Latest / Mechanical Engineering Made Simple / (#28) How Lean Manufacturing Cut Steel Mill Lead Time by 65% The Hybrid Pull System Revolution

Transcript

(#28) How Lean Manufacturing Cut Steel Mill Lead Time by 65% The Hybrid Pull System Revolution · Mechanical Engineering Made Simple

• 0:00
  Get ready to torque your nuts, harden your springs, and forge
• 0:03
  your mind into a bulletproof study guide.
• 0:10
  We are Mechanical engineering made simple.
• 0:13
  Welcome back design junkies to mechanical engineering made
• 0:15
  simple. So today we're doing a bit of an
• 0:17
  industrial demolition job. You could say we're tackling
• 0:21
  this well, this long held belief that lean manufacturing, the
• 0:26
  Toyota Production System magic only really works for nice clean
• 0:30
  discrete assembly lines. Yeah, like making cars or
• 0:33
  appliances. Things you can count easily.
• 0:35
  Exactly. But our mission today, we're
• 0:37
  taking those core lean ideas and throwing them right into the
• 0:40
  deep end. We're talking huge scale,
• 0:42
  continuous process industries, specifically steel
• 0:45
  manufacturing. Think massive furnaces, liquid
• 0:48
  metal stuff. That's fundamentally different
• 0:50
  from building an engine piece by piece.
• 0:52
  We're talking about taking muta waste elimination into
• 0:54
  environments where machines are gigantic, run nonstop, and you
• 0:57
  definitely can't just rearrange them over the weekend.
• 1:00
  And it's interesting, isn't it? Lean came from scarcity post WWI
• 1:03
  Japan. They had shortages of pretty
• 1:05
  much everything, capital, space, materials.
• 1:07
  Right. So the whole philosophy was
• 1:08
  built around using less. Less time, less inventory, less
• 1:12
  effort, but getting the same or even better output, higher
• 1:16
  quality, more variety. So the big question we're asking
• 1:19
  in this deep dive is can that same scarcity driven thinking
• 1:23
  really work when you're dealing with the sheer overwhelming
• 1:27
  scale and inflexibility of heavy industry like steel?
• 1:32
  That's the challenge. Let's dig in.
• 1:33
  OK, first let's quickly level set on the difference between
• 1:36
  these two worlds. Discrete manufacturing that's
• 1:38
  leans home turf. You make things you can count.
• 1:41
  Cars, phones, bolt. Individual units, yeah.
• 1:45
  Whereas continuous process industries, they make stuff you
• 1:47
  measure or meter oil, chemicals, cement.
• 1:51
  Or in our case, steel, liquid steel flowing.
• 1:52
  Right. Exactly.
• 1:53
  And historically, managers in these process industries,
• 1:56
  they've been pretty hesitant about lean tools.
• 1:58
  Why is that just the scale? Well, partly.
• 2:01
  You've got super high volume, but often very low product
• 2:04
  variety compared to discrete and the equipment, it's massive,
• 2:08
  fixed in place. You can't just create a cell
• 2:10
  like you would in a car plant when your blast furnace weighs
• 2:12
  thousands of tons. Right concepts like cellular
• 2:14
  manufacturing just seem impossible.
• 2:16
  They do. And then there are the setup
• 2:18
  times, the changeovers. They can be incredibly long and
• 2:22
  costly. That's a killer.
• 2:23
  I bet she can't just swap a die in 5 minutes.
• 2:26
  Not when changing over might mean shutting down a huge part
• 2:30
  of the continuous flow. We're talking potentially
• 2:32
  millions in lost production, maybe even risks to the
• 2:35
  equipment itself. So the traditional thinking is
• 2:39
  we need huge piles of inventory to buffer against these
• 2:42
  changeovers and potential downtime.
• 2:44
  OK, that makes sense from a risk perspective, but is that the
• 2:47
  whole story? Well, here's the key insight
• 2:49
  from the research. We looked at most of these
• 2:51
  so-called continuous processes. They're actually hybrids.
• 2:54
  Hybrids. How so?
• 2:56
  Think about that steel example. You have liquid steel, which is
• 2:59
  definitely continuous, flowing from the furnace through the
• 3:02
  cast. Very measured flow.
• 3:03
  But then what happens somewhere mid process, usually after
• 3:06
  casting, maybe some initial rolling, that liquid metal
• 3:09
  becomes solid. It gets cut into discrete
• 3:12
  countable units, slabs, coils, billets, whatever the form is.
