Using Metamodels of Optimal Prognosis for EMC Analysis in an Automobile

Modern cars—they’re sleeker and faster than ever before, but they’re also more electrified. It’s estimated that the modern vehicle contains up to 1,500 wires, stretching out to about a mile in length. Compare that to the 1950’s when cars had just 55 wires totaling about 50 feet in length. A new challenge now facing car manufacturers is how to handle the complexity of the cable harnesses required to connect the various subsystems.

A cable harness is a collection of electrical wires typically bound by a material such as rubber, vinyl, string, or a combination of materials. Harnessing cables provides an advantage over loose, dangling wires because the harness provides better protection against adverse effects from vibrations, abrasions, and moisture.

With wires in close quarters inside the harness, it is important to measure the electromagnetic (EM) quality of the cable. A quantitative parameter to estimate the EM quality of a specific cable layout is the crosstalk noise, or the influence on low voltage cables by the high voltage cables that are in proximity.

To determine the EM quality, a cable needs to be cut into cross sections. When a cable is measured with just one cross-section, it may test as not having any problems. In reality there is the possibility of wires moving while driving. As a consequence, multiple different cross-sections need to be tested to diagnose potential electromagnetic interference and compatibility (EMI/EMC) issues. Metamodeling is a technique that can be applied to the design, optimization, and assessment of harnesses. The methodology can be used for automotive purposes, as well as other industries including aerospace and defense.

Analyzing Optimization

Optimization and robustness analysis has become an important method in the development of industrial products. A framework for numerical optimization can be defined by specifying the design criteria as objectives and constraints. However, one challenge with this approach is efficiently handling the optimization function given all the possible combinations of multiple design parameters. Due to this, a degree of nonlinearity has to be considered in the optimization process.

Numerical modeling can simplify the process to evaluate and minimize cross talk, ensuring high accuracy of the virtual representation of the cable harness and reducing the solution time.  A simulation tool can be used to understand which cable parameters (length and location) have the greatest influence. Simulation can also determine the probability that a given cable’s configuration falls below a certain threshold of crosstalk to help synthesize efficient form-factor layouts. By using Ansys Discovery, Ansys EMC Plus, and Ansys optiSLang together, it is possible to understand the impact of the uncertainty in the cross-section of the cable based on a metamodel of optimal prognosis (MOP).

A Cable Crosstalk Experiment

To start, a Gaussian distribution, also known as a normal distribution, is used to find the cross sections. This method is a bell-shaped curve that assumes any measurement values will follow a normal distribution with an equal number of measurements above and below the mean. Figure 1 shows the variation in the cross-section of the wires at different locations along the cable.

“Initially, they are looking the same, then we are using a Gaussian distribution to kind of have the cable crossover and we are splitting it into multiple segments for that one particular wire,” EMA Senior RF Engineer Prasanna Padmanabhan said.

This is done for each wire in the cable.

Fig. 1. Cross-section algorithm

The implementation and the outcome of the cross-section algorithm are seen in Figure 1 and 2. Figure 2 shows the variation in the cross-section of the wires at different locations along the cable.

“So now (in Figure 2) you can see each of the wires within the cable. They are color-coded, and you can see how the wires are getting twisted from one end to the other,” Padmanabhan said. “All these individual lines or cross sections that you see, they’re all the individual segments.”

Fig. 2. Cross-section algorithm showing the positioning of each wire

Numerical Models

EMC Plus is a finite-difference time-domain (FDTD) solver that can be used to model and simulate cables and transmission lines to identify EMI and EMC problems. Figure 3 shows a cable in EMC Plus with a source box and load box along with a cable cross-section.

“All these cables have the same continuity when you start off because they have the same cross section as you go from the source all the way to the load,” Padmanabhan said. “But in reality, what happens is the cable, they cross over. So, the cross-section changes from one point to the other point from the start to the end.”

The dots in Figure 3 represent the beginning or the end of an individual cross-section of the segments on the cable and also the wire distribution inside of the cable.

Fig. 3. Numerical model in Ansys EMC Plus

Cable Optimization Workflow

To optimize the cable, a parameterization needs to be created for each cross section in each segment. To do this, an application programming interface (API) class was created to automatically edit the segments’ definition in the input XML file without the user having to manually edit the configuration file.

In Discovery, users open EMC Plus and call the API to specify the X and Y locations, as well as, the segment number, cable number, harness number, and other parameters associated with it.

“Now the user controls how many combinations they want to be able to simulate,” Padmanabhan said. “They control how the wires crossover within each segment going from point A to point B.”

At every controlled location, a Gaussian and Weibull distribution is applied to discretize the interpolated segments of the wire. optiSLang reads the EMC Plus project and analyzes its input configuration file. The software then creates all the parameterizations for the segments, cable, and harness. optiSLang will push through one value at a time for each segment until all combinations have been completed.

“What we can do when we create this many variations is that we can determine what is the worst-case crosstalk,” Padmanabhan said.

Results computed by the voltage probe are defined at the source (left) and load (right) for all the combinations and are shown in Figure 4.

Fig. 4. Probe on wire 4 at the source side (left) and the load side (right)

The histogram response of the probes shows triangular and Gumbel probability distribution functions at the source and load respectively. We get expected results on the source side as the input is a combination of Gaussian and Weibull distributions. Figure 5 shows a Gumbel distribution on the load side. It typically means extreme deviation in the results and is indicative of a potential failure or numerical noise.

Fig. 5. Probability of Failure Analysis

As a final step, optiSLang can calculate the impact of uncertainty at the source and load based on all the designs generated. Figure 6 shows the overall forecast quality by aggregating the effect of discretizing the segments at various locations along the cable.

Fig. 6. Metamodel of optimal prognosis

Conclusion

The complexity of harness layout in a modern vehicle requires the use of efficient numerical methods to calculate quantities that describe the quality of the layout.

Using EMC Plus along with optiSLang makes the computation of cross-talk on a cable with non-uniform cross-sections an easy task. A new API class was developed to control the cross-section distribution in EMC Plus. The user just needs to define local X and Y coordinates in the cross-section. The variation of the cross-section along the cable can be controlled by optiSLang. The logic of the variation can be defined in many ways, in this example, controlled splines distribution in predefined cross-sections driven by a script.

The API can be used to create an unlimited number of combinations for however many sections that the cable needs to be broken into, how many wires are in the cable and how they crossover. This allows the user to find the worst-case EMC scenario and optimal design pattern.