Lean Six Sigma Black Belt Project using JMP Sta s cal Discovery So ware
David Payne Andrew Ruddick Head of Con nuous Improvement Director Process Insight Consul ng JMP Discovery Summit, March 2015
Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
Process Insight Consul ng
We help you maximise the value of your data by driving practical application of stats into the heart of your business
• Exploiting the power of JMP ® to solve your toughest real world problems
• Data Analytics for non – statisticians o “We speak human ! ”
• Engaging the total organisation in problem solving Johnson Ma hey Divisions
Johnson Ma hey is a speciality chemicals company and a world leader in advanced materials technology.
Fine Chemicals Emission Control Precious Metal Process New Businesses Technologies Products Technologies - API - Fuel Cells Manufacture - Auto Catalysts - Refining, - Catalysts recycling & - Ba ery - Catalysts & - Heavy Duty management - Davy Technologies Chiral Diesel Technologies Technology - Noble metals, - Water - Sta onary metal joining & - Tracerco Technologies - Research Emissions Colour Chemicals Control Technologies - New Business development Macfarlan Smith
• Manufacture Ac ve Pharmaceu cal Ingredients (APIs) for world wide supply
• Major products are alkaloids derived from the opium poppy (opiates) and synthe c opioids.
• World’s largest supplier of opiate APIs with the widest product por olio of opiate products for the pain relief market
• The largest consumer of “poppy” raw materials (morphine, thebaine and oripavine)
Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
The Supply Chain
Grow poppies Harvest Poppies Separate seed & Transfer in to plant (UK & Portugal) pelle se straw from containers Project Scope Codeine Phosphate
Codeine Sulphate
Codeine Alkaloid Pure
Solvent extract Purify product, recover CPS is the raw Dihydrocodeine Morphine in belt solvent & isolate materials for the extractor Morphine as manufacture of “CPS” (concentrated many opiate Poppy Straw) products Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
The Problem & Benefits
• Big investment made but plant not delivering to target capacity
• Cri cal business need to increase output but process chao c - Extremely difficult problem to solve!
• But solu on will deliver huge value to the business
Project Scope
Poppy Growing Morphine Extraction Purification & Solvent Recovery
Pictures here
Out of Scope In Scope Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
Lots of Data to Collect & Analyse
• Field • Bulk density • Down me hours • Variety • Morphine content • Equipment breakdown • Fer liser • Belt speed • Processing delays • Pes cide • Gate Height • Technical delays • Drill date • Morphine rate • Cleaning mes • Harvest date • Poppy straw rate • Tar forma on • Weed content • Solvent rate • Filtra on issues • Seed content • Solvent ra o • Percola on issues • Pelle sing condi ons • Recircula on rate • Separa on issues • Morphine content • pH • Sugars • Concentra ons through • Starches the plant • Flows through the plant Poppy Growing Data
Extrac on Plant Data • Yield through the plant Plant Performance Data
Data stored in lots of different systems which we can bring together in JMP and easily update using simple scrip ng Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
Ini al Plant Control Findings
Operators can change extrac on condi ons to sa sfy various plant targets
Straw fed from silos to a hopper above the extractor
Solvent charged Rate to the extractor is to the belt to controlled by adjus ng extract the belt speed and gate Target = 0.26% Morphine height Morphine post extractor JMP Link to JMP to show charts below Implica ons of Poor Control
Dilution Water Flowrate
Morphine concentra ons and flows to the Bulk Density Solvent Flow In purifica on stage are not in control: Solvent Flowrate Gate Over concentrate – yield lost to waste stream Height and increased risk of choking plant HE34 HE35 Belt Speed Under concentrate – under u lised equipment & lost capacity T201 T32
Control chart of Tank T32 Product Flow Out concentra on Recirc. P34 P46 P48 P50 Rate Ini al Data Analysis Summary
• Morphine rate and Solvent feed into the plant are not well controlled
• Morphine concentra ons and solvent flow from the extractor are highly variable and off target
Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
Improvement Approach
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant
2. Demonstrate concentra on & flow control
3. Op mise concentra ons & flows to maximise output
Improvement Approach
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant
2. Demonstrate concentra on control
3. Op mise concentra ons & flows to maximise output
Ini al Data Analysis Findings – Feed Rate
Dilution Water Flowrate
Bulk density v charging rate Bulk Density Parameters thought to control charging rate show no correla on: Solvent Flowrate Gate Height Something else controlling it? – what?