• 3:17
  So there's a point where it becomes discrete.
• 3:19
  Exactly. It transitions from this non
• 3:21
  discrete flow to countable pieces and that specific point
• 3:25
  in the process, that transition, that's our opening, that's the
• 3:28
  toehold where we can start applying lean principles even if
• 3:32
  the very start of the process remains continuous.
• 3:34
  Interesting. So you don't try to force lean
• 3:37
  onto the blast furnace itself, but you find that discrete point
• 3:40
  downstream. Precisely, you leverage that
• 3:43
  hybrid nature. OK, so if we found our entry
• 3:46
  point, the next step in lean is always identifying the waste,
• 3:50
  right Muda? Absolutely.
• 3:52
  The goal is always eliminating anything that doesn't add value
• 3:55
  from the customers perspective. In heavy industry you see all
• 3:58
  the classic wastes, but some really stand out inventory
• 4:01
  obviously given what we just discussed.
• 4:03
  Those huge piles of coils or slabs just sitting there.
• 4:06
  Yeah. Overproduction often goes hand
• 4:08
  in hand with that inventory defects, which can be incredibly
• 4:12
  costly when you're dealing with massive batches and time related
• 4:16
  waste like those super long set up times we mentioned.
• 4:18
  So how do you find it all? It must be hidden in plain sight
• 4:21
  sometimes. The essential tool, really the
• 4:23
  starting point for any serious lean effort, is Value Stream
• 4:26
  mapping VM. Right, not just a simple flow
• 4:29
  chart. No, definitely not.
• 4:31
  VM tracks everything. Material flow, yes, but also
• 4:34
  information flow, the orders, the schedules.
• 4:36
  And critically, it tracks time. How long does each step take?
• 4:40
  How long does material wait between steps?
• 4:43
  How much inventory accumulates? It forces you to see the whole
• 4:45
  system, not just optimize one machine.
• 4:48
  Exactly. You visualize the entire value
• 4:50
  stream, from raw material to finished goods hitting the
• 4:53
  customer. And when the researchers applied
• 4:55
  VSM to the steel manufacturer ABS, the results were, well,
• 5:00
  pretty shocking. OK, let's hear it.
• 5:01
  What did the map reveal about their current state?
• 5:04
  For their annealed product line, the total production lead time
• 5:07
  start to finish was 46 days. 46 days, that's incredible.
• 5:11
  Over a month and a. Half.
• 5:13
  Yep. Now here's the kicker.
• 5:14
  They measured the actual value added time, the time the steel
• 5:17
  was actually being heated, rolled, cut, treat, you know,
• 5:20
  physically transformed. OK.
• 5:21
  How much is that? 5 days.
• 5:23
  Five days of actual work out of 46 days total lead time.
• 5:26
  Exactly. That means the value added time
• 5:29
  was only about 11, 7% of the total time the material spent in
• 5:32
  the system. So for 41 days, roughly 89% of
• 5:36
  the time that steel was just sitting.
• 5:39
  As inventory. As inventory waiting, waiting
• 5:42
  for the next process, waiting for a schedule change, waiting
• 5:45
  in a massive storage yard. And that right there exposed
• 5:49
  inventory as the single biggest source of waste and delay in
• 5:53
  their system. And that much inventory, it must
• 5:55
  mask us ton of other problems. Right, absolutely, it hides
• 5:58
  everything. Scheduling mistakes doesn't
• 6:00
  matter. Grab another coil from the pile.
• 6:02
  Quality issues from the caster. By the time you find the defect
• 6:05
  weeks later in the finishing mill, the conditions that caused
• 6:08
  it are long gone. Machine breakdowns just work
• 6:11
  through the buffer inventory. So reducing that inventory isn't
• 6:14
  just about saving space or capital.
• 6:16
  It's about forcing those underlying problems, the quality
• 6:19
  issues, the machine reliability, the scheduling hiccups out into
• 6:22
  the open where you have to fix them.
• 6:24
  That's the core principle. Inventory reduction reveals the
• 6:27
  rocks hidden beneath the water. OK so VSM shows a 46 day lead
• 6:31
  time mostly waiting. How do you redesign this using
• 6:35
  lean for the future state? Where do you start?
• 6:37
  You start by setting the pace you need to link production
• 6:40
  directly to customer demand. That means calculating the tact
• 6:44
  time. Tact time.
• 6:45
  The heartbeat of the system. Yeah, German word for rhythm.
