Professional Documents
Culture Documents
Volume 3 Issue 1
ABSTRACT
Milkrun denotes a material conveyance and replenishment method to deliver and retrieve
materials simultaneously. The method has gained popularity in the manufacturing industries
to improve logistic efficiency and facilitates lean transformation by simulating pull system,
reducing inventory and streamlining processes to eliminate waste. This paper introduces a
multi-step milkrun system development and implementation framework. Aligned with lean
philosophies, several components were extended through the framework, such as leader
standard work and continuous improvement. A case study, based on the final assembly of an
electronics assembly company is adopted to illustrate the execution of the implementation
framework. The approach successfully generates a significantly complete, stable and
integrated milkrun. As a practical implication, the work provides guideline to practitioners to
systematically develop and implement milkrun as primary material handling system in
production.
Fig.1:-Milkrun process
interaction between multiple actors [9]. gives attention to key findings of the
This brings sociotechnical considerations implementation, including lean tools
to the foreground of any milkrun helpful to establish an effective milkrun.
development which consciously underlines
reciprocal interrelationships between MULTISTEP MILKRUN
people and technology to minimize DEVELOPMENT AND
conflicts between efficiency and humanity DEPLOYMENT
[23] to constantly align customer Boysen, Emde, Hoeck, and Kauderer [5]
requirement and remove all operational commented warehouse decisions typically
wastes. Besides, human operators who be need to be made quickly without extensive
the ones required to understand, support computational resources. Further, the
and follow the rules and standards in most management generally demands a solution
production settings [18]. The interrelating method to be simple, intuitive, and reliable
lean concepts and tools affect sustainable to minimize the training costs. Design
milkrun development and deployment. considerations in material handling include
Toyota built up a lean system of respect material characteristics, flow rate, routing
for workers, emphasizing waste and scheduling and plant layout [11]. In
elimination, safety and self-displaying of these spirits, five basic steps were
capabilities by entrusting them with conceived to facilitate the development
greater responsibility and authority [17]. and deployment of a functional milkrun
Five fundamental principles of lean with support from warehouse, as shown in
thinking are specify value, identify the Figure 2. A common production goal is to
value stream, continuous process flow, eliminate inventory on production floor.
pull system and pursuit of perfection The framework only considers in-plant
through continuous improvement [29]. milkrun as logistic milkrun would require
Lean practices work in synchronization to deliberating other factors such as logistics,
create a streamlined, defect-free system to cost, materials with multiple suppliers etc.
produce finished products apace with To integrate both types of milkrun, the
customer demand and generation of resultant management complexity
minimum waste [15]. The quantifiable outweighs the benefits in most cases. In
benefits are significant and potentially addition, Mello, Hellingrath, and Martins
motivating industries to implement lean in (2019) opinoned in-plant milkrun would
their processes [8]. Lean compels complete inherent a higher pitch of delivery
change to business [21]. Integrating lean compared to logistic milkrun, therefore
into all aspects of supply chain maximizes decoupling both systems at warehouse
benefits within the organization [21]. would be the most practical option.
production floor space utilization. One For example, a high running assembly line
way to meet this end is to match inventory averagely consumes 85.6 parts per hour
level to a smaller delivery pitch (illustrated with standard deviation 1.8. Average
in the next section). delivery lead time is 1.03 hours with
standard deviation of 0.06 hours. A 90%
The desired inventory level, Idesired depends confidence level requires seven parts of
on demand level, demand fluctuation, safety stock, or 8% of the hourly
delivery promptness and delivery lead consumption (isafety = 0.08). The level of
time. Demand fluctuation can be best safety stock estimates the level of
estimated through historical data. Delivery inventory required to buffer against
lead time is the time required to deliver variations. Continuous improvement to
and replenish parts soon an order is placed. lower safety stock by stabilizing the
Related operations include material milkrun is vital. For instance, overstocking
picking, material feeding and delivery can be avoided by making loading time
truck traveling. Delivery should be made reliable [30].
just-in-time, as any delivery inconsistency
should be investigated and minimized. Step 2- Compute Delivery Pitch
Delivery pitch is the pace of delivery or
As any stockout is undesirable, delivery time interval between two deliveries.
lead time must be shorter than time to fully Delivery pitch fundamentally affects all
deplete inventory in production as aspects of milkrun such as inventory level,
preventive. A discreet amount of safety efficiency and utilization of the material
stock in the inventory cushions against handling process. The pitch can be
variability and fluctuation in demand; and determined by setting a desired level of
avoids overabundance of inventory [6] inventory, and vice versa. It is generally
determined safety stock by Equation (2); easier to set the pitch to control inventory
√( ) ( ) level. After the pitch is calculated, find the
(2) nearer time (time multiplier) that is easy
and simple for calculation and
Where, remembering, such as 6 hours, 4 hours, 2
SS = Safety Stock, Z = Service factor, µT = hours, 1 hour, 30 min and etc. The formula
Average delivery lead time, µd = Average to calculate the delivery pitch as per
demand per unit time, δT = Standard Equation (4)
deviation of delivery lead time, δd = ( )
(4)
Standard deviation of demand per unit
time. Using Z scores, proportions of the
Where,
demand fluctuation (in normal
Tpitch = delivery pitch (hr). Inventory level
distribution) representing the confidence
can be reduced by progressively reducing
level of safety stock to obviate stockout
delivery pitch.
can be decided. The demand fluctuation
can be obtained from a historical record of
Step 3- Set Up and Coordinate Process
hourly demand.
