SAFETY CONFORMITY PREDICTION FOR A BOTTLING PROCESS PLANT: A MULTIPLE REGRESSION APPROACH

Safety conformity is an industrial practice to obtain enhanced safety performance and improved worker-management-government relationships. In this paper, a novel method to model and predict the conformity of bottling process operations and activities to safety rules and accident prevention is presented. Inspired by the machine guarding literature in safety compliance, this research extends the regression model beyond the machine operations domain, to cover activities in beverage testing unit (BTU), shuttle vehicle flotilla (SVF), stockroom, and suppliers. Data from practice in a bottling plant in Nigeria demonstrates the model's effectiveness. The coefficients of determination (R2 = 0.8454, 0.3891, 0.8156, 1 and 0.8156), showed the predictive competence of the warehousing, manufacturing hallway (MH), BTU, SVF, and suppliers' variables, respectively. These R2 values were derived from computations and tabulated by the software program used, and BTU and suppliers' values for R2 were obtained as the same from the program software. The relative importance of the bottling segmental factors was evaluated through ANOVA. The bottling process data results revealed the most significant variables at (p<0.05; calculated probability). This insight offers safety managers with useful practice information to plan and control.


INTRODUCTION
Scholars in recent times have paid growing attention to the analysis of accidents and their prevention (Luken et al., 2006;Kletz, 2009;Prem et al., 2010;Kidain et al., 2014). Their findings have revealed that workers in general exhibit many manipulative attempts to dislocate devices for protecting machines (for instance, effort to disable or by-pass adequate machine guards or even dismantle them (Anderson et al., 2010;Maghsoudipourand Sarfaraz, 2011;Samant et al., 2012a,b;Kica and Rosenman, 2017;Jeon et al., 2019). From the preliminary literature survey, it became evident that the literature concerning manipulative attempts by workers to displace or dismantle equipment protective devices has suffered substantial setbacks. There appears to be no useful and dependable statistics relating to the magnitude of the challenge of equipment, device or guard manipulations in organisations. However, guesses are not reliable but scientifically valid statistics promise some values for decision making and must be pursued with rigour. The restricted view of the literature limiting safety conformity to the evaluation machine guard usage should be broadened.
Few literature reports follow. Burstyn et al. (2000) employed manifold-phase Poisson and pessimistic binomial regression models to analyse the conformity results of over 29,000 administrative activities in Canada. Ortiz et al. (2000) studied warning action compliance through an appraisal of the impacts of a user's viewpoint concerning the product risk. Diaz and Resnick (2000) built up observations and rationale of a representation to envisage safety compliance for a manufacturing resource. Luken et al. (2006) reported on the motivations for removing safety devices for machinery moving parts protection and that it is poorly studied in safety research. Zin and Ismail (2011) established the compliance factors of safety from the viewpoint of employers' behaviour as employees in Malaysia. The JCGM 106 document (JCGM106:2012) established a connection between the conformity guidelines and how prediction may be attained in the context of machine guarding compliance. Griffin and Hu (2013) analysed the impacts of particular leadership behaviours on the worker's safety accomplishment. Li et al. (2013) appraised over 600 workers in crude oil production in China on safety compliance. Gressgard (2014) studied the association between information swapping scheme practice, information swapping in an establishment scheme, and compliance in safety. Ansary and Burna (2015) appraised the safety conformity for the ready-made garment industry in Bangladesh. Hu et al. (2016) adopted the technology acceptance model to build up representations for the safety conformity of workers. de la Vara et al. (2016) offered a metamodel to intervene in the safety assurance challenge in critical systems. Atamagwa (2016) exploited how environmental conformity could negatively interfere with health and safety issues at work. Washington et al. (2017) offered a Bayesian method that captures uncertainty in aviation systems conformity as a decision making scheme. Uzor and Oke (2018) studied the use of machine guards in company employees. Contract workers (external) were completely omitted from the research. Globally, the conformity literature has covered several contexts such as social psychology (Crawford et al., 2002), social influence (Cialdini and Goldstein, 2004), theoretical development (Forbes, 2006;Pendrill, 2014;Carobbi and Pennecchi, 2016), satellite remote sensing (Wildlowski, 2015), food safety (Pierna et al., 2015), product consumers (Chatterjee et al., 2017), and multi-component material (Kuselmanet al., 2017).