• 6:47
  Exactly. It tells you how often you need
• 6:49
  to produce 1 unit, in this case a coil, to meet customer demand
• 6:52
  exactly. So for ABS, how did that
• 6:54
  calculation work? OK.
• 6:55
  The customer demand for this annealed product was 76,500 tons
• 6:59
  per month. You break that down, it's about
• 7:01
  2550 tons a day, which translates to roughly 127 coils
• 7:07
  per day based on their average coil weight.
• 7:09
  A. 127 coils needed every single day, right?
• 7:11
  Then you look at the available work time.
• 7:13
  A standard day has 4440 minutes, 24 hours times 60 minutes.
• 7:18
  Got it. And the formula is.
• 7:19
  TAC time equals available work time per day divided by customer
• 7:23
  demand per day. So $1440 minutes divided by 127
• 7:27
  Ziller coils. Which comes out to approximately
• 7:29
  $11.00 and three 3 minutes per coil.
• 7:31
  So the target is 1 finished. Coil needs to roll off the line
• 7:36
  every $11.00 three, 3 minutes to keep up with demand.
• 7:39
  No more, no less. That's the rhythm the finishing
• 7:42
  end needs to March to. OK, but here's the tricky part
• 7:44
  again. The front end, the blast
• 7:47
  furnace, the caster, that's still continuous, right?
• 7:50
  It's pushing out steel constantly.
• 7:52
  How do you reconcile that push with this new $11.00 on three
• 7:56
  three minute pull rhythm at the finishing end?
• 7:59
  That's where the hybrid pull system design comes in.
• 8:01
  You acknowledge the front end has to push at least to some
• 8:03
  extent. You can't just turn a blast
• 8:05
  furnace on and off based on an 11 minute cycle.
• 8:07
  Definitely not. So you strategically decouple.
• 8:10
  You designate one single point in the downstream discrete
• 8:13
  process as the pacemaker. This is the only point that gets
• 8:16
  a formal schedule based on customer orders.
• 8:18
  And where was that for ABS? In their future state map, it
• 8:22
  was the temper mill. Basically the final processing
• 8:24
  step before shipping for this product.
• 8:26
  Everything upstream of the temper mill only produces to
• 8:29
  replenish what the temper mill consumes.
• 8:31
  So the temper mill pulls from the steps before it.
• 8:33
  Exactly. And those massive uncontrolled
• 8:36
  inventory piles we talked about? They get replaced by small
• 8:39
  controlled supermarkets. Like a grocery store shelf but
• 8:42
  for steel coils. Sort of, yeah.
• 8:44
  You have a limited defined amount of inventory, maybe a few
• 8:47
  hours or a day's worth, not 41 days worth held between key
• 8:51
  processes. When the downstream process,
• 8:54
  like the temper mill, withdraws a coil from its upstream
• 8:58
  supermarket, a signal, a Kanban, is sent back.
• 9:01
  Like an empty slot on the shelf signaling make one more.
• 9:04
  Precisely. The Kanban tells the upstream
• 9:06
  process, OK, I've used one now you can make one to replace it.
• 9:10
  It ensures you only produce what's actually needed when it's
• 9:13
  needed. It converts that push system
• 9:15
  into a pull system after the continuous part.
• 9:18
  But you still need some buffer between the continuous hot end
• 9:21
  and the first supermarket, right To absorb variations.
• 9:23
  Yes, you manage that buffer carefully.
• 9:25
  It's calculated, not just accumulated.
• 9:28
  And importantly, you also need to level the production Hajinka.
• 9:31
  Leveling meaning not making all of one type of coil on Monday,
• 9:36
  then all of another type on Tuesday.
• 9:38
  Exactly. If your customer needs, say,
• 9:40
  five types of coils this week, you don't run huge batches of
• 9:44
  type A, then type B, etcetera. That creates massive swings and
• 9:48
  makes changeovers a nightmare. Hijunca means smoothing out the
• 9:52
  production mix. So you'd make a little bit of
• 9:54
  each type more frequently. Right.
• 9:56
  They introduced the concept of pitch, a regular timed interval
• 9:59
  for releasing small batches of work into the system in the ABS
• 10:03
  future state. This was calculated to be around
• 10:05
  every $1.00, five, five hours. Every 90 minutes, a small mixed
• 10:09
  batch of coils needed by the pacemaker would be authorized
• 10:12
  for production upstream. So TAC time sets the overall
• 10:15
  speed, Cambian supermarkets control inventory flow, and
• 10:18
  Hijunca smooths out the mix. That sounds much more
• 10:21
  controlled. It is, but making the system
• 10:24
  work relies heavily on the physical processes being stable
• 10:27
  and reliable. You can't have a pull system
• 10:29
  operating on an 11 minute cycle if your machines are breaking
• 10:32
  down randomly for hours. Right, that smooth flow depends
• 10:35
  entirely on predictable uptime. Which brings us to the technical
• 10:39
  fixes needed. Absolutely.
