The task facing calibrators is not getting any easier. “The complexity of calibration exploded exponentially when the Tier 4 emissions regulations were introduced,” explained Paolo Di Martino, Business Unit Director for Diesel Systems for IAV North America. “One reason is because each customer decided to solve the Tier 4 problem with their own unique solution. There is no ‘mainstream option,’ but there are multiple solutions, in particular with aftertreatment.” Aftertreatment devices they could use for emissions control include diesel oxidation catalysts (DOC), diesel particulate filters (DPF), and/or selective catalytic reduction (SCR).
Some manufacturers use all three control devices, while others use combinations of some of them. “Each one of these devices is in their own way a prima donna,” Di Martino said. “Each wants their own level of care. Each needs a certain set of incentives to do their job” and need to be calibrated.
Engines that are more efficient also use exhaust gas turbocharging, exhaust gas recirculation (EGR), and complex common-rail fuel injection to improve the quality of combustion. Not only must engineers calibrate these devices, they also need to specify corrections to compensate for aging. Because of these trends, the number of variables a calibration has to account for has increased exponentially from a few thousand to 30,000 to 50,000 within the last decade. The number of engine variants makes the job tougher because there is no single "reference" model to start with.
The state-of-the-art methodology for base engine calibration today is to use design of experiments (DoE) to create a statistical engine model, called a response surface model that is fitted from a sparse set of measurements.
“Through DoE, we can still measure a reasonable amount of data in a reasonable amount of time to generate the calibration,” said Di Martino. The DoE method lets engineers test fewer response points while varying inputs like speed, load, and start of injection timing. The resulting model is also reusable. It can be modified for new applications or new duty cycles without having to conduct new measurements.
To get the best model requires careful planning of the test program ensuring response gradients are captured and all operating ranges covered. Easily creating that model from the measurement requires a set of tools. To do both, IAV created the EasyDoE software package, making the process accessible and simple yet retaining specific "expert" settings, according to the company. An engine’s input parameters and variance limits are the basis for guiding a user through the process of planning the measurements and developing models.
Focus on transients, variety
You can think of the calibration process in three parts. First, optimizing an engine to create a base engine calibration at its various steady-state operating points, typically on a dynamometer test rig. While that might be thought of as "traditional" calibration, there is now a second stage that focuses on transient phases, ensuring that emissions are controlled when an off-highway machine surges to meet sudden demands.
“That puff of black smoke one used to see when the load changes does not occur anymore,” said Di Martino. “This transient calibration is more important than it was even 5 or 10 years ago,” with IAV upgrading its tool set and procedures to meet the demand (like many others.) The final, third step in calibration is a vehicle level dynamic calibration, usually through a mix of dyno and vehicle testing, with again a newer focus on transient performance.
Another important factor unique to the off-highway industry is the sheer variety of applications and duty cycles an individual engine design could fit into, according to Michael Franke, Director of Light-Duty Diesel and Commercial Engines at FEV. The company engages in engine development programs for agriculture, power generation, marine, construction, and railroad applications.
“They also operate in a large variety of duty cycles and under extreme environmental conditions,” said Franke. “For example, an agricultural engine can operate under low-load profiles when spraying fields or other times at very high-loads when cultivating soil.”
Post Tier 4 final—reducing cost
Now that all the major OEMs in off-highway have engines that meet Tier 4 Final regulations, what Franke sees on their agenda for many of their customers is lowering product cost, especially for engines below 120 kW. For those engines, “reducing product cost means eliminating DPF and EGR and relying solely on SCR for NOx control,” he said. This would save on both engine hardware and operating cost. “These engines would need de-NOx systems with conversions efficiencies greater than 90 to 93%,” he said.
That level of efficiency will have an impact on the control system. “It will have to be very accurate in terms of urea mixing and ammonia control,” he explained. He also sees engine providers and machine OEMs using more electrification, hybridization, and continuously variable transmissions in off-highway applications. This will make the calibrator’s job even more challenging.
FEV provides both control system strategies as well as calibration services to their off-highway customers. Franke likes to stress the open-looped control nature of model-based systems, using predictive algorithms rather than closed-loop with feedback from sensors. “Traditional closed-loop control systems are time consuming to develop and cost-intensive to use,” he said. They can also simply be too slow to be practical.
He also notes that model-based control algorithms can also be transferred to different applications with minimal recalibration effort. “What open-loop does is predicts output, say NOx emissions, depending on the operating load point and other conditions,” he said.
When controlling an SCR for NOx, the open-loop algorithm will control the amount of urea it feeds to the device depending on the modeled NOx prediction for the actual load and environmental condition. “This is a much faster response than a closed-loop system can achieve,” he said, and applies to other engine components as well.
