• Tidak ada hasil yang ditemukan

An Example of the Implementation of Transparency

Dalam dokumen How Implementation-Lean Manufacturing (Halaman 186-189)

For example, in a high-speed electrical manufacturing plant with multiple lines operat- ing at 6.65-second cycle times, it was hard to tell, except after several hours of operation, if the lines were producing to the plan. Even once we got a grasp on the production, if there were problems, we didn’t have any idea where the problems might be, so JIT (Just In Time) problem solving was something that wasn’t possible.

In addition, the technicians, line leader, and area supervisors were always busy. They were always correcting something; however, the stories of what was done and why they were done were filled with generalities and terms like “It was not operating quite right.”

These are clear signs of a number of things. Let me point out three:

• Fortunately, the workers, as well as the management, were both motivated and engaged. They were trying to do the right things and they were doing work that resembled their job description.

• Unfortunately, they were doing many wrong things.

• Yet more unfortunately, even when they did the right things, there was inadequate feedback of information to confirm that something positive had been done.

At this time, the relevant information to assess the production rate, which was read- ily available on the floor, included andons, rejected product segregated into collection bins and production boards that covered a full day’s operation.

Theandons had no recording, only warning features. Hence, they were not useful for problem solving unless we were right there when they were activated or cleared.

The scrap data was sorted and recorded hourly, but actual scrap was very low and not a factor in low production.

The production boards had hourly and cumulative production goals with actual production numbers entered at the top of each hour. These goals were calculated based on the hourly goal of 600 units per hour, taking into account lunch and rest breaks.

There were 21.5 available hours per day, so the daily production goal was 12,900 units.

This hourly goal was met nearly 50 percent of the time, but the daily goal was met less than 3 percent of the time. A review of the last month’s data showed production of 9,330 units/day, or 434 units per hour: a full 27 percent below goal.

As part of their management system, there was a daily production meeting on the floor, run by the line supervisor. It lasted about 15 minutes and had a good agenda, but when the production shortage was discussed, which was a topic nearly every day, the answers were—well, amazing is the best word I could use. First, they were very gen- eral, and in almost every case where a specific problem was discussed, it was decreed to be solved. Unfortunately, these same problems would reappear later, and it did not seem odd to anyone that, although these problems had already been solved, they still reappeared. Quite frankly, during this part of the meeting everyone was on “autopilot.”

It was apparent they had accepted these “amazing” explanations for so long that they sort of believed them themselves. Yet day after day the production remained below goal and they were forced into working seven days per week to meet a plan for five days. In this case we can conclude that the production data were inadequate—

there was no “transparency” of the production data. We could neither understand if we were on plan, nor could we solve problems when we did understand them.

After a meeting with the top management, we decided that taking on the low pro- duction was our number one goal. It was the key reason this product was highly unprof- itable. The first thing we did was try to break down the problem into solvable pieces.

We asked three questions. Is the production shortage due to:

• Quality losses

• Availability losses

• Cycle-time losses

We had the quality information, and at this step of the process, quality losses were insignificant. Our segregation bins gave us all the real-time information we needed. We could “see” that quality losses were not the answer to our low production problem. Or in Leanspeak, our transparency regarding quality yield was adequate for this issue.

Now we had to determine if availability was an issue. A quick check showed that material stock outs were virtually nonexistent but the technicians were working on the machines seemingly all the time. Significant downtime occurred, but we neither knew how much of it there was nor did we know what was causing the downtime. We had serious concerns here but had no information at all about the availability losses. Our transparency in this instance was not only inadequate, it was nonexistent, but we did have the andons.

Next, we looked into cycle time. Other than a few time studies done by the engi- neers, no data on cycle time was available. The measured cycle time was advertised to be 6.0 seconds. If the process performed at this cycle time, production should be 600 units per hour—the hourly goal referred to earlier. No measurement of cycle time—of any kind—was done on the floor. The transparency about cycle time was similar to availability information. It was inadequate and practically nonexistent.

