Predictive Maintenance of a Heat Exchanger - MATLAB & Simulink
Video Player is loading.
Current Time 0:00
Duration 14:06
Loaded: 0.00%
Stream Type LIVE
Remaining Time 14:06
 
1x
  • Chapters
  • descriptions off, selected
  • captions off, selected
      Video length is 14:06

      Predictive Maintenance of a Heat Exchanger

      By implementing a predictive maintenance program on a heat exchanger, process engineers can identify when to modify operations to extend heat exchanger life versus when to take the heat exchanger offline for cleaning. In this session, you will learn how you can use MATLAB® and Simulink® to aid in fouling monitoring and prediction by: 

      • Building a rigorous first principles model of the heat exchanger with Simscape™
      • Building a digital twin of the heat exchanger by tuning the parameters of the model to match field data with Simulink Design Optimization™
      • Generating synthetic data from the digital twin to simulate heat exchanger fouling
      • Modeling an exponential degradation process for estimating the remaining useful life (RUL) of the heat exchanger with Predictive Maintenance Toolbox™

      Published: 21 Nov 2021

      Hello, my name is Inho Kim, Sydney Application Engineer and MathWorks. Today, we will be talking about predictive maintenance of a heat exchanger using a Simscape simulation model. This project is a collaboration with one of our customers.

      The project is not quite finalized yet, so this talk will mostly touch on generic workflow and background of approach better than details. Without further ado, let's get started. The motivation of this project is to estimate heat exchanger degradation level, mainly focusing on fouling. The TEMA, Tubular Exchanger Manufacturers Association, standard health exchanges are quite common in downstream refinery and petrochemical plants.

      Depending on the complexity of process, there are over 30 heat exchangers operating preheating or cooling process fluids. Petrochemical plants usually have 200 to 350 heat exchangers operating according to hydrocarbon processing magazine. Most common types of heat exchangers in refineries and petrochemical plants are shell and tube type ones as you can see in the picture.

      Heat exchanger filings is a commonly occurring problem in heat exchangers, resulting in changing the heat transfer surface and reducing the overall heat transfer rate through that surface. During fouling, the surface of heat exchanger will develop another layer of solid material. This can happen for a variety of reasons, but as a result, the heat transfer coefficient at the surface is drastically reduced since the heat conducting while metal is not in contact with the fluids anymore.

      Instead, the wall is separated from fluid by a layer of fouling. Following material prevents efficient heat transfer and reduces the efficiency of heat exchanger. To prevent degradation of the heat exchanger efficiency, maintenances are scheduled years ahead, but the accuracy of commonly used implicit equations is not enough to estimate right timing of maintenance, which may lead to unnecessary cost of maintenance or late maintenance.

      Before we jump in Simscape modeling, let's go through a heat exchanger model with effectiveness number of transfer units method. The first equation is straightforward. Q1 and Q2 are the heat transfer rates into fluid 1 and fluid 2. The next terms are quite important for our modeling. Epsilon is the effectiveness parameter, which is a dimensionless number between 0 and 1.

      The Q max is the maximum possible heat transfer rate between fluid 1 and fluid 2 at a given set of operating conditions. Max can be calculated with the minimum value of the thermal capacity rate. Cmin and the difference of inlet temperatures of fluid 1 and fluid 2. The Cmin can be determined between mass flow rates and specific heat coefficients of fluid 1 and fluid 2.

      As you can see, temperatures and mess flow rates are from operating conditions which we need to come back, again, after checking effectiveness thumb. The shell and tube type hit exchangers with one shell path effectiveness analytical expression is given in the reference. Although it seems quite complex, let's break it down each time since there are only two variables. The thermal capacity ratio, Crel and the number of transfer units, NTU.

      The thermal capacity ratio is simply the minimum value of the thermal capacity rate over the maximum value of the thermal capacity rate, which roughly represents the ratio of mass flow rates. The number of transfer units is inverse of the minimum value of the thermal capacity rate times resistance between two fluids. This resistance can be further brought down to five times as shown in this cartoon.

