Sign In


TUT subproject 2

Utilisation of simulation data to support the maintenance of mobile work machines


Service operations have today a major role in the business of machine manufacturing companies. Many manufacturers of mobile work machines, e.g. excavators, cranes and rock crushers, are pursuing increased volume and improved performance from their service departments. Typically, these machines contain a lot of automation and control systems which provide measurement data about the operation of the machine through communication buses but may also make the localisation of failures time-consuming and the detection of evolving failures difficult.

Even though the operating characteristics and condition of an individual component could be accurately defined and modelled in laboratory conditions, reliable diagnostics is far more challenging in field conditions. There an individual component is just one of hundreds of components in a whole machine, which may operate in a constantly changing environment and loads. The relatively low cost of mobile work machines also restricts the use of sensors and measurement systems only for diagnostic purposes.

Diagnostics of mobile work machines is challenged by:

  • limited amount of sensors due to the relatively low cost of machines,
  • harsh and highly varying operating conditions,
  • large variety in the operating principles of owners and operators of machines and
  • general problematic related to data analysis and reasoning.

The use of simulation models and simulators is becoming a necessity in the development of highly automated machines. This is due to the complexity of automation system and its software. Hardware-in-the-loop (HIL) simulator systems enable the connection of simulation models with real components of machine, such as control modules and control buses. Simulation models and simulators are typically created during the early development phase of a machine. However, these are not effectively utilised in the later phases of product lifecycle.


The main goals of SIMPRO subproject TUT 2 were:

  • Develop procedure and methods for utilising simulation models and HIL simulators of mobile work machines to support the diagnostics and service of machines.
  • Develop analysis methods and implementation of statistical data analysis algorithms for recognition of machine operating states and operating condition.
  • Experiments with real machines, both undamaged and malfunctioning.

The process flow and main tasks of the developed maintenance procedure is presented in Figure 1.

Figure 1. Maintenance procedure utilising HIL simulators and analysis of time series data using the joint probability method.


For the analysis, the use of joint probability distribution is proposed. The idea is to model the behaviour of the system with probability density functions of the correlation coefficients using histograms, and test how well the future behaviour fits the model. When the correlation coefficients of the segmented multivariate data belong to sections of histograms where the probability is very low, then it is treated as rare occasion and the probability of an anomaly is high. Again, if the correlation coefficients belong to sections where the probability is high, it is treated as normal behaviour.

Benefits and use cases

The mobile work machines used in the experiments were autonomous machines, which were developed in earlier projects at IHA. The machines were GIM-machine, which is a modified version of the Avant Tecno Oy’s multipurpose wheel loader and IHA-machine, which is a modified version of Wille wheel loader (See Figure 2).

Figure 2. Autonomous GIM-machine and IHA-machine used for the testing of the developed procedure and algorithms.


The HIL simulator environment that was developed in IHA and used in the experiments is presented in Figure 3.

Figure 3. HIL simulator environment at IHA.


Figure 4 shows example of the joint probability distributions from the experiments. The results present mean values of the distributions from training phase and from tests with both undamaged and malfunctioning mobile work machines. In the case of damaged machine the fault was a jammed flushing valve in hydrostatic transmission.

Figure 4. Joint probability distributions of training and actual testing data.


The results showed clearly lower probability values for the test drives where fault was present. In these experiments the detection of anomalies for diagnostics was presented using a combination of a static threshold and a threshold based on the arithmetic mean of the joint probability distribution. This enables the detection of both single segments with low probability values indicating anomalies and also changing trends of the system. These mean in practise capability for the identification of 1) sudden critical faults and 2) slowly evolving failures.

In the presented case studies the machines were autonomous hydraulically driven mobile work machines and their operating behaviour was compared to the responses of Hardware-in-the-loop simulator. However, the utilisation of presented methods is not limited to those. The use of maintenance procedure and the analysis algorithms are applicable to many other machine systems and environments. Also the generation of simulation data does not require real time simulation or the use of any hardware components of machines as long as the simulated responses correspond sufficiently to the behaviour of real machine.

Subproject deliverables

  1. Hietala, J.-P. et al.
    Novel Procedure for Supporting Maintenance of Mobile Work Machines Using R&D Simulators
    Proceedings of the CM 2014 and MFPT 2014 conference. The 11th International Conference on Condition Monitoring and Machinery Failure Prevention Technologies, 10–12 June 2014, Manchester, UK. 9 p.
  2. Oksman, C.
    Utilization of R&D simulation models at the maintenance of mobile machines
    Bachelor's thesis, Tampere University of Technology, 2014
  3. Hietala, J.-P. et al.
    Preliminary Report on Methods, Procedures and Analysis Tools
    TUT research report, 10.1.2014., 16 p.
  4. Krogerus, T. et al.
    Anomaly Detection and Diagnostics of a Wheel Loader Using Dynamic Mathematical Model and Joint Probability Distributions
    14th Scandinavian International Conference on Fluid Power, SICFP15, May 20–22, 2015, Tampere, Finland
  5. Krogerus, T. et al.
    Joint probability distributions of correlation coefficients in the diagnostics of mobile work machines
    Journal of Mechatronics, The Science of Intelligent Machines. Elsevier publishing. In review
  6. Multanen, P. et al.
    Diagnostics of Mobile Work Machines Using Dynamic Mathematical Models and Joint Probability Distributions
    Poster, SIMPRO-SCarFace Joint Seminar 2, Espoo August 25–26, 2015