Contact details
1.3 The concepts of reliability and robust engineering in diesel engine system
1.3.2 The concept of variability
The concept of variability represents uncertain or uncontrolled noise factors and their impact on system sensitivity. It is a concept of nondeterministic design. Uncertainty introduces risk. A system sensitive to noise factors exhibits unstable performance (i.e., not robust). Traditionally, diesel engines were designed under deterministic conditions, namely under standard controlled laboratory testing conditions or certification conditions. Design quality could be guaranteed by controlling the product within the design specifications.
However, when the diesel engine is put into real world usage, many problems caused by various noise factors surface in off-design conditions. The system may be over-designed for all the possible operating conditions (e.g., excessive margins on emissions, peak cylinder pressure, or coolant heat rejection), and consequently the product cost is too high. Or, the system could be under- designed for certain noise conditions and consequently it fails. In diesel engine system design robust engineering is needed in order to avoid those two scenarios. Design for variability (or design for probability) is required in order to achieve a cost-effective design for the majority of the product population. Moreover, design for variability is needed in order to detect at an early stage system failure modes caused by noise factors and to ensure reliability.
The risk associated with uncertainty or variability can be classified into system risk and design risk for an engine product. The risks are managed through FMEA (failure mode effect analysis). A detailed explanation of FMEA was provided by Stamatis (2003). A system/design FMEA approach usually consists of the following three steps:
1. Assess risk by identifying potential failure modes, the likelihood of occurrence, and the severity of their effects.
2. Establish priorities by ranking the failure modes with the risk priority number (RPN).
3. Take action to implement design/process changes to minimize the risk.
A system FMEA process for diesel engine system design is illustrated in Fig. 1.9. The FMEA is initiated by a cross-functional team of engineering, manufacturing, and reliability after the concept design is finished. The FMEA needs to be updated continuously prior to design release. Arcidiacono and Campatelli (2004) introduced an approach of failure mode and effect tree analysis (FMETA) for the reliability design process.
Tolerances are defined as the limit at which some economically measurable action is taken (Fowlkes and Creveling, 1995). Tolerances establish acceptable limits in specifications. In the system/design FMEA, it is always assumed that the failure modes result from design deficiencies. It is important to make
Identify engine system functions Draw functional block diagramConstruct interface matrix Reliability-based P-Diagrams List the effects of failure modesAssess severity (S) of the effects Assess occurrence (O) of the effectsCalculate criticality C=S¥OPrioritize for risk reduction Assess detection (D) of failure modes Calculate Risk Priority Number RPN=S¥O¥D Reliability-based engine system design
Identify failure modes with high severity
Brainstorm potential system failure modes Brainstorm the causes of failure modes List design controls & process controls 1.9 System FMEA process in engine system design.
the design specifications wide enough in tolerance (i.e., not over-constrained) so that the manufacturing process variation has little or no impact on the system/product performance.
The causes of failure modes may come from system design, component design, material selection, prototype build, etc. The failure modes can be classified into two categories: noise related and non-noise related. Each category is handled by a specific robust engineering tool. The functional block diagram and interface matrix for system FMEA development can be used to identify the non-noise-related failure modes. The purpose of a functional block diagram (Fig. 1.10) is to identify all the inputs and outputs of the functional blocks as well as their interfaces in the forms of physical or force connection, energy transfer, material exchange, and information or signal exchange. The system interface matrix (Fig. 1.11) quantifies the interfaces in terms of strength and importance and their potential effects. It is a useful tool for managing interfaces and the potential causes of failure resulting from subsystem interactions in diesel engine system design. Interactions present if the behavior of one subsystem depends on that of another subsystem.
The parameter diagram (i.e., P-Diagram, Fig. 1.12) is used to identify the noise-related failure modes. The P-Diagram identifies intended inputs and outputs, noise factors, control factors, and error states. Noise factors are unintended interfaces or sources of disturbing influences that may cause a deviation, disruption, or failure of the function during the mission time of the component or the engine. They are the causes of failure modes. Noise factors are uncontrollable (i.e., impossible, impractical or expensive to control) in product design or in-use operation. Generally, noise factors include the following five sources:
∑ Piece-to-piece or unit-to-unit variation (e.g., manufacturing variability in component geometry or material properties, variations in product- controlled parameters such as engine compression ratio, valve timing, turbocharger variance, fuel injection timing, injector variance or drift, tolerances in the controllable variables).
∑ Internal environment noise (also called system interaction noise or proximity noise, i.e., the unwanted effect of one subsystem on another, or subsystem interactions due to the variation of the input from neighboring subsystems or from the in-vehicle system operating environment, for example, variability in sensor signals, variance and drift of the air flow sensor, variability in the exhaust gas temperature or engine-out emissions composition).
∑ External environment noise (e.g., ambient temperature, humidity, altitude, road surface condition).