Despite the extensive work on conformity, the literature failed to consider the whole plant in safety conformity studies. In this paper, a multiple regression model was used to model the safety conformity of a bottling plant in terms of the five segments of SVF, BTU, suppliers, MH and stockroom. The highlights from the literature are as follows:  There is a relationship between conformity, risk, cognition, and motivation.
 The technological acceptance model has been used to formulate the conformity of workers and critical systems have been studied.
 Conceptual models of technology acceptance and metal model assure safety.
 Application areas include board ships, environmental protection  Investigations related to bottling plants are sparse and none has been done considering the company-wide parameters in major divisions of SVF, BTU, suppliers, MH and stockroom.
The principal contributions of this paper are subsequently summarized:  Formulating the conformity problem as a multiple regression model  Elaborating and developing solution steps for the contemplated problem  Implementing the described solution using practical field data.

Safety compliance process parameters and definitions
Safety conformity is measured and predicted by a number of parameters, including:  Conformityrelates to specific safety standards or behaviours laid down for workers and suppliers to a bottling plant, which directs their modes of interaction in their daily dealings Kufa Journal of Engineering, Vol. 11, No. 2, April 2020 31  Index, the level in which safety standard is related to the current attained value, indicated as a quotient, which can be weighed or measured.
 Conformity index is the appraisal of the level in which the standard of human behaviour set is strictly adhered to in order to obtain maximum efficiency and productivity.
 Stockroom is a massive storehouse where raw materials for the manufacturing process, parts of equipment and finished products are kept, pending distribution.
 MH is a large building for producing beverages in a bottling company  BTU offers a technique used to ascertain if the product of a particular batch is up to standard.
 Shuttle vehicle flotilla (SVF) represents a large hall in which vehicles used for logistics purposes in the manufacturing plant are being serviced.
 Suppliers refer to individuals or organizations that are hired to provide goods or services for the bottling company.
In the bottling process plant, conformity is influenced by a myriad of factors that may be described as major while others are considered to be minor (Uzor and Oke, 2018). The size and activity levels of the stockroom, MH, the BTU procedure, the SVF, and suppliers' activities may be regarded as major influencing factors with a direct impact on conformance performance.
On the other side, issues such as the psychological preparedness of the workers and their motivations for the day's work, the amount of budgeted expenditure and approvals for the various sections, the weather conditions in the environment of the plant such as the temperature (heat), humidity and pressure may be considered minor and indirectly impacting on the performance of the systems. However, whether considered major and directly influential or minor and indirectly impacting factors, they change in an inexplicable manner, and consequently, it is difficult to predict conformance in bottling plants with personal understanding. For this complicated prediction assignment, the multiple linear regression model is a superior alternative. With its principal advantage of authorizing numerous autonomous variables to partake in the regression modeling procedure, the multiple regression model permits us to develop a prediction representation, approximating future conformity indices by relying on the preceding ones.

Model setting
The derivation of the multiple regression model used in this work was motivated by the reality in the bottling process plant; safety conformity issue is complex. Certainly, there are numerous possible casual agents related to the safety conformity problem and accordingly, numerous factors may be required to solve this complex safety conformity problem. In the context of this paper, and based on the literature review conducted, the researchers concluded that the suitable model needed should be able to deal with the complicated instances and should possess the following attributes. First, it is desired that the model should tackle intervals and quotient level variables. It is also important that the model appraise casual connections. Third, a model that may project the future outcomes is also desirable. The conclusion of researchers from the literature review is that the multiple regression model satisfies all these conditions and therefore adopted in the current research. The multiple regression (MR) is used in this work to predict the values of safety conformity in instances of none than two variables. In the literature, Uzor

RESULTS AND DISCUSSION
The data shown in Table 2 is the percentage conformity of each factor in the warehouse segment as obtained from the field work carried out in the bottling company. Here, CF is average conformity index through all the factors in the segment.  Table 3 was obtained with the values of p checked in relationship with 0.05. This value (i.e. p(0.05)), the most commonly used limit in multiple regression, is the significant level at which one rejects the null hypothesis, and was selected randomly in this study. In principle, the 5% (below one chance in twenty chances of being incorrect) level is commonly used in literature (Uzor and Oke, 2018) and adopted in the current study. A low p-value reveals that one can reject the null hypothesis. This The interpretation of the equation is that only FD, HD, HT, CN and EB are to be the independent variables.
