Minimizing unplanned downtime using data collection and processing

The flow of cardboard sheets through the factory is designed according to the nominal shape of the boxes, but after the boxes are stored in the warehouse, they can take a concave shape compared to the original. The concave shape is probably due to changes in humidity. See the image below to see the difference between the cardboard sheets before and after they have been stored in the warehouse.

The concave shape of the sheets causes the cardboard sheet to stick to different parts of the process and the robot's difficulty in picking them. To see the working flow of the cardboard sheet, see video below.

The first problem that can occur is when the robot picks the sheet from the tray. To move the sheet, the robot uses 8 suction cups to create a vacuum. This vacuum is also what gives feedback to the robot that the sheet is stuck and that it is good to move on to the next step in the process. When the cardboard sheets are bent, it is not possible for all suction cups to create a vacuum, see image below, which means that the robot does not continue its sequence but continues to try to create a vacuum with all 8 suction cups.

Another problem that can arise is when the cardboard sheet must pass through the punch and because of the concave shape it ends up wrong. If the cardboard sheet is concave, the cut-out may be offset in some direction, which may cause the pressure on the sheet to be misplaced. The fact that the cardboard sheet is bent also creates an increased risk that the sheet may be further deformed because if the height of the sheet is higher than the difference between the rollers, the cardboard sheet is pushed through the punch and a crashing sound occurs.

The last problem that can arise is when the cardboard sheet cannot be delivered from the robotic cell because the slot through which the sheet is transported is too narrow when the sheet is bent. See the image below to see how the carton sticks.

All the above problems can create unexpected downtime, but with the help of sensors, these downtimes can be predicted and avoided. The factory has been equipped with sensors to measure humidity, temperature, vacuum, distance and vibration. A vision camera also measures the flatness of the cardboard sheet before the robot picks it up. By measuring flatness together with other data, hopefully, certain relationships can be identified.

The vibration sensors should be able to determine if the cardboard sheet has not gone through the process, for example, by identifying the special sound that becomes when the cardboard gets stuck in the process or does not fall through the gap at the factory outlet.

The sensors used come from IFM, see image below. The data goes via a gateway and is stored in the cloud for processing. Application for receiving the data is done using Cybercom's “Machine Monitoring” which is a backend solution they developed. The data can also be visualized on a dashboard in real time and trend curves can also be developed.

The next step is to use machine learning to predict breakdowns. The operator could receive decision support to perform any action or the machine would act autonomously to prevent a breakdown, for example, by sorting away boxes that deviate too much from nominal flatness.