3.2 Process Improvement Techniques & Loss Elimination

Process Improvement Techniques & Loss Elimination

Quick Answer: True manufacturing reliability requires more than fixing broken components; it requires eliminating process instability and chronic manufacturing losses. By deploying Lean, Six Sigma (DMAIC), and Kaizen frameworks alongside Statistical Process Control (SPC) and process capability metrics ($C_p$ and $C_{pk}$), organizations eliminate the 'Hidden Factory' of chronic micro-stoppages, scrap, and rework. Differentiating common-cause variation from special-cause variation prevents operator over-adjustment, stabilizes machinery, and protects plant throughput.

---

Continuous Improvement Frameworks: Lean, Six Sigma, and Kaizen

CMRP Function 2.2 covers identifying production losses and establishing a continuous-improvement process. Lean, Six Sigma, and Kaizen are useful examples, but the outline does not prescribe one branded method.

1. Lean Manufacturing and the Elimination of Waste

Originating from the Toyota Production System (TPS), Lean focuses on maximizing customer value by systematically identifying and eliminating Muda (waste), Muri (overburden on operators and machinery), and Mura (unevenness or variability in workflow).

In industrial maintenance and manufacturing environments, Lean targets the 8 Operational Wastes (DOWNTIME):

• D — Defects: Producing scrap parts or experiencing maintenance rework due to incorrect assembly, improper torque, or misalignment.
• O — Overproduction: Running equipment beyond customer demand, creating excess inventory that ties up working capital and causes equipment wear.
• W — Waiting: Maintenance craft waiting for safety work permits, equipment cooldown, parts delivery from the storeroom, or engineering approvals.
• N — Non-utilized Talent: Failing to capture technician diagnostic feedback or ignoring operator ideas for autonomous equipment care.
• T — Transportation: Excessive movement of heavy motors, gearboxes, or rebuild components between plant areas and external repair shops.
• I — Inventory: Excess spare parts rotting on storeroom shelves, or massive work-in-progress (WIP) buffers hiding machine unreliability.
• M — Motion: Technicians walking miles per shift searching for special tools, rigging equipment, calibration instruments, or job documentation.
• E — Extra-processing: Over-maintaining assets through unnecessary calendar-based overhauls on healthy equipment that introduce infant mortality.

2. Six Sigma and the DMAIC Framework

While Lean targets velocity and waste elimination, Six Sigma focuses on reducing process variation and eliminating defects. Developed by Motorola and popularized by General Electric, Six Sigma targets an error rate of no more than 3.4 defects per million opportunities (DPMO), corresponding to a statistical process width of six standard deviations between the process mean and specification limits.

Six Sigma executes process improvement projects using the rigorous five-phase DMAIC roadmap:

1. Define: Articulate the business problem, failure frequency, project scope, baseline financial impact, and customer impact (Project Charter, Voice of Customer [VOC], High-level SIPOC: Suppliers, Inputs, Process, Outputs, Customers).
2. Measure: Collect reliable baseline data on process performance, cycle times, component lifetimes, and failure rates. Validate measurement systems through Gage Repeatability & Reproducibility (Gage R&R) to ensure instruments are accurate.
3. Analyze: Interrogate data to identify root causes of variation and failure. Utilize Fishbone (Ishikawa) diagrams, 5-Whys, Pareto charts, Failure Modes and Effects Analysis (FMEA), and multi-vari statistical testing (ANOVA, regression analysis).
4. Improve: Develop, pilot, and deploy targeted engineering and procedural solutions that directly address verified root causes. Redesign components, modify operating envelopes, or optimize lubrication regimes.
5. Control: Standardize operating procedures, implement Statistical Process Control (SPC) charts, establish visual standard work, train technicians, and track key process indicators to ensure gains do not erode over time.

3. Kaizen and Rapid Improvement Events

Kaizen (Japanese for 'change for the better') embodies the philosophy of continuous, incremental improvement driven by shop-floor teams. Rather than waiting for multi-million-dollar capital overhauls, Kaizen empowers front-line operators and maintenance technicians to implement daily, low-cost modifications.

