FMS Insight Monthly Flexibility Analysis

FMS Insight supports continuous improvement by assisting with a monthly review. We suggest that approximately once a month, all stakeholders review the operation of the cell and decide on potential improvements. The improve an FMS documentation goes into more details about how to make the most of these efficiency calculations.

At the top of the page are two radio buttons allowing you to analyze either the last 30 days or a specific calendar month. You can then select from a variety of reports to investigate the cell.

The best metric for continuous improvement of an FMS is not cost/piece but instead is the bottlenecks and utilization of system resources (the goal is to produce more stuff with the same machines and quality). The efficiency tab shows some charts and graphs for a monthly summary of the operation of the cell, displaying reports that in our experience are very helpful to find and fix bottlenecks. We suggest you review this data once a month and use these reports to gradually alter the flexibility or operation of the cell to improve the performance.

Efficiency

Buffer Occupancy

As material moves through the system, it buffers before various operations. It can buffer on a pallet waiting on the inbound rotary table, it can buffer on a pallet in the stocker waiting for either machining or unloading, or it can buffer between processes in a transfer queue. In general, material will buffer right in front of the bottleneck operation. The buffer occupancy chart can thus be used to determine which operation (machining, loading, unloading) is the current bottleneck and also how the bottleneck changes over time.

The buffer occupancy chart calculates a moving average of the quantity of material buffered in all these various places. The x-axis is the days of the month. For each time on the x-axis, a window around that point is used to calculate the average quantity of material in the buffer during the window of time. This average quantity is graphed on the y-axis.

Using a moving average window smooths out the jitter from individual pallet moves, for example when a pallet rotates into the machine and another pallet is quickly sent to the inbound rotary. The size of the window can be controlled by the slider in the top-right. The size of the window should be set so that the major trends are visible while short oscillations are smoothed out.

In an efficient, well-running system the bottleneck is always machining. This will be reflected in the buffer occupancy chart with the "Stocker[Waiting for unload]" line zero or almost zero, the "Rotary" for each machine always above zero, the "Stocker[Waiting for machining]" only positive if all rotary tables are full, and any in-process transfer queues small. If instead the load stations become the bottleneck, the buffer occupancy chart will show a rise either "Stocker[Waiting for unload]" or a rise in the in-process transfer queue between processes. Also, the rotary buffer occupancy will drop to zero.

Station Use

The Station Use heatmap shows the station usage over the month. On the x-axis are the days of the month and on the y-axis are the machines and load stations. There are three charts which can be selected in the top-right corner: "Standard OEE", "Planned OEE", and "Occupied".

• The "Standard OEE" chart computes the actual overall equipment effectiveness (OEE) over the month. For each station and each day, FMS Insight uses the log of parts produced and adds up the expected operation time for each part cycle and then divides by 24 hours to obtain a percentage that the station was busy with productive work. (If station cycles were longer than expected, this extra time is not counted in the OEE.) For each grid cell in the chart, the OEE percentage is drawn with a color with darker colors higher OEE and lighter colors lower OEE. A grid cell can be moused over to obtain extra information in a tooltip.

• The "Planned OEE" chart displays the simulated station usage for the downloaded jobs and is the prediction of what the OEE should look like based on all the schedules for the month.

• The "Occupied" chart computes the total percentage of time that a pallet is occupying the station. For each station and each day, FMS Insight uses the log of events to determine when a pallet arrives and departs from each station. The total time a pallet is at the station is then divided by 24 hours to obtain a percentage that the station was occupied. These percentages are then charted with darker colors higher occupancy.

The Station Use heatmaps are useful for several observations. First, the Occupied heatmap can be used to determine the overall usage of the load station. If the load station occupied percentages are very high, it can indicate that the load station is a bottleneck and preventing the cell from performing useful work. In a healthy cell, the load stations should be faster than the machines, so the load stations should be empty at least part of the time waiting for machining to finish.

The difference between the Standard OEE and Occupied heatmaps can be useful to determine if there is a lot of cycle interruptions. For example, if a program is expected to take 25 minutes but spends 40 minutes at the machine, the Standard OEE heatmap will be credited with 25 minutes and the Occupied heatmap will be credited with 40 minutes, causing the Occupied heatmap to be darker in color for that day. Now, cycle interruptions do occasionally happen; comparing the Standard OEE and Occupied heatmaps allows you to determine if these are one-off events or a consistent problem. If most days are darker on the Occupied heatmap, the cycle interruptions should be investigated in more detail. (For example, look at individual parts on the machine cycle charts, ensure the expected time entered into the schedule is actually correct, investigate machine maintenance records, etc.)

Finally, the Standard OEE heatmap helps visualize how balanced the machines were loaded over the month. We want to see all the machines consistently roughly the same color. If you see that a machine has a lighter color for a couple days, that indicates either the machine was down or that the daily mix for that day did not have enough flexibility. You should then consider picking a part and extending that part to run on the lightly loaded machine. To find such a part, you can use the part production chart below to see which part mix was run on this day to help find a part that might be changed to run on the lightly loaded machine.