      Heat transfer coefficient an area for both sides. Resistance from fouling for both sides, and resistance of wall. This breakdown gives us important idea of how we can set up the model with different conditions and scenarios. This red variables are depending on operating conditions and cases with a short time process, and this green variables are depending on the position of filing layers, which takes much longer times than red variables.

      Other than these variables, for example, the specific heat coefficients, the heat transfer coefficients, and the area are all constants in our model. This is an example Simscape model using thermal liquid network. The real systems can be much more complex with bypass lines, series of heat exchangers, and valves, but this model can give an idea how the thermal liquid network model can be set up.

      This part is the effectiveness and heat exchanger block for thermal liquid to thermal liquid. Simscape also offers different combinations of different network domains, such as gas to gas, gas to thermal liquid, and thermal liquid to moist air. Each side of liquid goes into heat exchanger defined with a thermal liquid property blocks, and the mass flow rate and temperature of liquids are defined with control mass flow rate source and controlled vegetable block, respectively.

      Temperature of both the kits are locked with this four sensors during simulations. Although we carefully went to the analytical expressions in previous slide, the effectiveness NTU heat exchanger block contains all necessary equations and parameters mask so users don't have to make equations to build algebraic loops for system equations. The operating condition related variables are carefully selected from field data, fed into the model as input signal using signal editors. Following factor in one of heat exchanger block parameter can be set as parameters to be estimated in the next step parameter estimation.

      The parameter estimation app is part of Simulink design optimization. Parameter estimation uses experimental data, which often comes from field data to back calculate using optimization techniques. This step often called constructing a digital twin since the model parameters are updated based on field data. The experiment data is from a subset of field data, which is under controlled operating conditions.

      Temperatures and mass forests of fluid 1 and fluid 2 at inlet of heat exchanger. This step is iterated over multiple subsets of filled data, which have similar controlled operating conditions to find the correlation between filing factor and efficiency of heat transfer. Historical data showed a clear degradation of heat exchanger rate over a long period of time.

      The range of following factor can be correlated with the degradation of efficiency of heat transfer. Here is a table of initial filing resistance depending on fluids. This can be used to determine the initial value and bounds for better estimations using optimization in parameter estimation. We also know that the operating conditions affect the level of effectiveness come from the analytical expressions of heat transfer.

      Different condition sets were selected to correct following factor estimations. Once we estimate the range of following factors to correlate them with heat exchange efficiency, we can perform four simulations to generate the degradation map to build a similarity-based remaining useful life model. By sweeping the following factors versus different time spans to vary the speed of efficiency degradation in linear degradation scenario, the baseline degradation remaining useful life cases can be generated.

      There are additional baselines can also be constructed by varying operating conditions as model inputs. As you can see, additional dimension of baselines are prepared too. At this point, if there is any measured filled data, including operating conditions at heat exchanger inlet and outlet temperatures can be used to estimate following factors with parameter estimation, which represent health indicator in this case to determine where the heat exchanger status is currently located in pre-generated degradation map.

      The current state shows as or a line of that in case the estimation was repeated over time, which follows one of linear degradation line or switching between operating condition maps. If there is any threshold for health indicator, the remaining useful life can be estimated between time difference of current status and intersection of degradation line and threshold. In summary, this presentation discussed about heat exchanger efficiency degradation with fouling.

      With analytical expression of transfer equations, constants and different types of variables were identified for using them in modeling and simulations. Simscape thermal liquid network and blocks were used to model heat exchanger network, then ranges of following factors were estimated using parameter estimation under different time span degradation and operating conditions. All these following factors were known, multi-dimensional degradation maps were generated.

      And finally, identified current degradation level and remaining useful life by back calculating filing factors using parameter estimation and then remap to figure out what is the current status of heat exchanger. Thank you for listening. Please let me know if you have any questions.

      View more related videos