FEV offers calibration engineers its TOPexpert Suite of tools for calibration, test planning, model-based engine analysis, and optimization. In addition, illustrating how important CAE simulated data is for front-loading the development cycle, Franke also noted that FEV introduced DoE methodologies to its CFD combustion analysis software, reducing time from weeks to days to understand how fuel injection, piston geometry, or air motion affects combustion and engine-out emissions.
Another important new tool for calibrators is FEV’s Micro-HiL tool, a virtual engine model connected to a real time processing ECU. “It is a desktop development environment that integrates with a simulated engine model,” he said, which is typically GT-POWER from Gamma Technologies.
Multiple variables, expanded ECUs
Simulink and MATLAB from MathWorks are tools frequently used to create control systems and control systems simulations. The commonly used differential equation simulator Simulink is a tool to design the controllers, estimators, and diagnostic components of an engine controller, according to Pete Maloney, Senior Principal Consulting Engineer for MathWorks. Another tool, Simulink Design Optimization, lets engineers numerically optimize parameters that affect dynamic control performance in the controllers, estimators, and diagnostics.
MathWorks also provides help to calibration engineers through the Model-Based Calibration Toolbox, which provides tools for optimally calibrating complex powertrain systems using statistical modeling (response surface modeling) and numerical optimization to develop base engine calibrations. The result is a statistical response surface engine model, which provides the engineer with a means of analytically developing a base engine calibration that will meet performance and emissions while minimizing fuel consumption. Their Model-Based Calibration Toolbox also provides a DoE planning tool.
“We also extend or optimize our core functionality to solve specific customer problems as a consultant,” said Maloney. His work as a consultant provides insight into another emerging customer requirement: Multiple input and multiple output control, and in particular model predictive control (MPC). “Our customers are asking for coordinated control of multiple variables and coordinating the actions of multiple actuators has to improve,” he said.
“We also think that MPC plays well with the emergence of multi-core ECUs,” he explained. “MPC will require more ECU power to accomplish, but fortunately that is coming along. We can see one core of an ECU used for, say, the air system, while another core will be dedicated to the fuel system, and aftertreatment devices on a third core.” It is a question of timing, but he thinks multiple core ECUs and routine implementation of MPC is inevitable.
Development cycles shortening
If the regulatory pressures increasing the complexity of off-highway engines were not enough, everyone interviewed for this article agreed that development cycles are getting shorter due to competitive pressures. “Not only is time to market shortening, but as always, the industry needs to continue to lower cost and increase quality,” said Talus Park, Skill Team Leader in Calibration & Certification for AVL Powertrain Engineering. “The industry is also facing a shortage of test vehicles, engines, dyno facilities, and trained calibration engineers.”
He sees another advantage of model-based development. “It allows us to decouple the test phase, on a dyno or in a vehicle, from the calibration phase, which now can be done in an office,” explained Park. Using a combination of measurement data and simulated data relieves some of the pressure on facilities and trained test personnel. AVL offers a number of software tools for modeling, testing, and calibration development, including the CRUISE-M and VSM vehicle level simulation modeling software, PUMA and CAMEO testing tools, and AVL-DRIVE for objective drivability assessment.
The company also recognizes that there are a number of tools in use besides their own. “We can’t expect everyone to convert to using AVL tools,” said Park. Instead, AVL concentrated on process. The result was the Integrated Open Development Platform (IODP).
Park describes this as enabling integration between tools in various development environments. “There is a movement in the industry for more standardization,” he said, pointing to the functional mock-up interface (FMI) standard as an important enabler for the IODP. He stresses the purpose is early evaluation of hardware while developing controls and calibration.
The IODP process starts in the virtual world. It progressively pursues development of hardware, controls, and calibration first on a desktop using model-in-the-loop and software-in-the-loop methods. It then advances to hardware-in-the-loop simulation and development. Finally, the system is validated through field-testing the equipment.
While FMI allows mixing simulation models from different tool providers, Park also stressed the need for simulations that are fast enough to run in real-time, which enables the mixing of models with real hardware. One such tool from AVL, CRUISE-M with embedded MoBEO models, meets this need for real-time calculations. MoBEO is a real-time semi-physical diesel engine and aftertreatment model and CRUISE-M is a real-time crank-angle resolved physical engine model for diesel and gasoline engines.
“Using these we can start calibration work before we have any actual engine hardware available,” Park said, also noting that with today’s fast, multi-core desktop computers even high fidelity models with dozens of degrees of freedom can run in real-time.