Part Production

The Part Production heatmap shows the distribution of completed parts over the month. On the x-axis are the days of the month and on the y-axis are the part types. For each part and each day, FMS Insight counts how many parts were produced that day. For each grid cell in the chart, the entry is drawn as a color with darker colors higher machine hours and lighter colors lower machine hours. A grid cell can be moused over to obtain extra information in a tooltip.

The part production OEE heatmap is mainly useful to visualize the part mix as it varies throughout the month, by comparing the relative color shades. Also, it can help find a part to change move onto a lightly loaded machine. For example, consider that a machine is found to be lightly loaded via the station OEE heatmap. That same day can be viewed on the part production OEE heatmap and the darkest colored part was the highest run that day and could be considered to be extended to be run on the lightly loaded machine.

Note that these heatmaps should only be used to brainstorm ideas. We would still to investigate if expanding yyy to include machine 2 would increase overall system performance. Are there enough pallets? How many extra inspections are required? Will this cause a traffic jam? These questions can be answered using simulation, SeedTactic: Designer, Little's Law, or a tool such as our SeedTactic: Planning.

Cycle Analysis

If the broad analysis above indicates a problem, the cycle analysis can be used to investigate the problem in more detail.

Machine Cycles

The machine cycle chart displays a point for each program cycle. The cycles can be filtered by a specific part, a specific program, a specific machine, a specific pallet or a combination. The x-axis is the days of the month and the y-axis is cycle time in minutes. The cycle time is the wall clock time between program start and program end of the machine. The legend at the bottom shows which colors correspond to which stations, and by clicking on stations in the legend you can enable or disable the viewing of specific stations. By clicking on a point you can obtain details about that specific cycle in a tooltip and open the material card for the cycle. Finally, the chart can be zoomed by clicking and dragging.

When a specific part and specific program are selected, a gray background and black horizontal line are shown. The gray background is calculated to be the statistical median and deviation of the points; the middle of the gray band is the median of the program time and the width of the band is determined by calculating the median absolute deviation of the median (MAD). In contrast, the horizontal black line is the expected cycle time entered during scheduling and sent in the scheduled jobs. Thus if the horizontal black line differs significantly from the middle of the gray band, this means that likely the expected time used in the simulation and scheduling is incorrect.

The example screenshot above is an example of a part program which looks good. We see that almost all of the machine cycles are around 40 minutes, within the gray band and the horizontal black line and there are only a couple machine cycles significantly longer than 40 minutes which likely come from program interruptions, tool problems, etc. Since there are only a couple outlier points, we might instead conclude that it is not worth the effort to spend a large amount of time focusing on this part. If instead the graph showed many points outside the gray band and longer than 40 minutes, it would be worth it to investigate and improve the part program and tooling.

The cycles can be toggled to display the raw data in a table instead of a chart. The table can be filtered, sorted, and restricted to a specific date range. The resulting raw data can be copied to the clipboard to be pasted into a spreadsheet for further analysis.

Load/Unload Cycles

The load/unload cycle chart displays a point for each load or unload event. The x-axis is the days of the month and the y-axis is time in minutes. The cycles can be filtered by a specific part or pallet. The chart will display one of three possible times for each operation: "L/U Occupancy", "Load (estimated)", and "Unload (estimated)".

The "L/U Occupancy" chart displays a point for the wall clock time that the pallet spends at the load station; the time from the pallet arriving at the load station until the operator presses the ready button. This is the time that FMS Insight collects and stores so this chart displays the actual raw data collected.

Despite being the actual data FMS Insight collects, the "L/U Occupancy" chart is hard to use to determine if there are problems or slowdowns in loading a specific part. The recorded time is the wall clock time from pallet arrival to departure so includes all the operations which occur at the load station which might include a load and an unload or just a load or just an unload. Thus, FMS Insight attempts to split the wall clock occupied time of the pallet at the load station among the various operations such as loading or unloading that occurred.

For each cycle, FMS Insight splits the time the pallet spends at the load station among each piece of material being loaded, transferred, or unload. To do so, FMS Insight uses the list of material loaded/unloaded/transferred and calculates the expected time the operation should take based on the values entered into the scheduled jobs. The expected time is then used to calculate a percentage consumption for each material, and this percentage is then applied to the actual wall clock time of the pallet occupancy. For example, consider an operation where part aaa is loaded and part bbb is unloaded. The schedule says that the loading of aaa should take 5 minutes and the unloading of bbb should take 8 minutes for a total time the pallet should spend at the load station of 13 minutes. Now say that the pallet actually spent 15 minutes at the load station. The load of aaa is expected to consume 5/13 = 38% of the cycle and the unload of bbb is expected to consume 8/13 = 62% of the cycle. The 38% is multiplied by the actual wall clock time of 15 minutes to give approximately 5.8 minutes for the load of aaa. Similarly for the unload of bbb, 15 minutes times 62% is approximately 9.2 minutes. Note how the "extra" 2 minutes (15 minutes compared to 13 minutes) for the entire cycle is divided among both the load and the unload operation. This calculation is repeated separately for each load/unload cycle.