∑ Customer usage (e.g., accidental or foreseeable misuse and abuse of the product, real world usage duty cycle or load, different types of fuel or
Gas flow Fuel flow Coolant flow Power Control signal
Fuel systemEngine controls and calibration Power cylinder & piston assembly Waste heat recovery system
Accessories CrankshaftHybrid powertrainCylinder head Valvetrain Intake manifoldExhaust manifoldAftertreatment TurbochargerEGR system
Engine brakeVehicle drivetrain
Cooling system (Optional) (Optional)(Optional) 1.10 Robustness tool – functional block diagram of diesel engine system.
Diesel engine system Effects (to) Vehicle drive- train
Hybrid power- train Hybrid power- train
Power cylinder & piston Power cylinder & piston
Waste heat recovery Waste heat recovery
Cylinder head Cylinder head
Valve- train
Valve- train
Connect- ing rod Connect- ing rod
Crank- shaft
Crank- shaft Turbo- charger
Turbo- charger
EGR system EGR system
Fuel system Fuel system
Cooling system Cooling system
Engine brake Engine brake
Acce- ssories Acce- ssories
Engine controls Engine controls
After- treat- ment After- treat- ment Engine cali- bration
Intake manifold Intake manifold
Exhaust manifold Exhaust manifoldVehicle drive- train
Causes (from)
System FMEA interface matrix P: Physical connection E: Energy transfer M: Material exchange I: Information exchange 2 = Necessary for function (i.e., required) 1 = Beneficial but not absolutely necessary for functionality (i.e., desired) 0 = Does not affect functionality (i.e., indifferent) –1 = Causes negative effects but does not prevent functionality (i.e., undesired) –2 = Must be prevented to achieve functionality (i.e., detrimental) 1.11 Robustness tool – interface matrix and subsystem interaction of diesel engine system.
1. Piece-to-piece variation (e.g., tolerance) 2. Internal noise due to neighboring subsystems 3. external
environmental noise (e.g., climate, road) 4. Customer usage profile and duty cycle 5. Changes over time
Iterate
Noise factors (uncontrollable due to variations)
Robust design process:
• Analyze the sources of noise and magnitudes
• Analyze the effects of noise factors or their combinations on error states (strong/weak/no interaction)
• Assess the robustness of the design
• Generate noise factor management strategy
• Parameter design, tolerance design, design for variability
• Reliability confirmation test
Robust design (a link) Error states (failure modes entered in FMEA)
Input
Control factors Engine system or
subsystem
Output response or quality characteristic
(ideal or intended function)
1.12 Robustness tool – parameter diagram (P-Diagram) and robust design process.
lubricant or coolant used in an engine, accumulation of soot in the DPF and the associated increase in the back pressure).
∑ Changes over time or vehicle mileage (i.e., time-dependent deterioration or aging, e.g., component wear, corrosion, fatigue, component strength reduction, catalyst’s physical or chemical degradation in conversion efficiency, build-up of impurities in recycled materials such as soot in engine oil, accumulation of ash in the DPF and the associated increase in the back pressure, EGR cooler fouling).
In system design or reliability verification, several noise factors may be combined to form a worse or the worst case noise level so that the number of tests can be reduced. When noise sources exist the pieces or units that marginally meet the minimum acceptable functional performance will suffer from a loss of function that may cause a failure. For example, a borderline component with maximum fuel injection quantity may fail at an extreme ambient condition. It should be noted that for a multi-cylinder engine, the cylinder-to-cylinder variation (for example, variations in EGR distribution, peak cylinder pressure, temperature, heat flux, exhaust runner pressure and temperature) is not a noise factor. The cylinder-to-cylinder variation is a controlled variable which can be improved by design. Usually, the cylinder’s
worst parameter should be selected as the characteristic parameter to represent the whole engine in probabilistic design. Moreover, the malfunction or catastrophic failure of a component or system is usually a noise factor. For example, when the EGR valve malfunctions to fully close at rated power, all the exhaust gas flows to the turbine and the compressor may over-speed.
Although the turbocharger matching in engine system design should be targeted for normal operation rather than such a malfunctioning failure mode, the effect of such a failure mode needs to be checked as part of a reliable and robust design.
The control factors are the measures that can be changed in design to affect the mean response of the component or system to reduce variability.
The response refers to the output of the component or system in the form of force, energy, material, signal, etc. For example, the set point of VGT vane opening is a control factor, but the tolerance of the VGT opening due to turbine actuator errors is a noise factor. Changing control factors can make the system function more robust (i.e., less sensitive to the influence of noises). The error states identify the failure modes. The error states reflect the deviation of the intended function, and they are potential failure modes.
There are seven types of failure modes, namely omission of action, excessive action, incomplete action, erratic action, uneven action, action too slow, and action too fast. For example, the error states for a diesel aftertreatment system may include excessive tailpipe emissions, excessive back pressure, increased BSFC, decreased time between DPF regeneration cycles, etc.
At the end, a robustness checklist can be developed to manage the noise factors and failure modes. Each failure event is assessed by the frequency of occurrence of the cause (O), the severity of the effect (S) and the ability of detection (D, Fig. 1.9). To quantify the risks, each failure mode may have a risk priority number, calculated as the product of occurrence, severity, and detection. Design changes are usually required in order to reduce the occurrence and the severity and consequently the risk.