Mechanistic Modeling and Simulation for Process Data Generation

The paper presents a new benchmark simulator for testing advanced fault detection and diagnosis methods as well as control and optimization schemes, among other training and educational applications: BDsim. The context is the biodiesel production through the transesterification reaction of oil by methanol, in continuous mode. This process was accurately modeled and simulated, including a wide variety of upsets that may occur at different levels: process, sensors, and actuators. Each process fault is modeled in a rigorous way to make the simulator a realistic test bed for advanced data-driven algorithms. Furthermore, faulty scenarios not covered by current simulators were incorporated, such as intermittent faults and equipment degradation phenomena. The user has freedom to select the faults to simulate as well as their timing and magnitude. The code is fully compatible with both Matlab and Octave. BDsim is publicly available, registered under the GNU General Public License.