With this review, we decided to implement an OEE (Overall Equipment Effective- ness) program. (See the description of OEE in Chap. 4.) OEE information would allow us to begin the understanding of production losses and segregate the information into quality losses, availability losses, and cycle time losses. Forms were made, training was done on gathering and entering data, and the information was set up to manage these data with feedback after each shift.

We soon found out that the losses were about 20 percent on cycle time and almost 10 percent of availability losses. Although OEE is always lagging information, it was valu- able information and told us we needed to work on both cycle time and availability.

However, when it came to understanding and improving the process in real time, we were no closer than a shift away from good information, so we needed to improve that.

Nonetheless, we set up some goals, created an improvement plant, and the first thing we attacked was cycle time. We did a controlled study and found the cycle time was really 6.65 seconds, which surprised everyone. It meant the goal of 600 units per hour was not even attainable. The 6.65-second bottleneck was a manual operation in the welding process. This welding machine was operated by a robot that had a micropro- cessor with a small display screen. Using some great imagination and innovation, the engineering supervisor found a way to program the microprocessor and display the cycle time for the manual operation. The operator now had real-time information regarding cycle time. Immediately, the cycle time began to drop and stabilize, and like- wise production increased. The drop in cycle time was amazing. In less then two weeks, it had improved to 5.5 seconds. We implemented several kaizen activities to improve the work station even further. One of these activities was to program the microprocessor to

display the average cycle time. It told what the average was, how many units were produced with that average, and when the averaging had started. We retrofitted the station with a reset button and—voilà!—we had an excellent piece of transparency.

Ultimately, after a few kaizen events, we were able to achieve cycle times of 4.6 seconds, consistently. Furthermore, now the operator, supervisor, and anyone interested could look at the display and see how the process was performing. Literally, we could find the instantaneous production rate. This was an imaginative solution that added signifi- cantly to the transparency of the system.

However, remember that availability losses were about 10 percent. The OEE infor- mation was valuable in helping us understand the losses, but not very helpful in solv- ing many of the problems due to the lagging nature of the information. Nonetheless, we were able to reduce some losses. Check sheet information logs on white boards were installed on the andons to gather information at the time the andon was activated. This helped distill the information further and made it available for future use. We now had a recording function for the andons. However, the review of the OEE information showed that 90 percent of the availability losses were due to machine adjustments. A quick review showed that most adjustments could not be explained and were just apparently “tinkering” by well-meaning technicians. We did some training, clarified some work procedures, and these availability losses immediately dropped like a rock—

as did the workload of the technicians. They were not comfortable with this at all, how- ever, since they were highly motivated, very energetic and weren’t used to having any spare time.

It might be asked how we ever met the hourly goal of 600 units when the process cycle time would not support it and our average production was only 433 units per hour? Well, the answer is both simple and revealing.

The simple part is this. The way production was measured was to count the produc- tion of trays that each held 120 units and multiply by the number of trays. No partial trays were counted. So, production was in multiples of 120. A tray of 120 units was the trans- fer batch. Also, since the line leaders knew that the managers wanted the production goal to be met, it was better to meet the goal some of the time than none of the time.

Consequently, at the top of the hour when production was counted, if a tray was nearly filled, the line leader might wait to enter the data. Of course, the next hour the produc- tion was even shorter than it would have been if the data had been reported accurately.

This 120-unit transfer batch created some problems beyond accounting for production, so we later cut the transfer batch size to 48 units. This not only helped the accounting but reduced the processing lead time, as you might expect.

So how is this so revealing? It was a symptom of the entire facility. Everyone wanted to do well and be seen doing well. As I said earlier, they were a motivated group. And this motivation served us well as we implemented other process changes and improvements.

We could go on, but let’s briefly review what we just discussed.

• First, our production system improved. The production rate increased by 42 percent with virtually no invested capital.

• Second, once the improvements had been made, we could now “see” the system status, and if there were problems, we could rapidly find and implement the needed countermeasures. We now had the information to do Rapid Response PDCA. Our system transparency had improved immensely, allowing our problem solving to improve immensely, too.

Dalam dokumen How Implementation-Lean Manufacturing (Halaman 186-189)