Data collec on issue? HE34 H– where & why? E35 Belt Speed
T201 T32
Gate height v charging rate Belt speed v charging rate
Recirc. P34 P46 P48 P50 Rate Ini al Data Analysis Findings – Feed Rate
Conclusions: Exis ng method for calcula ng straw feed rate is flawed…. ……Needs a new approach Improvement 1 – New Feed Rate Method
Bulk Density
Hopper
Gate Height
Extractor
Belt Speed
JMP Improvement 1 - Feed Rate Model
R Squared value shows how much of the variability the model explains – nearly 90%
Prediction Profiler in JMP
P value indicates model is sta s cally significant Improvement Approach
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant Complete
2. Demonstrate concentra on & flow control
3. Op mise concentra ons & flows to maximise output
Improvement Approach
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant Complete
2. Demonstrate concentra on & flow control
3. Op mise concentra ons & flows to maximise output
Current Concentra on Control Method
Dilution Water Flowrate
Bulk Density
Solvent Flowrate Gate Key Concentration Key Concentration Height control point control point Target 0.26% Target 0.55% HE34 HE35 Belt Speed
Many process T201 variables T32
Recirc. P34 P46 P48 P50 Rate
• Mul ple key concentra on control points • Many cri cal process variables • Con nuous processing plant with lots of interac ons JMP • No understanding of the effect one change has on another Improvement 2 - Current Control Method
Dilution Water Flowrate
Bulk Density
Solvent Flowrate Gate Height Rate of Morphine to the extractor
HE34 HE35 Belt Speed
T201 T32
Recirc. P34 P46 P48 P50 Rate Improvement 2 - Current Control Method
Dilution Water Flowrate
Bulk Density
Solvent Flowrate Gate Height Solvent flowrate to the extractor
HE34 HE35 Belt Speed
T201 T32
Recirc. P34 P46 P48 P50 Rate Improvement 2 - Current Control Method
Dilution Water Flowrate
Bulk Density
Solvent Flowrate Gate HeighRa o of solvent to Morphine in the extractor t
HE34 HE35 Belt Speed
T201 T32
Recirc. P34 P46 P48 P50 Rate Improvement 2 - Current Control Method
Dilution Water Flowrate
Bulk Density Concentra on of Morphine in product stream Solvent Flowrate Gate Height
HE34 HE35 Belt Speed
T201 T32
Recirc. P34 P46 P48 P50 Rate Current Control Method
Dilution Water Flowrate
BMorphineulk Density rate to Extractor
Solvent Flowrate Gate Height
HE34 HE35 Belt Speed
T201 T32
Solvent Flowrate
Recirc. P34 P46 P48 P50 Rate Tanks T32 Morphine Concentration
Ratio of Solvent to morphine Improvement 2a – Morphine Control Method
1. Set a constant Morphine Dilution Water Flowrate feed rate to the plant Bulk Density using JMP model
Solvent Flowrate Target to stabilise Gate concentration & Control Basis: Height flow here 2. Match this Sample straw and analyse for with a HE34 Morphine. Adjust feed rate to HE35 constant Belt Speed flowrate of maintain a constant Morphine solvent to flow in T201 the extractor T32
Recirc. P34 P46 P48 P50 Rate
JMP based feed rate model gives us the tool to stabilise concentra on
JMP Improvement Result for Morphine Control Improvement Results in JMP
Predic on Profiler
Improvement 2b – Ra o Control Method
2. Vary Morphine feed rate Dilution Water Flowrate to the plant using JMP Bulk Density model
Solvent Flowrate Target to stabilise Gate concentration & Control Basis: Height flow here 3. Maintain a Sample product stream constant HE34 and adjust feed rate to HE35 flowrate of Belt Speed solvent to maintain a constant the extractor Morphine concentra on T201 T32