The value of R2 for the average conformity index in Equation (1) was used to predict the conformity index in the stockroom section of the bottling processing plant. R2 which is otherwise known as the regression coefficient value for CF, the average conformity index was 1 at first. From the second regression equation, it was seen that the regression coefficient associated with the forklift drivers is positive, which means that as the number of forklift drivers increases, the level of conformity to safety rules and precautions also increases, the reason is not far-fetched, the forklift drivers are not cumbered with too much of responsibilities, hence, have enough time to adhere to safety rules.
Notice that the Equations (1) to (9) in this work were obtained from the multiple regression where C = Coefficient and F = Factor The procedure followed for stockroom parametric prediction was applied to the MH and the results in Table 5 are obtained. Equation (3)  where the terms are as defined in Fig. 1. R2 coefficient of determination   Table 6 summaries the re-computation.
The interpretation of the equation is that only S, FO and CN are to be the independent variables. The procedure carried out previously for stockroom and MH conformity determination is extended to BTU and Table 7 summarises the essential information. Kufa Journal of Engineering, Vol. 11, No. 2, April 2020 39 Table 9.  (7) was derived and the procedure carried out for the sections is repeated here and the important results are summarised in Table 10.
Multiple regression method was adopted in order to solve the bottling plant supplier's conformity prediction problem through the execution of five independent variables, as the case may be, at a time. The multiple regression method is employed to obtain optimal results in the evaluation of values of the independent variables for which the bottling plant supplier's conformity index is either lowest or highest. In computing the polynomial to the data collected from the bottling plant, the conformity level was derived from the variables such as security, kitchen and a few others as in the case of suppliers. In the recent model development, the supplier's multiple regression model tends to be linear in behavior. Microsoft Office Excel 2013 was used to obtain the correlation among the supplier's conformity parameters such as security, kitchen and so on. The supplier's safety conformity index was then computed with the aid of the multiple linear regression model. The multiple linear regression model stimulates the association between multiple predictors and a response, where the predictions are the independent variables and the response represents the dependent variable by fitting to the observed data a linear equation. In this case, the data was obtained from a bottling plant in the south-west of Nigeria. The data is shown in Table 10.
R 2 = 1, 0.2 is the constant, 0.2 is the slope for SS, K, C1, and C2 Hypothesis: CF is influenced by S, K, C1, C2, C3 Null hypothesis: CF is not influenced by S, K, C1, C2, C3 where the following are defined R 2 coefficient of determination Equation (8) was derived and the procedure carried out for the sections repeated here and the important results are summarised in Table 11.
The interpretation of the equation is that only S, K, C1 and C2 are to be the independent variables.  From an extensive literature review, it was found that through the model by Uzor and Oke (2018) was developed in the recent past and limited to machine guarding operation in manufacturing organizations, it exhibits scope for development and application with respect for safety conformity evaluation in a process plant. As such, the insight gained from the literature review reveals that the problem of safety conformity evaluation resembles what would have been solved by the use of a multiple regression model and consequently applied. The multiple regression model was utilized to establish the significant association between each of the principal criteria, such as stockhouse, MH, SVF, BTU, and suppliers. The multiple regression models also serve the purpose of establishing the impact of these principal criteria (parameters) on the response (conformity indices) of the safety system regarding the manufacturing process.
By adopting this perception, the goal of the current research is linked to establish those factors, which are largely significant to improve the safety conformity indices of the bottling process plant such as it become clear and convincing to direct attention in the expenditure of resources to the top priority factors identified by the Taguchi-Factor scheme. An innovative characteristic of the model is that it predicts the association between the principal criteria, i.e. dependent factors) and the sub-criteria i.e. independence factors. As noted in the literature, only one study has revealed that the principal criteria of the various process components impact on the outcome of safety conformity while the association is created within them and the component parameters. Nonetheless, the perception is limited to machine guarding and the scope of application excludes external workers (suppliers) servicing the company written and outside the company premises. The scope of the study analysed (Uzor and Oke, 2018) also evades the analysis and incorporation of the SVF. As such, from a perspective, this study is novel and no such recorded previously in the literature concerning safety conformity.