The result of FMS Insight's estimated splitting is graphed in the "Load Operation (estimated)" and "Unload Operation (estimated)" charts selected in the top-right corner. The screenshot above shows the "Load Operation (estimated)" chart. For each load/unload cycle, FMS Insight splits the time as described above and then plots a point with the y-coordinate the calculated time, filtering the data to a specific part and/or pallet. (In the example from the previous paragraph, the points would have y-coordinate 5.8 and 9.2 minutes.) Exactly like the machine cycles, FMS Insight calculates the median and median absolute deviation of the points and plots them as a gray band in the background. Finally, the expected time entered into the scheduled jobs is graphed as a horizontal black line.

Pallet Cycles

Select a pallet from the combo box in the top-right. Once selected, all pallet cycles are displayed. The x-axis is the days of the month and the y-axis is the pallet cycle time in minutes. A pallet cycle time is the wall clock time from when a pallet leaves the load station to the next time it leaves the load station. This will therefore include cart transfer time, buffer time, machining time, and load/unload time. By clicking on a point, you can obtain more information about the pallet cycle in a tooltip. Similarly to the station cycle chart, the pallet cycle chart can be zoomed by clicking and dragging or the zoom range can be manually set via the button in the bottom-right.

The pallet cycle chart is best used to determine if the cell is running lean and validate the daily allocation and scheduling technique to determine if there are traffic jams occurring in the pallets. In a lean, healthy cell, most pallet cycles should be low and reasonably consistent. If pallet cycle times vary wildly, there is likely traffic jams or other flow problems.

Cell

Quality

The quality shows all paths taken by parts through the system for each inspection. The inspections can be viewed either as a sankey diagram or as a table.

Tool Replacements

The tool replacement chart shows data about tool replacements over the month. Each tool is shown with the number of replacements, average usage at replacement, and average count at replacement. These numbers are approximate if a tool replacement happens during a cycle, since we do not know the exact time of the replacement. Each tool also shows a graph of all replacements over the whole month, with the red line the average and a black dot for each replacement. A replacement can be hovered for more information.

Schedules

The schedules report shows all the scheduled jobs for the month, details about the schedule, and a list of all serials machined on the schedule.

Cost/Piece

Cost-per-piece is a great metric for scheduling, operations management, future capital purchases, quoting work, accounting, and budgeting. The cost/piece tab is largely intended to serve as a verification that flexibility and operational changes are reducing part costs; you can compare part costs from before a change to after a change to understand if the change improved the system. Indeed, FMS Insight has a narrow view of just the cell itself and does not take into account any overhead or other costs. Therefore, for budgeting, quoting work, and other management decisions, we suggest that you export the cost breakdown or workorder data to your ERP and implement a cost report in your ERP taking into account all factors.

Cost/piece is not a great metric to use initially when searching for techniques to improve the cell's operation, since focusing only on cost/piece risks introducing quality problems and OEE reduction. Instead, cost/piece is a great metric to use after-the-fact to determine if an implemented change in production operations has had a meaningful impact on cost/piece.

Percentages

Calculating cost-per-piece requires splitting large fixed costs such as machine depreciation and labor across the parts that were produced. To do so, we use a monthly analysis window of the actual operations together with the planned use of resources to determine a percentage breakdown of the cost for each part.

To do so, consider that the system produces two part types, aaa and bbb, and aaa has a planned cycle time of 3 hours and bbb has a planned cycle time of 2 hours. For January we collect data on the total number of parts produced by the system. Consider that in January the system produced 400 aaa parts and 500 bbb parts. We then calculate the cost breakdown as follows. Since we produced 400 aaa parts, those aaa parts should have used 400*3 = 1200 machine-hours. Similarly, the bbb parts should have used 500*2 = 1000 hours. Thus, aaa used 1200 / (1000 + 1200) = 54.5% of the planned hours and bbb used 1000 / (1000 + 1200) = 45.5% of the planned hours. Similar calculations happen for labor and automation.

From a cost perspective, any bottlenecks or utilization slowdowns should be viewed as system problems and all parts are responsible for a portion of this cost. Indeed, The cost-per-piece metric is an actionable insight for quoting orders and justifying future capital investments and for these purposes any OEE problems are a problem for everything produced by the system. The above method does this by using planned cycle times to divide the total machine cost (which includes active and idle time) among the parts produced during the month based on their weights. A quick calculation of machine utilization gives (1200 + 1000) / (24*30*4) = 76% use. The above method divides the 24% of the time the machine is not in use among aaa and bbb based on their percentages of planned use. Attempting to identify OEE problems are better addressed using the efficiency reports above.

The data can be copied to the clipboard via the two-arrow-icon button in the top right.

Part Cost/Piece

Once the percentage breakdown for each part is calculated, the webpage allows you to enter the total machining, labor, and automation cost for the period. This value is then multiplied by the above cost breakdown percentages to get a total cost per piece. As mentioned above, we suggest that you copy the above cost breakdown percentages into your own spreadsheet or ERP system to be able to account for overhead, overtime, material costs, and other costs not visible to FMS Insight.