Four basic elements to milkrun process
The safety factor, isafety, can then be
are;
calculated in term of percentage using SS
1. Replenishment signal (Kanban Card)
as a fraction of consumption rate, Q per
2. Delivery
unit time, as the Equation (3);
3. Feeding
(3) 4. Picking
Today, Kanban cards and barcode replace cards for use in subsequent replenishment.
standard containers as alternative to The amount of material to be delivered
signals for replenishment. Kanban cards depends on the number of withdrawal
ease control over buffer and inventory Kanban received from the previous
level. According to [2] supplied container milkrun session. A dispatching schedule
with a Kanban card identified few must be defined and made known to all
information indicates the storage location drivers on duty. The delivery route should
of the material in the production. The be planned, designed and defined to
flexibility of using Kanban cards also deliver parts efficiently.
encourages continuous improvement. The
number of Kanban cards in circulation can The feeding section receives materials
be expanded if part consumption rate from the delivery section and loads them at
increases; or can be reduced if delivery the inventory holding area. Usually done
pitch is shortened. by manual labor called feeders, it is
The following formula (Equation (5)) can important that work needs to be performed
be used to calculate number of Kanban within the delivery pitch to prevent
Cards in circulation for each part: subsequent delay in milkrun and possible
( ) line down issues.
(5)
The formula takes into consideration that Picking needs to be attuned to delivery
(1+isafety) sets of inventory remain in the pitch, not vice-versa. The process is
production floor while another set of usually performed by manual labor called
consumed inventory will be returned to pickers, with the aid of machines such as
warehouse for replenishment. Regardless forklifts and automated storage/retrieval
of standard pack size, a minimum of two systems. As order picking is generally
cards are required. Ideally, one Kanban recognized as the most expensive
card is to be attached to one part. The warehouse operation, various picking
smaller the standard pack size, the easier is strategies could improve picking efficiency
to control inventory and this leads to [12]. For example, departments are further
higher inventory accuracy. organized into pick zones. Batch picking
and wave picking naturally complements
For example, a milkrun process running at milkrun. Batch picking or known as multi-
1-hour pitch, with a safety factor 10%. The order picking combines several orders into
consumption rate is two parts per hour. one pick instruction. As many customer
With a standard pack of one part per pack, orders need to be picked in a specific
a total of 4.2, rounded up to five Kanban period, batch picking is a mutual way to
cards would be required. At any given systemize the picking process [26]. Wave
point of time, three Kanban cards would picking assigns orders into grouping
remain on production floor and two (waves) which will be released in specific
Kanban cards would be circulated back to interval. The picking section receives
warehouse for picking process. empty containers and Kanban cards.
Materials and parts are picked accordingly
The delivery section connects component and loaded onto the delivery train and
between material source (warehouse) and ready for the next delivery. Data entry may
recipient (production floor). Picked be required to simplify the picking process
materials are loaded onto the delivery train by generating easy-to-access picking list.
and delivered to the production floor at The picking list should include part
specific pitch. On the return trip, the train description as well as actual location for
collects empty containers and Kanban the pickers to locate the parts. With each
Fig.3:-Pareto Chart
Tow tugger
Yellow
cart
Loaded carts were connected one another separate and alternating routes were
to resemble a train (Figure 4), prior to developed to deliver material to each cell
departure. The whole installation is at the specified pitch: Pengasus Express
referred as tow train hereafter. Two and Unicorn Express (Figure 5).
Pengasus Express leaves the departure Kanban cards before feeding them to the
every hour, starting from 7:00am. As it rack from outside and around the cells.
leaves the departure area, it travels straight Each cell is configured with customized
to Station 1. At the station, tugger driver material flow rack to enable First-In-First-
detaches the cart designated for a cell and Out (FIFO). Rack inlets were labelled
collect Kanban cards for parts consumed in according to part numbers and only
the past hour. Later, the tow train moves to matching parts can go into the rack. Each
Station 3 and repeats the process. Before rack was designed to hold exact amount of
returning to Warehouse, it makes a final quantities stipulated in Kanban cards to
stop at Station 5 to collect carts emptied in avoid excess inventories. Efficient feeding
previous delivery run. route and sequence were studied and
established.
