Lecture 4: BOTTLENECK IDENTIFICATION AND ELIMINATION
Semyon M. Meerkov, University of Michigan J ingshan Li, University of Wisconsin Madison Liang Zhang, University of Wisconsin Milwaukee
Copyright J . Li, S.M. Meerkov and L. Zhang 2012 Outline 4.1. Introduction 4.2. What is the bottleneck machine? 4.3. What is the bottleneck buffer? 4.4. Identification of BN-m and BN-b in serial lines 4.5. Identification of BN-m and BN-b in assembly systems 4.6. Potency of buffering 4.7. Designing continuous improvement projects 4.8. Measurement-based management 4.9. Case studies 4.10. Summary 4.11. Lab: PSE Toolbox function for BN-m and BN-b identification 4-2 Due to breakdowns and other perturbations, it is not uncommon that machining lines operate at 60%-70% of their capacity. In assembly systems, these numbers are 80%-90%. Therefore, continuous improvement is of central importance. 4.1. Introduction 4-3 In practice, continuous improvement projects (i.e., improving machines and buffers or purchasing new ones) are often designed without a rigorous justification. As a result, many continuous improvement projects do not measure up to expectations. The goal of this lecture is to present quantitative engineering methods for designing continuous improvement projects with rigorously predicted results. The approach developed here is based on identification and elimination of bottleneck machines and bottleneck buffers. 4-4 4.2. What is the Bottleneck Machine? Often, the worst machine is isolation is viewed as the bottleneck machine. This understanding is wrong because it is local in nature and does not look at the system as a whole. We define the bottleneck as the machine that affects the overall system performance in the strongest manner. To quantify this understanding, we introduce the following definition: 4-5 Definition: m i , i {1,, M}, is the bottleneck machine (BN-m) in a Bernoulli line if
m i , i {1,, M}, is the c-bottleneck (c-BN) in a line with continuous time models of machine reliability if
for all ; i j PR PR j i p p
>
for all . i j TP TP j i c c
>
i i j j 4-6 4-7 (c) Machine with the smallest capacity is not the c-bottleneck 4-8 cannot be measured on the factory floor
during normal system operation. Also, they cannot be evaluated analytically. So, how the BN-m can be identified?
or i i PR TP p c
4-9 4.3. What is the Bottleneck Buffer? Definition: b i , i {1,, M 1}, is the bottleneck buffer (BN-b) if
The smallest buffer is not necessarily BN-b:
How the BN-b can be identified?
4-10 4.4. Identification of BN-m and BN-b in Serial Lines It turns out that BN-m (or c-BN) and BN-b can be identified using blockages and starvations of the machines. Specifically, consider a production line and determine (either by measurements on the factory floor or by calculations) the probabilities (or frequencies) of blockage and starvation of each machine. Place BL i and ST i under each machine as follows: ST i 0 0.01 0.39 0.37 0.27
BL i 0.41 0.20 0.27 0.01 0 4-11 Assign arrows from one machine to another according to the following rule: If BL i >ST i+1 , assign the arrow pointing from m i to m i+1 . If BL i <ST i+1 , assign the arrow pointing from m i+1 to m i .
ST i 0 0.01 0.39 0.37 0.27
BL i 0.41 0.20 0.27 0.01 0 4-12 Bottleneck Indicator: The machine with no emanating arrows is the BN-m (or c-BN). If there are multiple machines with no emanating arrows, the one with the largest severity is the Primary BN-m (Pc-BN), where the severity of each local bottleneck is defined as follows:
BN-b is one of the buffer surrounding the BN-m (or c-BN); it is before BN-m if ST i >BL i ; it is after BN-m if ST i <BL i . 4-13 4-14 4-15 4-16 4-17 4-18 4.5. Identification of BN-m and BN-b in Assembly Systems The definitions for BN-m and BN-b remain the same. The arrow assignment rule also remains the same, with the only difference that the merge operation may be starved by multiple component lines. Given this arrow assignment, the Bottleneck Indicator remains the same as in serial lines. 4-19 4-20 4-21 4-22 4-23 4-24 4-25 4.6. Potency of Buffering Definition: The buffering of a production system is: weakly potent if the BN-m is the worst machine in the system (i.e., the machine with the smallest throughput in isolation); otherwise, it is not potent; potent if it is weakly potent and its production rate is sufficiently close to the BN-m efficiency (e.g., within 5% of the BN machine efficiency); strongly potent if it is potent and the total buffer capacity is the smallest possible to ensure the desired throughput).