When cross-functional teams tackle complex, stubborn operational roadblocks, they convene a Kaizen Event (or Kaizen Blitz). A Kaizen event is an intensive, highly focused 3- to 5-day workshop where a dedicated team—comprising operators, mechanics, electricians, process engineers, and supervisors—maps a localized process, identifies root-cause defects, prototypes physical or workflow modifications, and implements immediate changes on the shop floor before the event concludes.

---

Statistical Process Control (SPC): Principles and Variation Types

Pioneered by Walter A. Shewhart at Bell Laboratories and championed by W. Edwards Deming, Statistical Process Control (SPC) is the application of statistical methods to monitor, control, and optimize processes. The fundamental premise of SPC is that every manufacturing and operating process exhibits variation.

Common-Cause vs. Special-Cause Variation

A critical responsibility of reliability professionals is distinguishing between the two fundamental types of variation:

• Common-Cause Variation (Chance / Inherent Variation): The natural, random, background noise inherent in a stable process. It is caused by numerous small, unavoidable factors: slight ambient temperature fluctuations, minor raw material lot differences, or normal electrical supply variations. A process exhibiting only common-cause variation is statistically in control and predictable over time. Crucial Rule: Common-cause variation can only be reduced by fundamentally redesigning the system, upgrading machinery, or re-engineering the process. Treating common-cause variation as if it were a special cause is known as tampering (Deming's Funnel Experiment), which destabilizes the process and increases total variation.
• Special-Cause Variation (Assignable Variation): Unnatural, erratic variation caused by specific, identifiable external events or sudden mechanical changes. Examples include a seized bearing, a broken thermocouple, an operator entering an incorrect setpoint, a batch of contaminated lubricating oil, or a plugged slurry nozzle. Special causes render the process statistically out of control and unpredictable. Special-cause variation requires immediate operational and maintenance troubleshooting to identify and eliminate the assignable root cause.

Control Limits vs. Specification Limits

One of the most common confusions on the CMRP exam is the distinction between control limits and specification limits:

• Control Limits (UCL and LCL): Calculated entirely from historical process data. They reflect the actual performance and natural variability of the process ('the Voice of the Process'). They are traditionally set at three standard deviations ($3\sigma$) above and below the process mean.
• Specification Limits (USL and LSL): Established by customer requirements, engineering tolerances, or equipment design ratings ('the Voice of the Customer' or 'the Voice of Design'). They are independent of actual machine performance.

Core Control Charts

SPC charts plot sample statistics chronologically against calculated Upper and Lower Control Limits:

• $\bar{X}$ and $R$ Charts (Average and Range): Used for continuous variables with small subgroup sample sizes ($n = 2$ to $9$). The $\bar{X}$ chart monitors process central tendency (mean), while the $R$ chart monitors process dispersion (spread/range).
• $\bar{X}$ and $s$ Charts (Average and Standard Deviation): Used for continuous variables when subgroup sample sizes are larger ($n \ge 10$), where sample standard deviation ($s$) provides a more efficient estimator of variation than the range ($R$).
• $I-MR$ Charts (Individuals and Moving Range): Used when subgrouping is impossible or impractical, such as in continuous chemical processes, batch blending with long cycle times, or once-per-shift oil analysis measurements.

---

The Western Electric & Nelson Control Chart Rules

A process is out of statistical control if sample points fall outside control limits or if non-random patterns occur within the control limits. To detect special-cause variation early, industrial quality and reliability engineers utilize the Western Electric Rules (and expanded Nelson Rules), dividing the control chart region between the centerline (mean $\mu$) and control limits ($3\sigma$) into three equal $1\sigma$ zones: Zone C ($0$ to $1\sigma$), Zone B ($1\sigma$ to $2\sigma$), and Zone A ($2\sigma$ to $3\sigma$).

SPC Control Chart Interpretation Table

Process Capability Indices: $C_p$ vs. $C_{pk}$

While Statistical Process Control determines whether a process is stable (predictable over time), Process Capability Analysis determines whether a stable process is capable of consistently manufacturing product within engineering specification limits (USL and LSL).

Process capability is evaluated using two primary statistical indices: $C_p$ and $C_{pk}$.