Recirc. P34 P46 P48 P50 Rate 1. Sample tank and calculate the ratio of target concentration to actual
JMP based feed rate model s ll fundamental to plant control JMP Improvement Results in JMP Improvement Results
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant Complete
2. Demonstrate concentra on & flow control Complete
3. Op mise concentra ons & flows to maximise output
Improvement Results
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant Complete
2. Demonstrate concentra on & flow control Complete
3. Op mise concentra ons & flows to maximise output
Improvement Results
Rate increased incrementally over several weeks all the me using JMP to analyse data and ensure key parameters remained in control Control – Automated Data Clean Up & Dashboards
Daily Rou ne
Model (kg/hr) since last T32 (kg)
GRN Farm Variety Soil Type Geo Area Drill Date Cultivation Method Drill Type Harvest Date Harvester GRN In Extractor AMA Content of Poppy Straw (%)Poppy Straw Bulk Density (ex pelletiser)Date Time AMA Content at Belt HopperPoppy straw bulk density at belt (g/ml)Belt Speed (Hz)Gate Height (cm)Calculated Feed Rate from Model (te/hr)Calculated rate of AMA to the belt from Total AMA fed top the belt (from model) Recirculation Rate (m³/hr)Solvent Flowrate to belt (l/hr)T32 AMA Content (%)Flowrate (T32 to WF1) (l/hr)AMA Through Put (ex T32) (kg/hr) 26029 RR Burge Griffon Chalk Hants 05/04/2013 Plough JD 11-Aug-13 MS Hants 26109 2.01% 0.57 27/01/2014 07:45 0.59 28 54.0 0.4321 8.6849 27.4 8.54 3988 0.320 2811 9.00 26109 2.01% 0.57 27/01/2014 10:55 28 2603054.0 Rowland Griffon 8.54 Chalk3935 Hants 299803/04/2013 Plough Spirit 9-Aug-13 Dawe 26109 2.01% 0.57 27/01/2014 15:30 0.59 28 2603154.0 RR Burge/Woodhouse0.4321 8.6849Griffon 67.9 8.54 Chalk3400 0.310 2934 9.10 26109 2.01% 0.57 27/01/2014 19:45 27 2603254.0 Rowland Griffon 8.55 Chalk3364 Hants 292103/04/2013 Plough Spirit 9-Aug-13 Dawe 26109 2.01% 0.57 27/01/2014 23:30 0.59 27 2603354.0 Hosford0.4114 8.2695Osprey 67.5 8.54 Chalk3562 Dorset0.310 297408/04/20139.22 Min till Rapid 15-Aug-13 Dawe 26109 2.01% 0.57 28/01/2014 03:30 27 2603454.0 Hosford Osprey Chalk4231 Dorset 08/04/2013 Min till Rapid 15-Aug-13 Dawe 26109 2.01% 0.57 28/01/2014 06:50 27 54.0 3750 2913 26109 2.01% 0.57 28/01/2014 07:30 0.600 27 2603554.0 Hosford/Strang 0.4186 8.4144Griffon/Osprey12.8 8.54 4096 0.304 2978 9.05 Min till Dawe 26109 2.01% 0.57 28/01/2014 11:30 27 2603654.0 Crowmarsh Sparrowhawk 8.54 Chalk3432 Oxon 06/04/2013 Min till Rapid 20-Aug-13 MS Oxon 26109 2.01% 0.57 28/01/2014 15:30 0.625 27 2603754.0 Bryce/Foot0.4367 8.7767Griffon/Sparrowhawk70.2 Clay3348 Berks0.275 2858 7.86 Plough MS Hants 26109 2.01% 0.57 28/01/2014 15:50 27 2603854.0 Doble Elenoris/Griffon8.54 Gravel3932 Oxon 285808/04/2013 Plough Carrier 22-Aug-13 MS Oxon 26109 2.01% 0.57 28/01/2014 23:30 0.606 27 2603954.0 Hewison/Whitsbury0.4230 8.5014Griffon/Osprey40.8 8.54 3354 0.203 2866 5.82 Plough 26109 2.01% 0.57 29/01/2014 03:30 27 54.0 8.61 3510 26040 Foot/Hagbourne/CrowmarshOsprey/Sparrowhawk 26109 2.01% 0.57 29/01/2014 07:30 0.625 27 54.0 0.4367 8.7767 70.2 8.61 3635 0.273 2846 7.77 26109 2.01% 0.57 29/01/2014 08:15 27 2604154.0 Crowmarsh Sparrowhawk Chalk3610 Oxon 287806/04/2013 Min till Rapid 20-Aug-13 MS Oxon 26109 2.01% 0.57 29/01/2014 11:00 27 2604254.0 Woodhouse/Drax Griffon 8.61 Chalk4300 2907 