• 10:40
  Tool number one has to be total productive maintenance or TPM.
• 10:44
  You have to attack equipment reliability head on.
• 10:47
  What was the problem at ABS? Just infrequent maintenance.
• 10:50
  Partly they had long scheduled shutdowns like maybe furnace
• 10:53
  reline every couple of months taking almost 1000 minutes.
• 10:56
  But the real killer was the unplanned downtime.
• 10:59
  The study mentioned the casters unplanned uptime followed an
• 11:02
  exponential distribution, specifically Expo ( 20,160
• 11:08
  dollars. close parenthesis minutes.
• 11:09
  Expo ( 20,106 dollars. Close parenthesis minutes.
• 11:15
  What does that mean in practice? It means on average it might run
• 11:18
  fine for about $20,160 minutes, which is roughly 14 days.
• 11:23
  But then it could fail unpredictably, and then downtime
• 11:26
  could last for several hours. That kind of variation just
• 11:28
  destroys a poll system. So TPM aims to prevent those
• 11:32
  failures. Yes, it shifts the focus from
• 11:34
  reactive fix it when it breaks or even just long interval
• 11:37
  preventative maintenance to proactive and predictive
• 11:39
  maintenance. It involves operators and daily
• 11:41
  checks, cleaning and lubrication.
• 11:43
  It means doing shorter, more frequent plan maintenance tasks
• 11:46
  to catch problems early. Maybe a four hour targeted
• 11:49
  maintenance every three weeks instead of that massive 10 hour
• 11:51
  shutdown every two months. The goal being maximizing OEE
• 11:55
  overall equipment effectiveness. Exactly.
• 11:58
  Reducing breakdowns, minimizing minor stops and speed losses,
• 12:01
  Making the machines reliable enough to support that tack
• 12:04
  time. OK, makes sense.
• 12:06
  What's tool #2? Set of production, often called
• 12:09
  SMA. Single minute exchange of dyes,
• 12:11
  though it applies to any changeover.
• 12:13
  Tackling those long change over times we talked about if you're
• 12:17
  running smaller level batches thanks to High Junka, you must
• 12:20
  be doing changeovers much more often.
• 12:22
  Way more often, if they take hours, the whole pull system
• 12:24
  collapses. SMA DS core idea is simple but
• 12:28
  powerful. Analyze every single step of a
• 12:31
  changeover. And figure out what you can do
• 12:33
  before the machine actually stops.
• 12:35
  Precisely. You separate internal setup
• 12:37
  tasks that absolutely must be done while the machine is
• 12:39
  stopped. Like actually swapping A roll
• 12:41
  from external setup tasks that can be done while the machine is
• 12:44
  still running. Like getting the next roll
• 12:46
  ready, preparing tools, preheating things.
• 12:49
  So find ways to convert internal time to external time.
• 12:53
  That's the game for the steel mill.
• 12:54
  They targeted things like the hot strip mill and the cold
• 12:57
  reduction mill, where work roles might be changed, say, 18 times
• 13:01
  a day. 18 times OK setup time is critical there.
• 13:04
  Absolutely. So polishing the next roll,
• 13:07
  getting the tooling ready, placing hoses, all that needs to
• 13:10
  become external work happening concurrently while the current
• 13:14
  coil is running. You fight for every minute saved
• 13:17
  on this frequent changeovers. Makes sense.
• 13:19
  Reliable machines, fast changeovers.
• 13:21
  What else? The foundation You can't forget
• 13:23
  5S and visual system. I've S sort, straighten, shine
• 13:27
  or sweep, Standardize, sustain the housekeeping 1.
• 13:31
  It's way more than housekeeping. Especially in a heavy industry.
• 13:34
  There's often this culture of well, it's a steel mill.
• 13:36
  It's supposed to be dirty and cluttered. 5S directly combats
• 13:39
  that sorting, gets rid of junk straight, and organizes tools
• 13:42
  and parts shine keeps things clean, making abnormalities
• 13:46
  easier to spot. Like an oil leak on a clean
• 13:48
  floor versus a dirty. One exactly.