4-26 To determine if the buffering is weakly potent, the methods introduced in this lecture may be used. To determine if it is potent, the methods introduced in Lecture 3 may be used. To determine if it is strongly potent, the methods introduced in Lecture 5 may be used.
4-27 Along with its practical utility, the notion of buffering potency is important conceptually. Indeed, typically, production supervisors concentrate their attention on the machines their reliability, efficiency, capacity, etc. Attention to buffering is often missing, although the buffers are also important for system operation. The notion of buffering potency provides a language necessary to focus attention on the buffers.
4-28 4-29 4.7. Designing Continuous Improvement Projects 4.8. Measurement-Based Management The process of continuous improvement requires a mathematical model of the system at hand. This may be difficult to obtain (especially for large systems). Therefore, a simpler method, referred to as MBM, is proposed. It consists of the steps described next.
4-30 4-31 Examples of the first step can be given as follows:
4-32 The second step can be carried out based on measuring the blockages and starvations and using the expressions:
Thus, to carry out MBM, the manager must receive daily, or weekly information of the time of blockages and starvation of various units. 4-33 The third step is carried using the arrow-based method of BN identification:
4-34
The last step is up to the manager and his/her staff to determine which actions should be taken to eliminate the BN. 4-35 4.9. Case Studies 4.9.1. Automotive ignition coil processing system Bottleneck identification:
Improving m9-10 by 10% and b 9-10 by 1 leads to TP =505 parts/hour 4-36 Bottlenecks of the improved system:
Increasing b 5 by 1 leads to TP =511 parts/hour an acceptable performance.
4-37 Bottleneck identification:
The main reason for m 3 to be the bottleneck is starvation by empty carriers. Assuming the empty carriers are always available, we obtain 4.9.2. Automotive paint shop production system Bottlenecks of the improved system:
Increasing efficiency of m 3 by 4% leads to
and machine m 4 becomes the new bottleneck. 4-39 4.9.3. Automotive ignition module assembly system Conclusion: MHS is not potent. Increasing capacity of all buffers: Increasing capacity of buffer conveyor: over 9% TP improvement. Eliminating starvations of Op.1 and Op.9 and blockage of Op.18:
Last two actions have been implemented. 4.10. Summary In the same manner a medical doctor cannot treat patients without taking their vital signs, production systems cannot be managed without appropriate measurements. This lecture shows that the most important vital signs of a production system are blockages and starvations. Based on this information, managers can exercise MBM a rigorous way for achieving good performance of production systems.
4-43 4.11. Lab: PSE Toolbox Function for BN-m and BN-b Identification 4-44 BN-m and BN-b in serial lines with Bernoulli machines Input: M, number of machines p, reliability of each machine N, capacity of each buffer Output: BN-m and BN-b Production rate (PR) Work-in-process (WIP) Probability of starvation (ST) Probability of blockage (BL) 4-45 4-46 BN-m and BN-b in serial lines with exponential machines Input: M, number of machines , failure rate of each machine , repair rate of each machine c, speed of each machine N, capacity of each in-process buffer Output: Throughput (TP) Work-in-process (WIP) Probabilities of starvation (ST) and blockage (BL) Machine efficiency (e) BN-m and BN-b. 4-47 4-48 BN-m and BN-b in serial lines with general model of machine reliability
Input: M, number of machines T up , average uptime of each machine T down , average downtime of each machine
Output: Probabilities of starvation (ST) and blockage (BL) BN-m and BN-b.
4-49 4-50 BN in assembly systems with Bernoulli machines Input: M 0 , M 1 , M 2 , number of machines in assembly line, component line 1 and component line 2, respectively p 0 , p 1 , p 2 , Bernoulli reliability of each machine N 0 , N 1 , N 2 , capacity of each buffer Output: Production rate (PR) Probability of starvation (ST) Probability of blockage (BL) 4-51 4-52