The most important aspect in the predictive task is to obtain the final equations after the removal of non-significant terms from the regression model. A support for this task was declared in Uzor and Oke (2018). The final equations may be used to determine the level of safety conformity in each segment of the bottling process plant. However, the main issue is the predictive accuracy of the model. The approach to tackling this problem has been demonstrated in Uzor and Oke (2018) where the error analysis was used as the most reliable indicator of predictive accuracy.
Usually, error metrics such as the mean absolute deviation (MAD), and the mean squared error (MSE) are used. Thus, the accuracy of the predictive models formulated is discussed from the viewpoints of mean squared error (MSE) and an aspect ratio that evaluates the MSE to the average compliance in all periods. In analysing the field data, the theme that appeared was the accuracy of the predictive models. For all the segments (warehouse, MH, BTU, SVF, and suppliers), the mean squared error ranged from 0.0251 (least for suppliers) to 0.25 (highest for SVF) with an average value of 0.0554. Benchmarked against the literature (Uzor and Oke, 2018), the differences in values are over 8940. The differences between the multiple regression model of Uzor and Oke (2018) and the current research may be due to the number of variables incorporated in the models; large for the literature model (seven variables) and smaller for the current study (ranged from two to five variables). It is known that the more the number of variables considered in a work, the more the variability of the measures and the less the accuracy of the prediction from such a model may be. The aspect ratio is another emerging theme form the work. By positioning based on the aspect ratio, the order is SVF > suppliers > warehouse > BTU > MH. However, the value obtained from the literature (Uzor and Oke, 2018) in comparison with the current study is more than 34.
Sticking differences in the approach could be found in the literature (Uzor and Oke, 2018) and in the current study. While the literature considered the characteristics of the workforce, including the age of machine and operators, number of damaged and functioning guards, noncompliance index, operator's experience and material factor, the current study did not treat these workforce characteristics but compliance to stated guidelines within the group. In the literature, the coefficient of determination (R 2 ) was obtained as 0.9980, an extremely dependable value that approves the use of the model (Uzor and Oke, 2018). On the other hand, the current study reveals segmental coefficient of determination of 1 for all the analysis carried out.
In this study, the duration is spread over twelve months while noise and the physical characteristics of the equipment and operator, such as age of machine and experience of operator were not considered. However, these factors may impact on the validity of results.
Noise emerging from a functional guard may distract the operator's attention and motivates the worker to by-pass guards. Nonetheless, an extended period covering up to sixty months would have enhanced the validity of the data. In addition, the multiple regression used assures linearity among variables. However, non-linear associations may be suspected among the variables and a new model such as non-linear mathematical platform of response surface methodology may have enhanced the validity of the data also. The most significant problem in this research is the unusual commitment of the workers to work during the study period; the consciousness that their performance is monitored stimulated more commitment, giving a false impression of their performance. Perhaps an alternative approach of video recording coupled with no previous announcements of study to workers would improve the validity of the work. In a future study, it is recommended to consider optimising the variables using Taguchi method.
Notwithstanding, any conclusions relating to the safety conformity of workers would be restricted to the semi-automated plants in developing countries of the world.

Conclusions
In the current research, the safety conformity of a bottling process plant was investigated. The subsequent observations are pointed out: 1. Multiple regression analysis offers a competent and methodical approach to predict bottling process parameters for response evaluation. The technique adjusts the process such that non-significant parameters are not sensitive to the impact of the system, thus promoting a robust blueprint method 2. The outcome reveals that (a) for stockroom conformity, only FD, HD, HT, and EB are the most significant independent variables; (b) the MH shows only S, Fo and CN to be the most significant independent variables (c) the BTU conformity assessment has only LT, WT, ET and O to be the relevant and most significant variables (FT, W and Bc) to be the most significant independent variables (e) for suppliers' conformity, only S, K, C and C2