Potential Capability ($C_p$)

The Process Capability Index ($C_p$) measures the potential capability of the process, comparing the allowable engineering tolerance width against the actual natural spread ($6\sigma$) of the process. It assumes the process distribution is perfectly centered between specification limits:

$C_p = \frac{\text{USL} - \text{LSL}}{6\sigma}$

Where:

• $\text{USL} = \text{Upper Specification Limit}$
• $\text{LSL} = \text{Lower Specification Limit}$
• $\sigma = \text{Standard deviation of the process}$ (estimated from within-subgroup variation, e.g., $\frac{\bar{R}}{d_2}$ or $\frac{\bar{s}}{c_4}$)

Limitation of $C_p$: $C_p$ evaluates only the dispersion (spread) of the process. It completely ignores whether the process mean is centered near nominal or whether it has shifted directly on top of a specification limit. A process with an outstanding $C_p = 2.0$ could be producing 50% scrap if its mean has drifted outside the specification window.

Actual Capability ($C_{pk}$)

The Actual Process Capability Index ($C_{pk}$) measures actual process performance by accounting for both process dispersion and process centering relative to the nearest specification limit. It is computed as the minimum of the upper capability ($C_{pu}$) and lower capability ($C_{pl}$):

$C_{pu} = \frac{\text{USL} - \mu}{3\sigma}$ $C_{pl} = \frac{\mu - \text{LSL}}{3\sigma}$ $C_{pk} = \min\left(C_{pu}, C_{pl}\right) = \min\left(\frac{\text{USL} - \mu}{3\sigma}, \frac{\mu - \text{LSL}}{3\sigma}\right)$

Where:

• $\mu = \text{Actual process mean}$

Mathematical and Geometric Relationship Between $C_p$ and $C_{pk}$

• When $C_p = C_{pk}$: The process is perfectly centered on the midpoint of the specification limits ($\mu = \frac{\text{USL} + \text{LSL}}{2}$).
• When $C_p > C_{pk}$: The process is off-center. While the process spread is narrow enough to meet tolerances, the mean has shifted toward either the USL or LSL. The difference between $C_p$ and $C_{pk}$ reflects the degree of off-centering (quantified by the metric $k = \frac{|m - \mu|}{(\text{USL} - \text{LSL})/2}$, such that $C_{pk} = C_p(1 - k)$).
• When $C_{pk} = 1.00$: The 3-sigma process tail touches the nearest specification limit. Assuming a normal distribution, approximately 0.27% (2,700 parts per million) will be defective.
• When $C_{pk} < 1.00$: The process is incapable. The distribution overlaps the specification limit, generating ongoing off-spec scrap or rework.
• When $C_{pk} \le 0$: The process mean has shifted completely outside the specification limits, producing over 50% defective product.

Worked Engineering Example

A high-precision CNC lathe turns hardened steel pump shafts. The engineering specification for shaft journal diameter is $50.000\text{ mm} \pm 0.030\text{ mm}$.

• $\text{USL} = 50.030\text{ mm}$
• $\text{LSL} = 49.970\text{ mm}$
• Total Tolerance Spread: $\text{USL} - \text{LSL} = 0.060\text{ mm}$

SPC sampling reveals:

• Process Mean ($\mu$) = $50.010\text{ mm}$ (shifted $0.010\text{ mm}$ toward USL due to tool fixture thermal expansion)
• Process Standard Deviation ($\sigma$) = $0.005\text{ mm}$

Calculate $C_p$: $C_p = \frac{50.030 - 49.970}{6(0.005)} = \frac{0.060}{0.030} = 2.00$ Interpretation: With $C_p = 2.00$, the machine's inherent mechanical precision is outstanding; its spread spans only half the allowable tolerance band.

Calculate $C_{pk}$: $C_{pu} = \frac{50.030 - 50.010}{3(0.005)} = \frac{0.020}{0.015} = 1.33$ $C_{pl} = \frac{50.010 - 49.970}{3(0.005)} = \frac{0.040}{0.015} = 2.67$ $C_{pk} = \min(1.33, 2.67) = 1.33$ Interpretation: Even though $C_p = 2.00$, the off-center mean pulls $C_{pk}$ down to $1.33$. Whether that result is acceptable depends on the approved capability criterion and evidence that the process is stable. Investigate the shift and, if authorized, re-center the process while controlling the risk of oversize output.

