Research On Concrete Pump Truck’s Manual Testing Process Recognition Based On Data Driven Method

In the production process of large mechanical equipment,testing equipment is often the last step of the production process and also the most important method of quality control.However,the testing operation is tedious and complicated.It is not only necessary to operate the mechanical equipment,but also necessary to observe the appearance and operation condition of the machine.So,most mechanical equipment needs to be tested manually.It is difficult for enterprise managers to obtain the testing progress in time and determine whether workers strictly follow the operation specifications.Excessive testing or a lack of testing process may exist,which brings huge risks to enterprise operation cost control and product quality control.The development of Io T and big data technology brings opportunities for improving testing operation.Using remote real-time monitoring data to recognize the processes of equipment testing,can provide guidance to avoid the unreasonable phenomenon in testing processes.Thus,it can improve the testing efficiency and reduce the cost of testing.This paper takes concrete pump truck as the research object,and designs method to recognize the sub-steps of pumping process under high pressure and low pressure.In this paper,the recognition method of pumping process under high pressure and low pressure based on two-step recognition is proposed,and the recognition accuracy of each sub-step is between 80% and 90%.The main work and innovations of this paper are as follows:(1)The procedure of pre-processing the time series data in the case of serious data missing is constructed.Due to the complex working environments,data from the concrete pump trucks’ sensors tends to occur occasional missing.This paper proposes to use different missing processing methods to recover data as much as possible.Secondly,in order to obtain regular time interval data,this paper selects the appropriate time interval to resample the time series data.Finally,in order to make the subsequent training get better results,the data is smoothed and standardized in turn.(2)The idea of data augmentation is used to expand the original data set.In the condition of a small number of samples,the data augmentation idea of clipping inspired we expand the original data set by extending the annotated start and end points of the testing process appropriately and setting the sliding window with a small step.The number of positive samples is expanded by more than ten times,which greatly increases the number of training samples.(3)A process recognition method based on two-step recognition is proposed.The difficult recognition task is divided into two easier tasks.In the first step,machine learning classifier is used to lock the recognition interval in a small range.In the second step,rule-based method is used to accurately identify the sub-steps of pumping process under high pressure and low pressure in the locked interval.The method achieves good recognition accuracy and is expected to save more than 9.72% fuel consumption during testing process.