• 13:49
  Systematize and standardize. Make it stick.
• 13:52
  It frees up space and moves safety, boosts morale, and
• 13:55
  actually helps quality by making problems visible.
• 13:57
  And visual systems. What's an example there?
• 13:59
  Things like clear floor markings for inventory locations, or in
• 14:03
  the ABS case, they proposed a simple visual Kanban post near
• 14:07
  the shipping dock. Instead of relying purely on
• 14:10
  computer systems or paperwork which can have delays or errors,
• 14:14
  this post visually showed the status and position of coils
• 14:17
  ready for shipment. So a dock worker could instantly
• 14:20
  see, OK, this truck needs these three coils and they are
• 14:23
  physically right here in these marked spots and their status
• 14:26
  card is green. Right.
• 14:28
  It reduces errors like shipping the wrong coil or one that
• 14:31
  hasn't passed final inspection. Simple visual cues to prevent
• 14:35
  mistakes and keep things flowing.
• 14:37
  OK, so we have the hybrid pull system design supported by TPM
• 14:40
  for reliability, SMAD for quick changeovers, and 5S visual
• 14:44
  systems for foundation and air proofing.
• 14:47
  Did it work? What did the simulation show?
• 14:49
  The researchers modeled this whole future state system using
• 14:52
  simulation software arena I believe it was to predict the
• 14:55
  impact and the. Bottom line, the lead time.
• 14:57
  The results were pretty dramatic.
• 14:59
  The simulation showed that implementing the hybrid pull
• 15:01
  system along with the improvements from TPM and SMED
• 15:05
  reduced the average production lead time from the original $34
• 15:08
  and 2666 days they focused on in the detailed model.
• 15:11
  Wait, wasn't it 46 days initially?
• 15:13
  Ah. Good catch.
• 15:14
  The initial VSM showed 46 days across the entire process
• 15:18
  observed. The simulation model likely
• 15:20
  focused on a specific, slightly narrower part of that stream for
• 15:23
  detail analysis, which had a baseline of $34 and 266 days.
• 15:27
  Still a very long time. OK.
• 15:29
  So from 30 4.26 days down to. Down to 12.12 days.
• 15:34
  Wow. From over a month to under 2
• 15:36
  weeks, that's almost a 65% reduction.
• 15:38
  A massive improvement, yeah, nearly 2/3 of the lead time just
• 15:42
  that's. Huge.
• 15:43
  What does that do to? The value added percentage was a
• 15:44
  dismal 11%. Before it jumped significantly,
• 15:48
  the simulation showed the value added time as a percentage of
• 15:50
  total lead time increased to about 39%.
• 15:53
  Still room for improvement, sure, but a huge leap from 11%.
• 15:56
  Much less waiting, much more working.
• 15:58
  So it really proves the point, doesn't it?
• 16:01
  Even in these massive, supposedly inflexible continuous
• 16:04
  process industries, the core lean principles can be applied
• 16:08
  if you adapt them smartly. It shows the constraint often
• 16:12
  isn't the giant machinery itself, but the management
• 16:15
  system, the scheduling, the inventory policies, the
• 16:18
  maintenance approach, the lack of flow synchronization.
• 16:21
  That hybrid approach, finding that discrete point seems key.
• 16:25
  It's critical you have to find that interface between the
• 16:28
  continuous push and the potential for discrete pull, and
• 16:32
  manage it with tools like supermarkets and Kanban.
• 16:35
  It's a powerful demonstration that waste really is everywhere,
• 16:39
  Even baked into industries we think of is inherently
• 16:41
  inefficient due to scale. Continuous improvement is never
• 16:44
  really finished. Absolutely not.
• 16:46
  And the next logical step, the real frontier for this kind of
• 16:49
  thinking in heavy industry. Extending it beyond the plant
• 16:51
  walls. Exactly.
• 16:53
  Integrating the entire supply chain, looking for waste, not
• 16:56
  just inside the mill, but in the delivery of raw materials from
• 16:59
  suppliers and in the distribution network, getting
• 17:02
  finished goods to the final customer.
• 17:04
  That's where the next big gains are more likely hiding.
• 17:07
  Fascinating stuff, applying car factory thinking to melt shops
• 17:10
  and rolling mills and getting incredible results.
• 17:13
  Until the next equation design junkies, we are mechanical
• 17:16
  engineering made simple. Stay balanced, stay precise, and
• 17:19
  never stop building.