---

Interpreting Capability Indices Carefully

Capability thresholds belong to the customer, product, and quality system; they are not universal SMRP maintenance mandates. If the process is stable, approximately normal, and the specification is two-sided, a centered $C_{pk}=1.00$ places each specification limit three standard deviations from the mean, while $C_{pk}=1.33$, $1.67$, and $2.00$ represent progressively greater separation from the nearer limit.

Do not combine a short-term six-sigma distance with a long-term 1.5-sigma-shift defect rate without stating the model. A centered normal process with a six-standard-deviation distance to each limit has a different predicted defect rate from a process whose mean is assumed to shift. Sample estimates also have uncertainty.

Process capability is not a control chart. First establish statistical stability, validate the measurement system, and confirm that engineering specification limits are legitimate. Then compare capability with the documented customer or regulatory requirement. Maintenance action follows evidence of an equipment-related cause; a poor capability result alone does not prove a spindle, bearing, sensor, or lubrication fault.

The 'Hidden Factory': Uncovering Unmeasured Capacity Losses

A central concept in SMRP Pillar 2 is the Hidden Factory (a term coined by quality pioneer Armand Feigenbaum). The Hidden Factory refers to the unmeasured, undocumented portion of plant capacity absorbed by correcting defects, running rework loops, sorting defective batches, dealing with chronic micro-stoppages, and running machinery at reduced speeds.

In conventional accounting and reactive maintenance environments, leadership tracks only major, catastrophic downtime events (e.g., equipment stops $> 15$ or $> 30$ minutes). Consequently, the myriad chronic micro-losses occurring every hour remain invisible:

• Chronic Micro-Stoppages (< 5 minutes): A carton jamming in a guide rail, a photoelectric sensor blinding with dust, a pneumatic cylinder hesitating on return stroke, or a bag unloader sticking. Operators clear these jams manually in 30 seconds without logging them into the CMMS. Yet 50 such occurrences per shift steal an hour of productive capacity and subject clutches, motor contactors, and mechanical drive linkages to constant stop-start thermal shocks.
• Reduced Speed Losses (Idling and Slow Running): Running a bottling line at 75% of nameplate design speed because running at 100% causes chronic bottle tipping or conveyor motor tripping. The plant loses 25% of its installed capital capacity simply because mechanical alignment and sensor tuning have degraded.
• Rework and Recycling Loops: Scrap plastic reground and fed back into extruders, or off-spec chemical batches routed back to intermediate holding tanks for re-distillation. This consumes vast amounts of boiler steam, compressor air, electrical power, and mechanical asset life without generating a single incremental pound of sellable product.

Quantifying the Hidden Factory

Industrial research indicates that the Hidden Factory typically consumes 20% to 40% of total plant capacity in non-optimized facilities. Reliability leaders expose and eliminate the Hidden Factory by:

1. Installing automated PLC-connected downtime tracking systems that log every stop down to the single second.
2. Conducting detailed Overall Equipment Effectiveness (OEE) gap analyses, disaggregating losses into Availability, Performance, and Quality components.
3. Empowering operators through Autonomous Maintenance (Total Productive Maintenance - TPM) to eliminate chronic micro-stops via root-cause defect elimination.

Value Stream Mapping (VSM): Flow and Lead-Time Optimization

Value Stream Mapping (VSM) is a Lean diagnostic tool that creates a comprehensive visual diagram of the material and information flows required to bring a product or service from initial raw material receipt through final customer delivery. While traditionally applied to manufacturing assembly lines, VSM is heavily utilized by reliability leaders to streamline maintenance turnaround workflows, storeroom logistics, and equipment overhaul cycles.

Anatomy of a Value Stream Map

A standard VSM captures two distinct flows:

1. Information Flow (Top Half of Map): Illustrates how customer orders, production schedules, MRP forecasts, and maintenance work orders are transmitted from enterprise software down to the shop floor.
2. Material / Physical Flow (Bottom Half of Map): Traces physical movement from left to right, documenting processing steps, intermediate inventory buffers, cycle times, changeover times, equipment uptime, and labor counts.