Min till 26109 2.01% 0.57 29/01/2014 16:15 0.617 27 2604354.0 Tenantry0.4309 8.6607Elenoris34.6 8.62 Chalk3808 Hants0.355 260304/04/20139.24 Plough Pronto 20-Aug-13 Dawe 26109 2.01% 0.57 29/01/2014 17:00 27 2604454.0 Drax Griffon Chalk3948 Dorset 06/04/2013 Min till Rapid 13-Aug-13 Dawe 26109 2.01% 0.57 29/01/2014 19:30 27 2604554.0 Drax Griffon 6.67 Chalk3932 Dorset 06/04/2013 Min till Rapid 13-Aug-13 Dawe Join mul ple 26109 2.01% 0.57 29/01/2014 23:30 0.633 27 54.0 0.4424 8.8926 64.5 7.41 3939 0.351 2609 9.16 13-Aug-13 26269 2.07% 0.57 30/01/2014 03:30 27 2610354.0 Drax Griffon#VALUE! 7.49 Chalk3953 Dorset0.289 06/04/20130.00 Min till Rapid Dawe 26269 2.07% 0.57 30/01/2014 07:30 27 2610454.0 Hildred Griffon#VALUE! 8.09 Clay Cap4034 Oxon0.347 238806/04/20138.29 Min till Spirit 11-Aug-13 MS Oxon 26269 2.07% 0.57 30/01/2014 11:30 27 2610554.0 Drax Griffon Chalk Dorset 06/04/2013 Min till Rapid 13-Aug-13 Dawe 26269 2.07% 0.57 30/01/2014 12:00 27 2610654.0 Drax/Chalcraft Griffon 8.05 3086 Rapid Dawe 26269 2.07% 0.57 30/01/2014 15:35 27 2610754.0 Hildred Griffon#VALUE! 8.05 Clay Cap3542 Oxon0.351 273006/04/20139.58 Min till Spirit 11-Aug-13 MS Oxon tables in JMP 26108 Owen Day Elenoris Sand Lincs 06/04/2013 Plough Combination 20-Aug-13 Fox/MS Lincs 26109 Blankney Griffon/SparrowhawkBrash Lincs 08/04/2013 Plough Combination 15-Aug-13 Fox/MS Lincs 26110 Blankney Griffon/SparrowhawkBrash Lincs 08/04/2013 Plough Combination 15-Aug-13 Fox/MS Lincs 26111 Blankney Griffon/SparrowhawkBrash Lincs 08/04/2013 Plough Combination 15-Aug-13 Fox/MS Lincs 26112 RR Burge Griffon Chalk Hants 05/04/2013 Plough JD 11-Aug-13 MS Hants 26221 Chalcraft /O Day/N JacksonElenoris/Griffon
Process data in Excel from JMP Script to mul ple sources auto-clean data
JMP Script to create dashboard Improvement Results
Split improvements into 3 stages:
1. Control feeding rate and therefore Morphine into the plant Complete
2. Demonstrate concentra on control Complete
3. Op mise concentra ons & flows to maximise output Complete
Contents • Macfarlan Smith & Process Insight Consul ng
• The Morphine supply chain
• Define the problem
• Collec ng plant performance data
• Using JMP to analyse data sets
• Bring the process under control
• Future work using JMP
Next Steps – Ver cal Integra on of CI work
• Growing • Extrac on • Process loca ons condi ons condi ons to • Growing • Purifica on maximise condi ons condi ons yield • Poppy variety • Morphine • Process • Fer lizer type product condi ons to minimise • Pes cide type specifica on impuri es • Drill dates • Minimise • Harves ng process point variability • Pellet type • Increase • Reduced Extrac on Op misa on product RFT product rate variability Poppy Growing Op misa on Opiate Processes Op misa on We Could Not Have Done This Without JMP
• Pin point the issues in a fog of data
• Visualise the data to easily spot pa erns, trends & outliers
• Powerful modelling, profiling & what-if simula on
• Easy communica on of process issues through simple visualisa on
• Ease of data clean up, manipula on & dash- boarding Process Insight Consul ng
We help you maximise the value of your data by driving practical application of stats into the heart of your business
• Exploiting the power of JMP ® to solve your toughest real world problems
• Data Analytics for non – statisticians o “We speak human ! ”
• Engaging the total organisation in problem solving Ques ons?
Thanks JMP!