Unicorn Express takes the similar route for
delivery every hour, starting from 7:30 am. Picking section
It stops at Station 2 and 4 for delivery Figure 7 illustrates the workflow of each
process, and Station 5 for empty cart section where the detached Kanban cards
collection. Figure 4 outlines both delivery are collected and sent back to Warehouse.
routes. Stop points for tow train (termed as At Warehouse, the detached Kanban cards
Bus stops) with sign boards (Figure 6) are scanned using barcode scanner and the
were placed at each station with detailed information enters the SAP system to
instruction and arrival timetable. These bus generate a picking list. A picking list
stops serve as visual tools to facilitate contains information such as part number,
process monitoring and accountability. quantity, pack size, bin location etc. Picker
Figure 6: Bus stop picks parts, loads them onto the carts and
attaches them with the corresponding
Feeding section Kanban cards. Finally, the picker brings
A feeder receives parts from tow train on a the carts to the departure area ready for the
cart and returns detached Kanban card to next replenishing trip. Picker also scans
the tugger driver. Feeder later verifies the attached Kanban cards to close the
these parts are attached with the correct picking order.
Step 4 - Establish Standard and Training Step 5-Stabilize, Monitor and Improve
Several training sessions were conducted A provision of three weeks aims to
for managers, supervisors, Warehouse and stabilize the milkrun process. Leader
production operators to improve their standard works support the monitoring and
understanding on milkrun. The technical accountability process. In border terms,
training is considered to reduce time for audits will be conduct on Kanban cards,
implementation [28] because most of them milkrun schedule, departure punctuality,
have to be aware of their job inventory accuracy, daily core team
responsibilities and roles in milkrun. The meeting etc. Leader standard works were
training introduced the concept of milkrun,
displayed on a review board and visible to
milkrun rules and prerequisites. As a part
everyone. Pockets are reserved to keep
of interactive training, a Lego simulation
completed audit forms. The process status
of the milkrun was included. The first
round of the simulation focused on the was displayed and updated on the review
shortcomings of a production system board (Figure 8). These forms were
without milkrun system. The subsequent collected and documented by process
rounds introduced and resulted to an owners at the end of work. All metrics of
improved milkrun system. The trainees the milkrun process such as milkrun status,
were later trained on their assigned roles. inventory accuracy, departure accuracy,
Work instruction included task downtime etc were reviewed daily. Green
prioritization, safety and precautions, magnetic button was used to indicate a
technical handling of equipment, reporting, normal condition and red to indicate an
housekeeping and maintenance. For abnormal condition. Further actions were
example, tugger drivers must be written in the ―Remarks‖ section and being
experienced in battery replacement and removed upon completion of the actions.
charging.
Kaizen for continuous improvement minutes later and 19 minutes earlier than
After studying the collected results, the scheduled time, with a standard
several improvements (kaizen activities) deviation of 8.95 min. As implications,
were undertaken to fine-tune the milkrun. late departure led to insufficient parts and
A case to stabilize milkrun departure is material at cells, starving these processes
presented here. Within three weeks of upon exhausting the safety stock. Special
monitoring, production line was down deliveries had to be called, which were in
three times due to insufficient material. A essence operational waste and cost to
time study was conducted on the tow train Warehouse. Meanwhile, early delivery
departure punctuality, traveling time of caused WIP accumulation at cells and
tow train between stops and time for the affected Kanban cards circulation. In
tow train to complete a milkrun. Figure 9 additions, this deprived time to unload all
shows that tow trains departed from the carts and consequentially inadequate carts
warehouse with time variations within 16 collection to perform the next delivery.
As observed, tugger drivers were unaware The first signal alerts the tugger driver and
of the delivery time while attending to pickers. The tugger driver on-duty should
other tasks. To sustain a functional drop all tasks at hand and proceed to drive
milkrun process, the tow train must depart the tow tractor to departure area. With ten
for material delivery no later or earlier minutes to spare, tugger driver retrieves
than one minute from the scheduled time. parts at designated location and loads them
An alarm system could improve the onto carts. Meanwhile, pickers complete
delivery consistency and help all operators their picking process and attach the cart
in the milkrun system to work smoothly. onto the tow train. The ten-second
The alarm system generates signals at countdown alerts the tugger driver ready
specific interval audible to the entire for departure. All preparation for departure
warehouse. It also presents visual must be completed and tugger driver
information such as departure time, current stands by on the tow train. The countdown
task and delivery destinations. The alarm also alerts the team leaders to monitor the
system comprised of a complete set of process. One minute after the departure
computer, Windows 7 Operating System signal, the information of the next
and a speaker. The system was mounted departure will be displayed.
eight feet above ground at the tow train
departure point as shown in Figure 10. The Management used Single-Point Tutorial
alarm system integrated Windows Task (SPT) for quick briefing and training on
Scheduler, Windows Media Player and the alarm system. SPT includes easy-to-
Microsoft PowerPoint for the following understand WI, visual instructions and
tasks: warning to be read by operators. Operators
1. Trigger an audio signal ten minutes signed the SPT acknowledgement form to
prior to departure. indicate their awareness to the changes.
2. Trigger an audible and visual ten After the implementation, the tow trains
seconds countdown to departure. departed significantly more punctually
3. Display the next departure information (Figure 11). The milkrun is significantly
for numbered lists stabilized, more manageable and
predictable.
Lastly, shop floor conditions before and after the deployment of milkrun are compared and
tabulated in Table 2.