ZF Group, Germany, the world’s largest independent powertrain producer, puts each new transmission design through a demanding series of physical tests that take several weeks to complete. In the past, it took months to manually pre-process the data in order to prepare it for the fatigue analysis programs that are used to estimate life of critical components. Recently, LMS engineers used the streamline processing module within LMS CADA-X Time Data Processing Monitor (TMON) and TecWare data analysis software packages to automate almost the entire process.
The data viewing process now takes about three weeks and all other analysis can be completed in a single batch run overnight - allowing ZF to complete their testing program three months sooner than before. The flexibility of the module makes it possible for ZF engineers to easily vary the analysis parameters to meet specific individual customer specifications and comply with new requirements as they arise.
The ZF Group develops and manufactures transmissions, steering assemblies, axles and chassis components as well as complete powertrain systems for passenger cars, commercial vehicles and off-road equipment. The company also produces drive machinery for boats, railway cars and helicopters. The automobile transmission division designs and builds automatic transmissions for all classes of passenger cars with rear, front or four-wheel drive, manual transmissions for cars and light commercial vehicles and continuously-variable transmissions for cars. With 6,300 employees and sales of approximately $1.25 billion, this division has production operations in Saarbr2cken and Brandenburg, Germany, Sint-Truiden, Belgium and Batavia, Ohio, USA. The ZF Group recently developed the world’s first 6-speed automatic transmission with an extra gear in the high gear range, which lowers engine speeds and at the same time reduces noise and fuel consumption by 5 to 7%. Compared to a typical 5-speed automatic transmission, weight is reduced by 13% and acceleration increased by up to 5%.
Automobile OEMs using ZF transmissions require that certain fatigue life requirements are met. In order to prove that transmissions meet these requirements, ZF engineers install them in the customer’s vehicle along with sensors so that a wide range of signals can be measured, including in most cases rpm and torque in and out of the transmission and the pressure and temperature of the hydraulic fluid in the various torque converters. The vehicles are then driven over test tracks that represent a wide range of different driving conditions such as city, highway, off-road and mountain driving. The raw time history data acquired during these tests must then undergo a considerable amount of processing before it can be used by other engineers within the firm to evaluate the fatigue life characteristics of the transmission. In the past, this has involved a complicated manual process that took several weeks during which work on the fatigue qualification studies otherwise ground to a halt.
TMON corrects errors
The first step in this process was converting the time history data to the specific format used by the TMON software used for data analysis. TMON provides an extensive set of tools for the manipulation and interpretation of large time history data sets. An engineer then views the data in TMON in order to detect measurement errors, such as spikes or failures of a channel that might skew the results. These areas in the time history, as well as areas where the driver reported a problem, can be eliminated by cutting out time intervals. The engineers also examine the data for sensor drift and, if necessary, use the editing features in TMON to correct it. Another time-consuming manual step required in the past was using the rpm input and output tracing to determine the gear that the transmission was in at each period of the test. The traces were then divided into intervals based on certain predefined values of the gear, rpm and torque. The intervals were then ordered based on the values of these variables. The last step, which used a different LMS software package called TecWare, condensed the trace intervals so that only the most damaging cycles remain using certain accepted methods such as rainflow and time at level.
Delays of manual process
“The amount of time required by this manual data analysis caused significant delays in our product development process,” said Ralf Schmidt, Manager of the Metrology Department. “We noticed that the data analysis process required only a small amount of decision making by the engineer. For the most part, he was simply manipulating data. We wondered if this task could be automated. We spoke to our supplier of data acquisition and analysis software, LMS International. They told us that both of the software tools that we use, TMON and TecWare, have built in programming languages that make it possible to automate virtually any data analysis task. We asked LMS to send one of their engineers to work with us, learn our procedures in depth and write the routines needed to automate them.”
Bart Vandenplas, the LMS engineer assigned to the project, used the Streamline Processing (SLP) capabilities in TMON to define each task performed by ZF engineers and generate a batch file that does each of them in sequence. The input and output of these tasks and some parameters, such as sampling frequency, can be interactively changed by the user. Some of the discrete tasks, such as resample trace, were already available in SLP while others were programmed using the TMON programming language, UPA.
Data preparation routineThe TMON routine:
- sets specific attributes on multiple traces which later allows operations to be performed on all traces carrying this attribute
- performs mathematical operations such as square root and addition
- defines intervals of a trace based on various criteria and then joins them to create a new trace
- determines the gear ratio at each point in the time history and creates a new trace containing this information
- calculates a matrix of gears shifted during the past, such as how many times from first gear to second gear
- calculates various statistics such as minimum, maximum and root-mean-square of the different signals and
- identifies the time intervals during which specific hydraulic torque converters and turbines were active.
The values calculated in steps 5, 6 and 7 above are used to classify the traces in order to create fatigue predictions based on the specific requirements of the individual customers. An example is a rainflow classification of torque into the transmission during an interval in which the transmission is in third gear or a multiaxial time at level classification of the turbine rpm with the transmission in a state in which the torque converter is closed. Many similar classifications have to be made for each transmission program. For many of these classifications, only the data for a specific gear or a specific turbine state has to be processed. To meet this requirement, processing begins on the sampled data prior to classification; for example, the intervals where the transmission is in first gear are extracted from the rpm into transmission trace.
Data reduction routineTecWare is used to create the rainflow and time-at-level classifications of the input data that are required to validate the fatigue performance of the transmission. The process described above defines signal ranges and how many different bins they need to be classified into. For example, a rainflow classification of a torque with a range of 50Nm to 350Nm with 60 bins. TecWare contains a scripting language called Twbatch that runs the classification job in batch. For each classification, an input file is created that specifies the classification method, data source and traces that are to be classified, the classification ranges and the number of bands. The classification ranges and number of bands are automatically replaced with the values that were determined by SLP. This approach ensures that existing classification schemes can easily be selected by the user and that new schemes can be created without difficulty.
This automated data analysis process has dramatically reduced the amount of time required to complete the testing and validation process for custom transmissions. The only part of the process that still has to be performed manually is viewing the data and selecting valid traces. Data viewing now takes around two to three weeks and all of the other computational steps can be completed in a single night. This means that the entire process can now completed in 4 to 5 weeks instead of the 3 months that was required in the past. “The substantial time savings that we have achieved have helped us respond more quickly to our customers’ needs, bring new programs to market in less time and reduce our engineering costs,” Schmidt concluded.