The Timeline and Process Lead Time

At the bottom of the VSM sits the Timeline ladder, which starkly contrasts two metrics:

• Value-Added Time (VA): Time spent actively transforming the physical shape, chemical composition, or function of the product in a way that the customer is willing to pay for (e.g., machining, stamping, painting, chemical reaction). In typical plants, VA time represents less than 5% of total lead time.
• Non-Value-Added Time (NVA / Waste): Time spent waiting in storage, sitting in staging areas, waiting for quality testing, or transporting between bays. NVA time comprises the remaining 95%+ of lead time.
• Process Cycle Efficiency (PCE): $\text{PCE} = \frac{\sum \text{Value-Added Time}}{\text{Total Process Lead Time}} \times 100\%$

Maintenance Turnaround Application

When applied to major equipment overhauls or shutdown/turnaround/outage (STO) events, VSM maps the complete work package cycle: from work order generation, engineering approval, parts kitting in the MRO storeroom, equipment lockout/tagout (LOTO), staging of cranes, actual repair, to post-maintenance commissioning. Eliminating NVA waiting time compresses outage durations, returning assets to revenue generation days earlier.

---

5S Workplace Organization and Visual Management in Maintenance

Originating in Japan, 5S is a foundational Lean methodology designed to create an organized, clean, visual, and highly disciplined workplace. In reliability operations, 5S transforms cluttered maintenance shops and chaotic production floors into high-performing environments where equipment anomalies and safety hazards are instantly apparent.

The Five Pillars of 5S

1. Sort (Seiri): Separate necessary from unnecessary items. Inspect the maintenance shop and equipment area, eliminating scrap metal, broken tools, abandoned spare parts, obsolete gaskets, and empty fluid drums. Utilize the Red Tag technique: any item without a verified, near-term functional purpose is red-tagged, moved to a holding area, and discarded or returned to the storeroom after 30 days.
2. Set in Order (Seiton): Arrange necessary items systematically so they are easy to locate, retrieve, use, and return. Establish shadow boards for hand tools, labeled racks for rigging gear, and dedicated floor striping for mobile hydraulic carts. Adhere to the core visual rule: 'A place for everything, and everything in its place.'
3. Shine (Seiso): Clean the workplace and machinery thoroughly. In reliability engineering, cleaning is inspection. As technicians and operators wipe down gearboxes, clean motor cooling fins, and sweep pump skids, they actively search for early failure symptoms: dripping oil seals, weeping flange gaskets, missing guard bolts, cracked conduit, and excessive heat.
4. Standardize (Seiketsu): Establish visual standard operating procedures (SOPs), color-coding standards, and checklists so that the first three S's are sustained uniformly across all shifts. Examples include standardizing torque wrench calibration schedules and posting visual lube inspection standards directly on machinery.
5. Sustain (Shitsuke): Build cultural habit and organizational self-discipline through regular 5S audits, leadership gemba walks, recognition programs, and integrating 5S compliance into technician performance evaluations.

The Visual Factory in Maintenance and Reliability

Visual management uses conspicuous, controlled cues to help an operator, technician, or manager distinguish normal from abnormal conditions quickly. The cue must match the approved engineering limit and should not substitute for an instrument, alarm, or procedure where those are required:

• Rotational Direction Arrows: Bright vinyl directional arrows affixed to pump casings, fan shrouds, and conveyor drives prevent reverse motor rotation following electrical maintenance.
• Lubricant Color-Coding: Matching color-coded grease guns, bulk oil transfer containers, and grease fittings to specific ISO VG oil viscosities prevents catastrophic lubricant cross-contamination.
• Gauge Operating Ranges: Pressure, temperature, and differential pressure dial gauges marked with green (normal), yellow (caution), and red (danger/trip) pie-wedge decals, allowing immediate visual verification without memorizing operating limits.
• Oil Level Sight Glasses: Bulls-eye sight glasses and columns marked with high and low operating lines, ensuring oil levels are maintained within the hydrodynamic lubrication zone.
• Bolt Witness Marks (Torque Strikes): Fluorescent lacquer lines drawn across torqued bolt heads and mating flanges reveal mechanical loosening at a glance during daily walkdowns.