needed). This resulted in an 8% improvement. The next step was to improve the conveyor logic by monitoring blockages. The logic was changed at three points, this allowed more pallets on the loop and resulted in an additional increase in output of 7%.
