Report on Best Practice Story 4 “Anomaly Detection”

In Use Case (UC) 4, the consortium partners, Arendar, PDTec, VPE, RapidMiner and IPS work together under the leadership of PolyMerge. The UC addresses an industrial potential- and data analysis in production. The focus is divided into three main groups, data collection and data analysis, processing of the obtained data and a potential analysis. In the work carried out to date in UC 4, the available process data from the plastics welding process presented as an example in the UC were primarily identified and interpreted regarding their significance.

Plastic welding

Plastic welding is a process based on the plastic being converted from a solid to a molten state in the weld area of the component halves. In most cases, this is done by heat supplied from outside. In the case of the hot plate welding process considered in the project, this is a heated piece of metal, the so-called heating element (HE). The welding cycle follows a repetitive pattern. The component halves are fixed in so-called component holders. The heating element is placed between the component halves by means of a movable axis. The component halves are then positioned so that the weld area (weld land) comes into contact with the heating element. Contact with the hot metal surface causes the plastic to start melting. Since plastic components are usually uneven or warped, at the beginning of the melting process the moving axes of the machine melt off and displace a small part of the weld bar so that full-surface contact of the weld bar with the heating element occurs (approximation phase). Once the so-called alignment distance has been reached, the component axes stop, and the component halves remain on the heating element until a defined melt cushion is reached by heat conduction into the depth (heating time). After the heating time has elapsed, the component halves are removed from the heating element and the latter is retracted between the component halves (changeover phase). Once the heating element is removed, the component halves are pressed together, displacing part of the melt into a weld bead, and the component halves form the material bond with each other by cooling (joining phase and cooling phase, see Figure 1).

Figure 1: Process – Hot plate welding method

Aim of the work at PolyMerge

With the expert knowledge of the process available at PolyMerge, the following insights are expected from the data collection for plastic welding:

The heating element must be at a specific, material-dependent temperature to melt the plastic. The temperature window here ranges between 120 and 480 °C, depending on the plastic being processed. As part of the data storage of the UC, this temperature, which is already measured and monitored by the controller of the machine’s programmable logic controller (PLC) anyway, should also be stored continuously for analysis in the UC. If it is necessary to reduce the amount of stored data, it is possible to store only the temperature shortly before the start of the adjustment and warm-up phase until its end. However, a full temperature recording would be preferable.

The axes of the machine, which move the HE and the component halves of the welded parts into the respective positions and, if necessary, also apply pressure, can be monitored by several parameters:

From the start of the movement, which is triggered by the PLC, the time can be measured until the planned end position is reached. In the case of pneumatic drives, this can be done by measuring the time between the triggering of the valve and the signal from the end position sensor (or displacement encoder). In the case of a servo-motor actuator, by the internal travel measuring system and the time signals between the initial and end positions. The time needed to reach a certain position can give information about the friction of the cylinder rods and the guides. If a longer time is required to reach the position, a potential defect may be present here. In the case of a pneumatic cylinder, there could also be a defect in the compressed air system (leakage, clogging, cylinder fatigue), which is causing the cylinder to be pressurized too low.

If the drives are additionally equipped with a pressure sensor (pneumatic cylinder) or if the electric current (servo drive), which is required for the movement, is monitored, additional information about the friction forces of guides and drives can be obtained. With electric drives, it would even be possible to evaluate acceleration and deceleration values. For pneumatic actuators, there is also the additional possibility of measuring the amount of air (flow) required for a movement. However, the sensors for this are usually comparatively expensive, so that this type of monitoring is rarely used with spatial resolution, i.e. per actuator. In current machines, this is currently not monitored at all in such detail, so that new insights can be gained in the context of the UC.

Potentials for the process phases

For the individual process phases, the data can provide the following statements and indicate problems in the process:

Alignment phase:

A combined travel-time measurement can be used to determine whether it always takes the same amount of time to reach the alignment position. A delay could be caused by problems with the drive (sluggish), the material (wrong plastic) or the HE temperature (too low). If the position is reached too quickly, there could also be a material mix-up, or a HE temperature that is too high. By additionally monitoring the force, or the required drive energy (current or air pressure), it is possible to further narrow down the potential problem. For example, if the force required is lower than in the comparative data when the adjustment time is shortened, the viscosity of the plastic could be lower (wrong material, temperature too high).

Warm-up phase:

At the end of the heating phase and at the beginning of the changeover phase, the molten weld bars of the component halves are detached from the HE. A load cell installed in the system can measure and monitor the detachment force. In conjunction with a displacement measurement, a force-displacement curve can be recorded and analyzed. At present, there are no findings as to whether these data contain any information about the properties of the process. Therefore, it is interesting within the scope of the project to gain more in-depth knowledge about these data and to analyze them with the help of the existing expert knowledge.

Joining phase:

In this phase, the molten weld webs are pressed together under pressure and the material bond is created in the welded joint. Force-displacement and force-time curves are also recorded and analyzed here. In this process step, as in the approximation phase, influences from the drives, guides, melt temperature and the material itself can be detected and characterized.

Through the conceptual preliminary work in the current period, these possibilities were discussed and the effort weighed up.

Regarding the transferability of the findings from the UC, conceptual considerations have also been made as to how the sensor data can be beneficially used in related automation applications. The data of movements with simultaneous recording of the duration for the movement can be transferred to any actuators that move or position components, to press-in processes of inserts or clip processes. The extent to which the knowledge gained from this adds value and the integration proves financially beneficial must be assessed on a case-by-case basis.

The energy requirement, for example, allows conclusions to be drawn about possible wear on the system. In relation to the component, faulty component tolerances in press-fit or clipping processes, for example, could be responsible for increased force or energy requirements.

Outlook

In future work, data will be collected and analyzed both in laboratory test specimens with artificially introduced defects and in industrial series production. For this purpose, ARENDAR IT Security GmbH has provided an ARENDAR with numerous input modules that can receive data from the machine. The goal is to use Artificial Intelligence to detect conspicuous patterns that can indicate defects in the welding process or in the welded component.

Author and contact person: 

Share This Story, Choose Your Platform!