Causes unclear: Rejects on the assembly line are too high, cycle times are increasing minimally but steadily, but at first glance it is not clear what the cause is. With large assembly lines, it is often difficult to locate problematic stations or processes in order to identify the potential for optimization. Collecting production data using a software application called 'piaOptimum' can be the solution.
Bottlenecks in complex interlinked assembly systems can be identified and, thanks to data transparency, production efficiency can be quickly optimized. The analyses and evaluations that are possible with the application will be explained using the example of a large assembly system for rear axle transmissions at a German automotive supplier.

The assembly plant comprises around 50 stations, all of which are monitored using the software application. It is also possible to take a closer look at individual stations or line sections. However, in order to be able to make statements about OEE (Over-all Equipment Effectiveness) losses, it makes sense to look at the entire plant in order to identify bottlenecks and faults as a whole. A station is defined in the software as a part of the system where one or more processing steps are carried out that have the same target cycle time from the start of processing one component to the start of processing the next component as the system as a whole; in this case 55 seconds. This means that a rear axle drive is completed every 55 seconds. The stations are generally set up in series.
However, there are also exceptions where it is not technically possible to complete the processing step in 55 seconds. However, in order to prevent slower stations from creating a bottleneck in the assembly process, these stations are designed as double or multiple integer multiples of the target cycle time. As a result, a system cycle time of 55 seconds can be achieved again.

Data connection via OPC UA without PLC intervention

By using the software application, the user wants to create transparency for optimizing the output of his system. This requires data relating to the basics of overall equipment effectiveness (OEE), i.e. information about the component quality at the end of a processing step (i.e. a station), the associated cycle times and any faults and messages that occur. All data points can be configured easily via the software's web interface. No intervention in the machine control system is necessary; piaOptimum enables a flexible data connection using the standardized OPC UA protocol - provided that the minimum requirements of the line control system are met. While the data for component quality and cycle times can be easily generated via OPC UA, details about messages and faults can only be retrieved via HMI panels, whose compatibility with the customer's respective network structure must first be checked and in some cases adapted.

Station 8 is used below as an example to show how an analysis at overall cycle level (from one processing start to the next) and a partial cycle analysis ultimately resulted in a demonstrable improvement in the cycle time.

Analysis and optimization of a press station

Figure 1: Extract from an overall clock analysis as a boxplot diagram

© PIA

The boxplot diagram (Fig. 1) is an excerpt from an overall cycle analysis. The 50 % of the middle bars are shown here in ascending order. The software application uses this tool to hide outlier cycles that cannot be regarded as 'normal' cycles but are also recorded. The stations are displayed in line order on the x-axis and the time in seconds on the y-axis. The white line in a red or blue box represents the median, which represents the station's cycle as an outlier-proof average. The box height shows the fluctuation of these cycles. The target cycle of the system is 55 seconds - shown in the graph as a horizontal blue line.

The analysis shown was carried out in the period from calendar week 23/2018 to calendar week 25/2018. At station 8, a press station, it can be seen that the median of 58.5 seconds was slightly above the target value. Optimization of the cycle time was therefore necessary here (as at stations 7 and 10, which are not considered in more detail here). At station 8, the following processing steps take place in the target cycle time: 'Actuators ready to hand', 'Auxiliary time', 'Component in position', 'Clamp component', 'Pressing process', 'Units back'. A deeper analysis of the partial movements made it possible to achieve a reduction in the overall cycle time. The optimization measures were carried out in calendar week 26/2018, after which it became clear that the adaptation of a special sub-cycle called 'auxiliary time' was the main contributor to the reduction in the overall cycle time.

Figure 2: Machine cycle share of station 8 compared to the reference period

© PIA

The software application offers the option of using a reference partial cycle widget to view the analysis down to the individual movements in the station and to compare both total and partial cycles with a reference. The three weeks CW 23/2018 to CW 25/2018 were defined as the reference period, and the entire following month of July 2018 as the analysis period. The reference is always shown as a blue bar in the graph, while the current analysis values are shown in red, yellow or green, depending on the result of the comparison.
The machine cycle rate in July fell by 4 % compared to the pre-optimization period (Figure 2). When looking at the individual steps in the station, the main culprit can be identified at first glance: Although the 'Actuators ready to grip', 'Component in position' and 'Units back' steps take slightly longer, the optimization of the 'Non-productive time' movement was the real deciding factor.

Machine cycle reduced by optimizing the 'non-productive time'

The sequence shows that after this step, all subsequent steps start earlier and the overall processing time could therefore be shortened. The 'non-productive time' is a non-automated movement performed by an operator. The optimization consisted of providing the operator with a better tool for processing - a small change with an amazing effect on the total cycle time of this station and ultimately also on the GAE of the plant.

Of course, this improvement must also be recognizable in the overall cycle analysis (Fig. 3). For the following month of July, the median cycle time was reduced to below the target cycle time - 53 seconds in this case - and the fluctuations were also greatly reduced.

Document the reporting process

In addition to the total cycle and partial cycle analysis, piaOptimum can be used to display other informative evaluations and use them for production optimization, such as displays of production statistics - the granularity can be defined as required: the entire line, individual groups or stations over any free period down to ten-minute packages - or collective and detailed views of faults and messages that have occurred.

Figure 4: Reporting history of station 6, both in terms of duration and number

© PIA

After selecting a specific analysis period, it is possible to display all the faults and messages that have occurred according to specific message numbers or in relation to stations in the sequence (Fig. 4). In this example for station 6, message 2385 was examined more closely: "Trumpf TruDisk laser source reports fault". The diagrams show the duration of the message or fault 2385 on the one hand and the frequency with which this message occurred on the other. This distinction is important because although a fault may only occur once, it may cause the system to be down for several hours.
Other faults that occur frequently can perhaps be rectified within seconds by acknowledging or tightening a screw. The above message occurred several times over the course of July, lasting several hours and causing station 6 to shut down. The underlying problem was eventually resolved by further detailed analysis of the specific error code of the laser welding component.